diff -uprN binutils-2.11.90.0.23/bfd/ChangeLog binutils-2.11.90.0.24/bfd/ChangeLog --- binutils-2.11.90.0.23/bfd/ChangeLog Fri Jul 6 09:26:10 2001 +++ binutils-2.11.90.0.24/bfd/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,40 @@ +2001-07-11 H.J. Lu + + * elf64-alpha.c (elf64_alpha_check_relocs): Set the ALLOC|LOAD + flags when creating the reloc section if the ALLOC flag in the + source section is set. + +2001-07-11 Steve Ellcey + + * reloc.c (bfd_reloc_code_type): Add IA64 relocs + BFD_RELOC_IA64_LTOFF_FPTR32MSB and BFD_RELOC_IA64_LTOFF_FPTR32LSB + * bfd-in2.h: Regenerate. + * libbfd.h: Regenerate. + * elfxx-ia64.c (ia64_howto_table): Define how to handle + new relocations. + (elfNN_ia64_reloc_type_lookup): Handle new relocations. + (elfNN_ia64_check_relocs): Likewise. + (elfNN_ia64_install_value): Likewise. + (elfNN_ia64_relocate_section): Likewise. + +2001-07-11 Jakub Jelinek + + * elf32-sparc.c (_bfd_sparc_elf_howto_table): Remove support for + R_SPARC_UA64. + (elf32_sparc_check_relocs): Likewise. + Only create .rela section for alloced sections in shared libraries. + (elf32_sparc_relocate_section): Likewise. + Remove redundant check. + Optimize unaligned reloc usage. + * elf64-sparc.c (sparc64_elf_relocate_section): Remove redundant + check. + +2001-07-11 H.J. Lu + + * elf64-alpha.c (elf64_alpha_check_relocs): Only use the same + ALLOC|LOAD flags as the source section for debugging sections + when creating the reloc section. + 2001-07-05 Jim Wilson * linker.c (_bfd_generic_link_add_one_symbol, case BIG): Use diff -uprN binutils-2.11.90.0.23/bfd/bfd-in2.h binutils-2.11.90.0.24/bfd/bfd-in2.h --- binutils-2.11.90.0.23/bfd/bfd-in2.h Tue Jun 19 11:57:32 2001 +++ binutils-2.11.90.0.24/bfd/bfd-in2.h Sat Jul 14 10:29:15 2001 @@ -2672,6 +2672,8 @@ this offset in the reloc's section offse BFD_RELOC_IA64_PCREL64LSB, BFD_RELOC_IA64_LTOFF_FPTR22, BFD_RELOC_IA64_LTOFF_FPTR64I, + BFD_RELOC_IA64_LTOFF_FPTR32MSB, + BFD_RELOC_IA64_LTOFF_FPTR32LSB, BFD_RELOC_IA64_LTOFF_FPTR64MSB, BFD_RELOC_IA64_LTOFF_FPTR64LSB, BFD_RELOC_IA64_SEGREL32MSB, diff -uprN binutils-2.11.90.0.23/bfd/configure binutils-2.11.90.0.24/bfd/configure --- binutils-2.11.90.0.23/bfd/configure Fri Jul 6 09:26:10 2001 +++ binutils-2.11.90.0.24/bfd/configure Sat Jul 14 10:29:15 2001 @@ -1116,7 +1116,7 @@ fi PACKAGE=bfd -VERSION=2.11.90.0.23 +VERSION=2.11.90.0.24 if test "`cd $srcdir && pwd`" != "`pwd`" && test -f $srcdir/config.status; then { echo "configure: error: source directory already configured; run "make distclean" there first" 1>&2; exit 1; } diff -uprN binutils-2.11.90.0.23/bfd/configure.in binutils-2.11.90.0.24/bfd/configure.in --- binutils-2.11.90.0.23/bfd/configure.in Fri Jul 6 09:26:10 2001 +++ binutils-2.11.90.0.24/bfd/configure.in Sat Jul 14 10:29:15 2001 @@ -7,7 +7,7 @@ AC_INIT(libbfd.c) AC_CANONICAL_SYSTEM AC_ISC_POSIX -AM_INIT_AUTOMAKE(bfd, 2.11.90.0.23) +AM_INIT_AUTOMAKE(bfd, 2.11.90.0.24) dnl These must be called before AM_PROG_LIBTOOL, because it may want dnl to call AC_CHECK_PROG. diff -uprN binutils-2.11.90.0.23/bfd/elf32-sparc.c binutils-2.11.90.0.24/bfd/elf32-sparc.c --- binutils-2.11.90.0.23/bfd/elf32-sparc.c Fri Jun 1 21:47:13 2001 +++ binutils-2.11.90.0.24/bfd/elf32-sparc.c Sat Jul 14 10:29:15 2001 @@ -120,7 +120,7 @@ reloc_howto_type _bfd_sparc_elf_howto_ta HOWTO(R_SPARC_NONE, 0,0, 0,false,0,complain_overflow_dont, bfd_elf_generic_reloc, "R_SPARC_NONE", false,0,0x00000000,true), HOWTO(R_SPARC_NONE, 0,0, 0,false,0,complain_overflow_dont, bfd_elf_generic_reloc, "R_SPARC_NONE", false,0,0x00000000,true), HOWTO(R_SPARC_NONE, 0,0, 0,false,0,complain_overflow_dont, bfd_elf_generic_reloc, "R_SPARC_NONE", false,0,0x00000000,true), - HOWTO(R_SPARC_UA64, 0,4,64,false,0,complain_overflow_bitfield,bfd_elf_generic_reloc, "R_SPARC_UA64", false,0,(~ (bfd_vma)0), true), + HOWTO(R_SPARC_UA64, 0,0, 0,false,0,complain_overflow_dont, sparc_elf_notsupported_reloc, "R_SPARC_UA64", false,0,0x00000000,true), HOWTO(R_SPARC_UA16, 0,1,16,false,0,complain_overflow_bitfield,bfd_elf_generic_reloc, "R_SPARC_UA16", false,0,0x0000ffff,true), HOWTO(R_SPARC_REV32, 0,2,32,false,0,complain_overflow_bitfield,bfd_elf_generic_reloc, "R_SPARC_REV32", false,0,0xffffffff,true), }; @@ -554,11 +554,10 @@ elf32_sparc_check_relocs (abfd, info, se case R_SPARC_LO10: case R_SPARC_UA16: case R_SPARC_UA32: - case R_SPARC_UA64: if (h != NULL) h->elf_link_hash_flags |= ELF_LINK_NON_GOT_REF; - if (info->shared) + if (info->shared && (sec->flags & SEC_ALLOC)) { /* When creating a shared object, we must copy these relocs into the output file. We create a reloc @@ -1223,7 +1222,6 @@ elf32_sparc_relocate_section (output_bfd || r_type == R_SPARC_LO10 || r_type == R_SPARC_UA16 || r_type == R_SPARC_UA32 - || r_type == R_SPARC_UA64 || ((r_type == R_SPARC_PC10 || r_type == R_SPARC_PC22) && strcmp (h->root.root.string, @@ -1412,8 +1410,7 @@ elf32_sparc_relocate_section (output_bfd case R_SPARC_LO10: case R_SPARC_UA16: case R_SPARC_UA32: - case R_SPARC_UA64: - if (info->shared) + if (info->shared && (input_section->flags & SEC_ALLOC)) { Elf_Internal_Rela outrel; boolean skip; @@ -1463,6 +1460,28 @@ elf32_sparc_relocate_section (output_bfd outrel.r_offset += (input_section->output_section->vma + input_section->output_offset); + /* Optimize unaligned reloc usage now that we know where + it finally resides. */ + switch (r_type) + { + case R_SPARC_16: + if (outrel.r_offset & 1) + r_type = R_SPARC_UA16; + break; + case R_SPARC_UA16: + if (!(outrel.r_offset & 1)) + r_type = R_SPARC_16; + break; + case R_SPARC_32: + if (outrel.r_offset & 3) + r_type = R_SPARC_UA32; + break; + case R_SPARC_UA32: + if (!(outrel.r_offset & 3)) + r_type = R_SPARC_32; + break; + } + if (skip) memset (&outrel, 0, sizeof outrel); /* h->dynindx may be -1 if the symbol was marked to @@ -1535,12 +1554,8 @@ elf32_sparc_relocate_section (output_bfd ++sreloc->reloc_count; /* This reloc will be computed at runtime, so there's no - need to do anything now, unless this is a RELATIVE - reloc in an unallocated section. */ - if (skip - || (input_section->flags & SEC_ALLOC) != 0 - || ELF32_R_TYPE (outrel.r_info) != R_SPARC_RELATIVE) - continue; + need to do anything now. */ + continue; } break; @@ -2024,7 +2039,7 @@ elf32_sparc_object_p (abfd) } else if (elf_elfheader (abfd)->e_flags & EF_SPARC_LEDATA) return bfd_default_set_arch_mach (abfd, bfd_arch_sparc, - bfd_mach_sparc_sparclite_le); + bfd_mach_sparc_sparclite_le); else return bfd_default_set_arch_mach (abfd, bfd_arch_sparc, bfd_mach_sparc); } diff -uprN binutils-2.11.90.0.23/bfd/elf64-alpha.c binutils-2.11.90.0.24/bfd/elf64-alpha.c --- binutils-2.11.90.0.23/bfd/elf64-alpha.c Wed May 23 11:36:03 2001 +++ binutils-2.11.90.0.24/bfd/elf64-alpha.c Sat Jul 14 10:29:15 2001 @@ -2589,8 +2589,10 @@ elf64_alpha_check_relocs (abfd, info, se sreloc = bfd_make_section (dynobj, rel_sec_name); if (sreloc == NULL || !bfd_set_section_flags (dynobj, sreloc, - ((sec->flags & (SEC_ALLOC - | SEC_LOAD)) + (((sec->flags + & SEC_ALLOC) + ? (SEC_ALLOC + | SEC_LOAD) : 0) | SEC_HAS_CONTENTS | SEC_IN_MEMORY | SEC_LINKER_CREATED diff -uprN binutils-2.11.90.0.23/bfd/elf64-sparc.c binutils-2.11.90.0.24/bfd/elf64-sparc.c --- binutils-2.11.90.0.23/bfd/elf64-sparc.c Fri Jun 1 21:47:13 2001 +++ binutils-2.11.90.0.24/bfd/elf64-sparc.c Sat Jul 14 10:29:15 2001 @@ -2283,12 +2283,8 @@ sparc64_elf_relocate_section (output_bfd ++sreloc->reloc_count; /* This reloc will be computed at runtime, so there's no - need to do anything now, unless this is a RELATIVE - reloc in an unallocated section. */ - if (skip - || (input_section->flags & SEC_ALLOC) != 0 - || ELF64_R_TYPE_ID (outrel.r_info) != R_SPARC_RELATIVE) - continue; + need to do anything now. */ + continue; } break; } diff -uprN binutils-2.11.90.0.23/bfd/elfxx-ia64.c binutils-2.11.90.0.24/bfd/elfxx-ia64.c --- binutils-2.11.90.0.23/bfd/elfxx-ia64.c Fri May 25 12:17:56 2001 +++ binutils-2.11.90.0.24/bfd/elfxx-ia64.c Sat Jul 14 10:29:15 2001 @@ -360,6 +360,8 @@ static reloc_howto_type ia64_howto_table IA64_HOWTO (R_IA64_LTOFF_FPTR22, "LTOFF_FPTR22", 0, false, true), IA64_HOWTO (R_IA64_LTOFF_FPTR64I, "LTOFF_FPTR64I", 0, false, true), + IA64_HOWTO (R_IA64_LTOFF_FPTR32MSB, "LTOFF_FPTR32MSB", 2, false, true), + IA64_HOWTO (R_IA64_LTOFF_FPTR32LSB, "LTOFF_FPTR32LSB", 2, false, true), IA64_HOWTO (R_IA64_LTOFF_FPTR64MSB, "LTOFF_FPTR64MSB", 4, false, true), IA64_HOWTO (R_IA64_LTOFF_FPTR64LSB, "LTOFF_FPTR64LSB", 4, false, true), @@ -480,6 +482,8 @@ elfNN_ia64_reloc_type_lookup (abfd, bfd_ case BFD_RELOC_IA64_LTOFF_FPTR22: rtype = R_IA64_LTOFF_FPTR22; break; case BFD_RELOC_IA64_LTOFF_FPTR64I: rtype = R_IA64_LTOFF_FPTR64I; break; + case BFD_RELOC_IA64_LTOFF_FPTR32MSB: rtype = R_IA64_LTOFF_FPTR32MSB; break; + case BFD_RELOC_IA64_LTOFF_FPTR32LSB: rtype = R_IA64_LTOFF_FPTR32LSB; break; case BFD_RELOC_IA64_LTOFF_FPTR64MSB: rtype = R_IA64_LTOFF_FPTR64MSB; break; case BFD_RELOC_IA64_LTOFF_FPTR64LSB: rtype = R_IA64_LTOFF_FPTR64LSB; break; @@ -2007,6 +2011,8 @@ elfNN_ia64_check_relocs (abfd, info, sec case R_IA64_LTOFF_FPTR22: case R_IA64_LTOFF_FPTR64I: + case R_IA64_LTOFF_FPTR32MSB: + case R_IA64_LTOFF_FPTR32LSB: case R_IA64_LTOFF_FPTR64MSB: case R_IA64_LTOFF_FPTR64LSB: need_entry = NEED_FPTR | NEED_GOT | NEED_LTOFF_FPTR; @@ -2814,6 +2820,7 @@ elfNN_ia64_install_value (abfd, hit_addr case R_IA64_GPREL32MSB: case R_IA64_FPTR32MSB: case R_IA64_PCREL32MSB: + case R_IA64_LTOFF_FPTR32MSB: case R_IA64_SEGREL32MSB: case R_IA64_SECREL32MSB: case R_IA64_LTV32MSB: @@ -2824,6 +2831,7 @@ elfNN_ia64_install_value (abfd, hit_addr case R_IA64_GPREL32LSB: case R_IA64_FPTR32LSB: case R_IA64_PCREL32LSB: + case R_IA64_LTOFF_FPTR32LSB: case R_IA64_SEGREL32LSB: case R_IA64_SECREL32LSB: case R_IA64_LTV32LSB: @@ -3678,6 +3686,8 @@ elfNN_ia64_relocate_section (output_bfd, case R_IA64_LTOFF_FPTR22: case R_IA64_LTOFF_FPTR64I: + case R_IA64_LTOFF_FPTR32MSB: + case R_IA64_LTOFF_FPTR32LSB: case R_IA64_LTOFF_FPTR64MSB: case R_IA64_LTOFF_FPTR64LSB: { diff -uprN binutils-2.11.90.0.23/bfd/libbfd.h binutils-2.11.90.0.24/bfd/libbfd.h --- binutils-2.11.90.0.23/bfd/libbfd.h Fri Jun 1 21:47:14 2001 +++ binutils-2.11.90.0.24/bfd/libbfd.h Sat Jul 14 10:29:15 2001 @@ -1006,6 +1006,8 @@ static const char *const bfd_reloc_code_ "BFD_RELOC_IA64_PCREL64LSB", "BFD_RELOC_IA64_LTOFF_FPTR22", "BFD_RELOC_IA64_LTOFF_FPTR64I", + "BFD_RELOC_IA64_LTOFF_FPTR32MSB", + "BFD_RELOC_IA64_LTOFF_FPTR32LSB", "BFD_RELOC_IA64_LTOFF_FPTR64MSB", "BFD_RELOC_IA64_LTOFF_FPTR64LSB", "BFD_RELOC_IA64_SEGREL32MSB", diff -uprN binutils-2.11.90.0.23/bfd/reloc.c binutils-2.11.90.0.24/bfd/reloc.c --- binutils-2.11.90.0.23/bfd/reloc.c Fri Jun 1 21:47:15 2001 +++ binutils-2.11.90.0.24/bfd/reloc.c Sat Jul 14 10:29:15 2001 @@ -2936,6 +2936,10 @@ ENUMX ENUMX BFD_RELOC_IA64_LTOFF_FPTR64I ENUMX + BFD_RELOC_IA64_LTOFF_FPTR32MSB +ENUMX + BFD_RELOC_IA64_LTOFF_FPTR32LSB +ENUMX BFD_RELOC_IA64_LTOFF_FPTR64MSB ENUMX BFD_RELOC_IA64_LTOFF_FPTR64LSB diff -uprN binutils-2.11.90.0.23/binutils/ChangeLog binutils-2.11.90.0.24/binutils/ChangeLog --- binutils-2.11.90.0.23/binutils/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/binutils/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,7 @@ +2001-07-11 Nick Clifton + + * MAINTAINERS: Add notes about branch checkins. + 2001-07-05 Daniel Berlin * readelf.c (display_debug_macinfo): New function, display diff -uprN binutils-2.11.90.0.23/binutils/MAINTAINERS binutils-2.11.90.0.24/binutils/MAINTAINERS --- binutils-2.11.90.0.23/binutils/MAINTAINERS Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/binutils/MAINTAINERS Sat Jul 14 10:29:15 2001 @@ -91,3 +91,15 @@ spelling mistakes, blatantly incorrect c also blatantly obvious), and so on. Obvious fixes should always be small, the larger they are, the more likely it is that they contain some un-obvious side effect or consequence. + + --------- Branch Checkins --------- + +If a patch is approved for check in to the mainline sources, it can +also be checked into the current release branch. Normally however +only bug fixes should be applied to the branch. New features, new +ports, etc, should be restricted to the mainline. (Otherwise the +burden of maintaining the branch in sync with the mainline becomes too +great). If you are uncertain as to whether a patch is appropriate for +the branch, ask the branch maintainer. This is: + + Philip Blundell diff -uprN binutils-2.11.90.0.23/binutils.spec binutils-2.11.90.0.24/binutils.spec --- binutils-2.11.90.0.23/binutils.spec Fri Jul 6 09:27:39 2001 +++ binutils-2.11.90.0.24/binutils.spec Fri Jul 13 18:29:01 2001 @@ -9,7 +9,7 @@ Summary: A GNU collection of binary utilities. Name: binutils -Version: 2.11.90.0.23 +Version: 2.11.90.0.24 Release: 1 Copyright: GPL Group: Development/Tools diff -uprN binutils-2.11.90.0.23/gas/ChangeLog binutils-2.11.90.0.24/gas/ChangeLog --- binutils-2.11.90.0.23/gas/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,59 @@ +2001-07-14 matthew green + + * configure.in (i386-*-netbsdelf*): New target. + * configure: Regenerate. + +2001-07-11 Jakub Jelinek + + * testsuite/gas/sparc/unalign.s: Remove .uaxword test. + * testsuite/gas/sparc/unalign.d: Adjust accordingly. + +2001-07-10 Mark Elbrecht + + * config/obj-coff.c (coff_frob_symbol): Don't merge if the storage + class of the non-debug symbol is C_NULL. + +2001-07-10 Anders Norlander + + * doc/as.texinfo (Incbin): Fix grammatical errors. + +2001-07-10 Nick Clifton + + * depend.c: Use FOPEN_.. macros in calls to fopen(). + * input-file.c: As above. + * itbl-ops.c: As above. + * listing.c: As above. + * output-file.c: As above. + * read.c: As above. + +2001-07-08 Anders Norlander + + * read.c (s_incbin): New .incbin function. + * read.c (potable): Add "incbin" pseudo-op. + * read.h: Add s_incbin prototype. + * doc/as.texinfo (incbin): Document .incbin pseudo-op. + * NEWS: Mention new feature. + +2001-07-07 Nick Clifton + + * ecoff.c (add_file): Only set debug_type to DEBUG_NONE if it is + DEBUG_UNSPECIFIED. + +2001-07-06 John Healy + + * cgen.c (gas_cgen_save_fixups): Modified to allow more than one + set of fixups to be stored. + (gas_cgen_restore_fixups): Modified to allow the fixup chain to be + restored to be chosen from any that are saved. + (gas_cgen_swap_fixups): Modified to allow the current set of + fixups to be swapped with any other set that has been saved. + (gas_cgen_initialize_saved_fixups_array): New routine. + * cgen.h: Modifed prototypes for gas_cgen_save_fixups, + gas_cgen_restore_fixups, and gas_cgen_swap_fixups. Added definitions + or MAX_SAVED_FIXUP_CHAINS. + * config/tc-m32r.c (assemble_two_insns): Changed calls to fixup + store, swap and restore fuctions to reflect the new interface. + 2001-07-06 Nick Clifton * config/tc-m68k.c (md_estimate_size_before_relax): Catch and diff -uprN binutils-2.11.90.0.23/gas/NEWS binutils-2.11.90.0.24/gas/NEWS --- binutils-2.11.90.0.23/gas/NEWS Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/NEWS Sat Jul 14 10:29:15 2001 @@ -1,5 +1,8 @@ -*- text -*- +New psuedo op: .incbin to include a set of binary data at a given point +in the assembly. Contributed by Anders Norlander. + The MIPS assembler now accepts -march/-mtune. -mcpu has been deprecated but still works for compatability. diff -uprN binutils-2.11.90.0.23/gas/cgen.c binutils-2.11.90.0.24/gas/cgen.c --- binutils-2.11.90.0.23/gas/cgen.c Mon May 21 10:35:39 2001 +++ binutils-2.11.90.0.24/gas/cgen.c Sat Jul 14 10:29:15 2001 @@ -94,62 +94,115 @@ queue_fixup (opindex, opinfo, expP) ++ num_fixups; } -/* The following three functions allow a backup of the fixup chain to be made, - and to have this backup be swapped with the current chain. This allows - certain ports, eg the m32r, to swap two instructions and swap their fixups - at the same time. */ -/* ??? I think with cgen_asm_finish_insn (or something else) there is no - more need for this. */ +/* The following functions allow fixup chains to be stored, retrieved, + and swapped. They are a generalization of a pre-existing scheme + for storing, restoring and swapping fixup chains that was used by + the m32r port. The functionality is essentially the same, only + instead of only being able to store a single fixup chain, an entire + array of fixup chains can be stored. It is the user's responsibility + to keep track of how many fixup chains have been stored and which + elements of the array they are in. + + The algorithms used are the same as in the old scheme. Other than the + "array-ness" of the whole thing, the functionality is identical to the + old scheme. + + gas_cgen_initialize_saved_fixups_array(): + Sets num_fixups_in_chain to 0 for each element. Call this from + md_begin() if you plan to use these functions and you want the + fixup count in each element to be set to 0 intially. This is + not necessary, but it's included just in case. It performs + the same function for each element in the array of fixup chains + that gas_init_parse() performs for the current fixups. + + gas_cgen_save_fixups (element): + element - element number of the array you wish to store the fixups + to. No mechanism is built in for tracking what element + was last stored to. + + gas_cgen_restore_fixups (element): + element - element number of the array you wish to restore the fixups + from. + + gas_cgen_swap_fixups(int element): + element - swap the current fixups with those in this element number. +*/ + +struct saved_fixups { + struct fixup fixup_chain[GAS_CGEN_MAX_FIXUPS]; + int num_fixups_in_chain; +}; -static struct fixup saved_fixups[GAS_CGEN_MAX_FIXUPS]; -static int saved_num_fixups; +static struct saved_fixups stored_fixups[MAX_SAVED_FIXUP_CHAINS]; -void -gas_cgen_save_fixups () +void +gas_cgen_initialize_saved_fixups_array () { - saved_num_fixups = num_fixups; - - memcpy (saved_fixups, fixups, sizeof (fixups[0]) * num_fixups); - - num_fixups = 0; + int i = 0; + while (i < MAX_SAVED_FIXUP_CHAINS) + stored_fixups[i++].num_fixups_in_chain = 0; } -void -gas_cgen_restore_fixups () +void +gas_cgen_save_fixups (int i) { - num_fixups = saved_num_fixups; - - memcpy (fixups, saved_fixups, sizeof (fixups[0]) * num_fixups); + if (i < 0 || i >= MAX_SAVED_FIXUP_CHAINS) + { + as_fatal("Index into stored_fixups[] out of bounds."); + return; + } + stored_fixups[i].num_fixups_in_chain = num_fixups; + memcpy(stored_fixups[i].fixup_chain, fixups, + sizeof (fixups[0])*num_fixups); + num_fixups = 0; +} - saved_num_fixups = 0; +void +gas_cgen_restore_fixups (int i) +{ + if (i < 0 || i >= MAX_SAVED_FIXUP_CHAINS) + { + as_fatal("Index into stored_fixups[] out of bounds."); + return; + } + num_fixups = stored_fixups[i].num_fixups_in_chain; + memcpy(fixups,stored_fixups[i].fixup_chain, + (sizeof (stored_fixups[i].fixup_chain[0]))*num_fixups); + stored_fixups[i].num_fixups_in_chain = 0; } -void -gas_cgen_swap_fixups () +void +gas_cgen_swap_fixups (int i) { - int tmp; - struct fixup tmp_fixup; + int tmp; + struct fixup tmp_fixup; - if (num_fixups == 0) - { - gas_cgen_restore_fixups (); - } - else if (saved_num_fixups == 0) - { - gas_cgen_save_fixups (); - } - else - { - tmp = saved_num_fixups; - saved_num_fixups = num_fixups; - num_fixups = tmp; - - for (tmp = GAS_CGEN_MAX_FIXUPS; tmp--;) - { - tmp_fixup = saved_fixups [tmp]; - saved_fixups [tmp] = fixups [tmp]; - fixups [tmp] = tmp_fixup; - } + if (i < 0 || i >= MAX_SAVED_FIXUP_CHAINS) + { + as_fatal("Index into stored_fixups[] out of bounds."); + return; + } + + if (num_fixups == 0) + { + gas_cgen_restore_fixups (i); + } + else if (stored_fixups[i].num_fixups_in_chain == 0) + { + gas_cgen_save_fixups (i); + } + else + { + tmp = stored_fixups[i].num_fixups_in_chain; + stored_fixups[i].num_fixups_in_chain = num_fixups; + num_fixups = tmp; + + for (tmp = GAS_CGEN_MAX_FIXUPS; tmp--;) + { + tmp_fixup = stored_fixups[i].fixup_chain [tmp]; + stored_fixups[i].fixup_chain[tmp] = fixups [tmp]; + fixups [tmp] = tmp_fixup; + } } } diff -uprN binutils-2.11.90.0.23/gas/cgen.h binutils-2.11.90.0.24/gas/cgen.h --- binutils-2.11.90.0.23/gas/cgen.h Fri Mar 9 11:16:34 2001 +++ binutils-2.11.90.0.24/gas/cgen.h Sat Jul 14 10:29:15 2001 @@ -53,9 +53,12 @@ extern const char * gas_cgen_parse_opera /* Call this from md_assemble to initialize the assembler callback. */ extern void gas_cgen_init_parse PARAMS ((void)); -extern void gas_cgen_save_fixups PARAMS ((void)); -extern void gas_cgen_restore_fixups PARAMS ((void)); -extern void gas_cgen_swap_fixups PARAMS ((void)); +/* Routines and macros for saving fixup chains. */ +extern void gas_cgen_save_fixups PARAMS ((int)); +extern void gas_cgen_restore_fixups PARAMS ((int)); +extern void gas_cgen_swap_fixups PARAMS ((int)); +extern void gas_cgen_initialize_saved_fixups_array PARAMS ((void)); +#define MAX_SAVED_FIXUP_CHAINS 50 /* Add a register to the assembler's hash table. This makes lets GAS parse registers for us. diff -uprN binutils-2.11.90.0.23/gas/config/obj-coff.c binutils-2.11.90.0.24/gas/config/obj-coff.c --- binutils-2.11.90.0.23/gas/config/obj-coff.c Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/config/obj-coff.c Sat Jul 14 10:29:15 2001 @@ -1174,18 +1174,21 @@ coff_frob_symbol (symp, punt) if (!SF_GET_DEBUG (symp)) { - symbolS *real; + symbolS * real; + if (!SF_GET_LOCAL (symp) && !SF_GET_STATICS (symp) && S_GET_STORAGE_CLASS (symp) != C_LABEL && symbol_constant_p(symp) && (real = symbol_find_base (S_GET_NAME (symp), DO_NOT_STRIP)) + && S_GET_STORAGE_CLASS (real) == C_NULL && real != symp) { c_symbol_merge (symp, real); *punt = 1; return; } + if (!S_IS_DEFINED (symp) && !SF_GET_LOCAL (symp)) { assert (S_GET_VALUE (symp) == 0); @@ -1199,6 +1202,7 @@ coff_frob_symbol (symp, punt) else S_SET_STORAGE_CLASS (symp, C_STAT); } + if (SF_GET_PROCESS (symp)) { if (S_GET_STORAGE_CLASS (symp) == C_BLOCK) @@ -1208,6 +1212,7 @@ coff_frob_symbol (symp, punt) else { symbolS *begin; + begin = *(symbolS **) stack_pop (block_stack); if (begin == 0) as_warn (_("mismatched .eb")); @@ -1215,9 +1220,11 @@ coff_frob_symbol (symp, punt) next_set_end = begin; } } + if (coff_last_function == 0 && SF_GET_FUNCTION (symp)) { union internal_auxent *auxp; + coff_last_function = symp; if (S_GET_NUMBER_AUXILIARY (symp) < 1) S_SET_NUMBER_AUXILIARY (symp, 1); @@ -1225,6 +1232,7 @@ coff_frob_symbol (symp, punt) memset (auxp->x_sym.x_fcnary.x_ary.x_dimen, 0, sizeof (auxp->x_sym.x_fcnary.x_ary.x_dimen)); } + if (S_GET_STORAGE_CLASS (symp) == C_EFCN) { if (coff_last_function == 0) @@ -1236,6 +1244,7 @@ coff_frob_symbol (symp, punt) coff_last_function = 0; } } + if (S_IS_EXTERNAL (symp)) S_SET_STORAGE_CLASS (symp, C_EXT); else if (SF_GET_LOCAL (symp)) diff -uprN binutils-2.11.90.0.23/gas/config/tc-m32r.c binutils-2.11.90.0.24/gas/config/tc-m32r.c --- binutils-2.11.90.0.23/gas/config/tc-m32r.c Sat May 12 00:06:37 2001 +++ binutils-2.11.90.0.24/gas/config/tc-m32r.c Sat Jul 14 10:29:15 2001 @@ -554,6 +554,8 @@ md_begin () scom_symbol.section = &scom_section; allow_m32rx (enable_m32rx); + + gas_cgen_initialize_saved_fixups_array(); } #define OPERAND_IS_COND_BIT(operand, indices, index) \ @@ -832,7 +834,7 @@ assemble_two_insns (str, str2, parallel_ /* Preserve any fixups that have been generated and reset the list to empty. */ - gas_cgen_save_fixups (); + gas_cgen_save_fixups (0); /* Get the indices of the operands of the instruction. */ /* FIXME: CGEN_FIELDS is already recorded, but relying on that fact @@ -941,7 +943,7 @@ assemble_two_insns (str, str2, parallel_ || (errmsg = (char *) can_make_parallel (&first, &second)) == NULL) { /* Get the fixups for the first instruction. */ - gas_cgen_swap_fixups (); + gas_cgen_swap_fixups (0); /* Write it out. */ expand_debug_syms (first.debug_sym_link, 1); @@ -953,7 +955,7 @@ assemble_two_insns (str, str2, parallel_ make_parallel (second.buffer); /* Get its fixups. */ - gas_cgen_restore_fixups (); + gas_cgen_restore_fixups (0); /* Write it out. */ expand_debug_syms (second.debug_sym_link, 1); @@ -972,7 +974,7 @@ assemble_two_insns (str, str2, parallel_ make_parallel (first.buffer); /* Get the fixups for the first instruction. */ - gas_cgen_restore_fixups (); + gas_cgen_restore_fixups (0); /* Write out the first instruction. */ expand_debug_syms (first.debug_sym_link, 1); diff -uprN binutils-2.11.90.0.23/gas/configure binutils-2.11.90.0.24/gas/configure --- binutils-2.11.90.0.23/gas/configure Sat Jun 9 00:13:01 2001 +++ binutils-2.11.90.0.24/gas/configure Sat Jul 14 10:29:15 2001 @@ -2366,6 +2366,7 @@ for this_target in $target $canon_target i386-*-beospe*) fmt=coff em=pe bfd_gas=yes ;; i386-*-beoself* | i386-*-beos*) fmt=elf bfd_gas=yes ;; i386-*-bsd*) fmt=aout em=386bsd ;; + i386-*-netbsdelf*) fmt=elf em=nbsd bfd_gas=yes ;; i386-*-netbsd0.8) fmt=aout em=386bsd ;; i386-*-netbsd*) fmt=aout em=nbsd bfd_gas=yes;; i386-*-openbsd*) fmt=aout em=nbsd bfd_gas=yes;; diff -uprN binutils-2.11.90.0.23/gas/configure.in binutils-2.11.90.0.24/gas/configure.in --- binutils-2.11.90.0.23/gas/configure.in Sat Jun 9 00:13:01 2001 +++ binutils-2.11.90.0.24/gas/configure.in Sat Jul 14 10:29:15 2001 @@ -242,6 +242,7 @@ changequote([,])dnl i386-*-beoself* | i386-*-beos*) fmt=elf bfd_gas=yes ;; i386-*-bsd*) fmt=aout em=386bsd ;; i386-*-netbsd0.8) fmt=aout em=386bsd ;; + i386-*-netbsdelf*) fmt=elf em=nbsd bfd_gas=yes ;; i386-*-netbsd*) fmt=aout em=nbsd bfd_gas=yes;; i386-*-openbsd*) fmt=aout em=nbsd bfd_gas=yes;; i386-*-linux*aout* | i386-*-linux*oldld) fmt=aout em=linux ;; diff -uprN binutils-2.11.90.0.23/gas/depend.c binutils-2.11.90.0.24/gas/depend.c --- binutils-2.11.90.0.23/gas/depend.c Fri Mar 9 11:16:40 2001 +++ binutils-2.11.90.0.24/gas/depend.c Sat Jul 14 10:29:15 2001 @@ -1,5 +1,5 @@ /* depend.c - Handle dependency tracking. - Copyright 1997, 1998, 2000 Free Software Foundation, Inc. + Copyright 1997, 1998, 2000, 2001 Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -21,15 +21,16 @@ #include "as.h" /* The file to write to, or NULL if no dependencies being kept. */ -static char *dep_file = NULL; +static char * dep_file = NULL; -struct dependency { - char *file; - struct dependency *next; -}; +struct dependency + { + char * file; + struct dependency * next; + }; /* All the files we depend on. */ -static struct dependency *dep_chain = NULL; +static struct dependency * dep_chain = NULL; /* Current column in output file. */ static int column = 0; @@ -86,9 +87,11 @@ quote_string_for_make (file, src) { char *p = src; int i = 0; + for (;;) { char c = *p++; + switch (c) { case '\0': @@ -102,6 +105,7 @@ quote_string_for_make (file, src) the end of a file name; and backslashes in other contexts should not be doubled. */ char *q; + for (q = p - 1; src < q && q[-1] == '\\'; q--) { if (file) @@ -190,7 +194,7 @@ print_dependencies () if (dep_file == NULL) return; - f = fopen (dep_file, "w"); + f = fopen (dep_file, FOPEN_WT); if (f == NULL) { as_warn (_("Can't open `%s' for writing"), dep_file); diff -uprN binutils-2.11.90.0.23/gas/doc/as.texinfo binutils-2.11.90.0.24/gas/doc/as.texinfo --- binutils-2.11.90.0.23/gas/doc/as.texinfo Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/doc/as.texinfo Sat Jul 14 10:29:15 2001 @@ -3452,6 +3452,7 @@ Some machine configurations provide addi * hword:: @code{.hword @var{expressions}} * Ident:: @code{.ident} * If:: @code{.if @var{absolute expression}} +* Incbin:: @code{.incbin "@var{file}"[,@var{skip}[,@var{count}]]} * Include:: @code{.include "@var{file}"} * Int:: @code{.int @var{expressions}} @ifset ELF @@ -4123,6 +4124,22 @@ Assembles the following section of code Like @code{.ifeqs}, but the sense of the test is reversed: this assembles the following section of code if the two strings are not the same. @end table + +@node Incbin +@section @code{.incbin "@var{file}"[,@var{skip}[,@var{count}]]} + +@cindex @code{incbin} directive +@cindex binary files, including +The @code{incbin} directive includes @var{file} verbatim at the current +location. You can control the search paths used with the @samp{-I} command-line +option (@pxref{Invoking,,Command-Line Options}). Quotation marks are required +around @var{file}. + +The @var{skip} argument skips a number of bytes from the start of the +@var{file}. The @var{count} argument indicates the maximum number of bytes to +read. Note that the data is not aligned in any way, so it is the user's +responsibility to make sure that proper alignment is provided both before and +after the @code{incbin} directive. @node Include @section @code{.include "@var{file}"} diff -uprN binutils-2.11.90.0.23/gas/ecoff.c binutils-2.11.90.0.24/gas/ecoff.c --- binutils-2.11.90.0.23/gas/ecoff.c Wed May 2 23:30:53 2001 +++ binutils-2.11.90.0.24/gas/ecoff.c Sat Jul 14 10:29:15 2001 @@ -2220,7 +2220,7 @@ add_file (file_name, indx, fake) if (! symbol_table_frozen && debug_type == DEBUG_UNSPECIFIED) debug_type = DEBUG_ECOFF; } - else + else if (debug_type == DEBUG_UNSPECIFIED) debug_type = DEBUG_NONE; #ifndef NO_LISTING diff -uprN binutils-2.11.90.0.23/gas/input-file.c binutils-2.11.90.0.24/gas/input-file.c --- binutils-2.11.90.0.23/gas/input-file.c Fri Mar 9 11:16:42 2001 +++ binutils-2.11.90.0.24/gas/input-file.c Sat Jul 14 10:29:15 2001 @@ -1,5 +1,5 @@ /* input_file.c - Deal with Input Files - - Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1999, 2000 + Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1999, 2000, 2001 Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -19,12 +19,10 @@ Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ -/* - * Confines all details of reading source bytes to this module. - * All O/S specific crocks should live here. - * What we lose in "efficiency" we gain in modularity. - * Note we don't need to #include the "as.h" file. No common coupling! - */ +/* Confines all details of reading source bytes to this module. + All O/S specific crocks should live here. + What we lose in "efficiency" we gain in modularity. + Note we don't need to #include the "as.h" file. No common coupling! */ #include #include @@ -35,34 +33,30 @@ static int input_file_get PARAMS ((char *, int)); /* This variable is non-zero if the file currently being read should be - preprocessed by app. It is zero if the file can be read straight in. - */ + preprocessed by app. It is zero if the file can be read straight in. */ int preprocess = 0; -/* - * This code opens a file, then delivers BUFFER_SIZE character - * chunks of the file on demand. - * BUFFER_SIZE is supposed to be a number chosen for speed. - * The caller only asks once what BUFFER_SIZE is, and asks before - * the nature of the input files (if any) is known. - */ +/* This code opens a file, then delivers BUFFER_SIZE character + chunks of the file on demand. + BUFFER_SIZE is supposed to be a number chosen for speed. + The caller only asks once what BUFFER_SIZE is, and asks before + the nature of the input files (if any) is known. */ #define BUFFER_SIZE (32 * 1024) -/* - * We use static data: the data area is not sharable. - */ +/* We use static data: the data area is not sharable. */ static FILE *f_in; static char *file_name; /* Struct for saving the state of this module for file includes. */ -struct saved_file { - FILE *f_in; - char *file_name; - int preprocess; - char *app_save; -}; +struct saved_file + { + FILE * f_in; + char * file_name; + int preprocess; + char * app_save; + }; /* These hooks accomodate most operating systems. */ @@ -92,6 +86,7 @@ input_file_is_open () /* Push the state of our input, returning a pointer to saved info that can be restored with input_file_pop (). */ + char * input_file_push () { @@ -105,7 +100,8 @@ input_file_push () if (preprocess) saved->app_save = app_push (); - input_file_begin (); /* Initialize for new file */ + /* Initialize for new file. */ + input_file_begin (); return (char *) saved; } @@ -116,7 +112,7 @@ input_file_pop (arg) { register struct saved_file *saved = (struct saved_file *) arg; - input_file_end (); /* Close out old file */ + input_file_end (); /* Close out old file. */ f_in = saved->f_in; file_name = saved->file_name; @@ -140,7 +136,7 @@ input_file_open (filename, pre) assert (filename != 0); /* Filename may not be NULL. */ if (filename[0]) { /* We have a file name. Suck it and see. */ - f_in = fopen (filename, "r"); + f_in = fopen (filename, FOPEN_RT); file_name = filename; } else @@ -157,7 +153,8 @@ input_file_open (filename, pre) c = getc (f_in); if (c == '#') - { /* Begins with comment, may not want to preprocess */ + { + /* Begins with comment, may not want to preprocess. */ c = getc (f_in); if (c == 'N') { @@ -165,7 +162,7 @@ input_file_open (filename, pre) if (!strcmp (buf, "O_APP\n")) preprocess = 0; if (!strchr (buf, '\n')) - ungetc ('#', f_in); /* It was longer */ + ungetc ('#', f_in); /* It was longer. */ else ungetc ('\n', f_in); } @@ -179,13 +176,14 @@ input_file_open (filename, pre) } /* Close input file. */ + void input_file_close () { + /* Don't close a null file pointer. */ if (f_in != NULL) - { - fclose (f_in); - } /* don't close a null file pointer */ + fclose (f_in); + f_in = 0; } @@ -218,12 +216,10 @@ input_file_give_next_buffer (where) if (f_in == (FILE *) 0) return 0; - /* - * fflush (stdin); could be done here if you want to synchronise - * stdin and stdout, for the case where our input file is stdin. - * Since the assembler shouldn't do any output to stdout, we - * don't bother to synch output and input. - */ + /* fflush (stdin); could be done here if you want to synchronise + stdin and stdout, for the case where our input file is stdin. + Since the assembler shouldn't do any output to stdout, we + don't bother to synch output and input. */ if (preprocess) size = do_scrub_chars (input_file_get, where, BUFFER_SIZE); else @@ -242,5 +238,6 @@ input_file_give_next_buffer (where) f_in = (FILE *) 0; return_value = 0; } - return (return_value); + + return return_value; } diff -uprN binutils-2.11.90.0.23/gas/itbl-ops.c binutils-2.11.90.0.24/gas/itbl-ops.c --- binutils-2.11.90.0.23/gas/itbl-ops.c Wed May 2 23:30:54 2001 +++ binutils-2.11.90.0.24/gas/itbl-ops.c Sat Jul 14 10:29:15 2001 @@ -183,17 +183,18 @@ itbl_parse (char *insntbl) { extern FILE *yyin; extern int yyparse (void); - yyin = fopen (insntbl, "r"); + + yyin = fopen (insntbl, FOPEN_RT); if (yyin == 0) { printf ("Can't open processor instruction specification file \"%s\"\n", insntbl); return 1; } - else - { - while (yyparse ()); - } + + while (yyparse ()) + ; + fclose (yyin); itbl_have_entries = 1; return 0; diff -uprN binutils-2.11.90.0.23/gas/listing.c binutils-2.11.90.0.24/gas/listing.c --- binutils-2.11.90.0.23/gas/listing.c Sat Mar 31 19:57:27 2001 +++ binutils-2.11.90.0.24/gas/listing.c Sat Jul 14 10:29:15 2001 @@ -485,7 +485,7 @@ buffer_line (file, line, size) } last_open_file_info = file; - last_open_file = fopen (file->filename, "r"); + last_open_file = fopen (file->filename, FOPEN_RT); if (last_open_file == NULL) { file->at_end = 1; @@ -1121,7 +1121,7 @@ listing_print (name) } else { - list_file = fopen (name, "w"); + list_file = fopen (name, FOPEN_WT); if (list_file != NULL) using_stdout = 0; else diff -uprN binutils-2.11.90.0.23/gas/output-file.c binutils-2.11.90.0.24/gas/output-file.c --- binutils-2.11.90.0.23/gas/output-file.c Fri Mar 9 11:16:43 2001 +++ binutils-2.11.90.0.24/gas/output-file.c Sat Jul 14 10:29:15 2001 @@ -1,5 +1,5 @@ /* output-file.c - Deal with the output file - Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1996, 1998, 1999 + Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1996, 1998, 1999, 2001 Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -16,7 +16,8 @@ You should have received a copy of the GNU General Public License along with GAS; see the file COPYING. If not, write to - the Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ + the Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA + 02111-1307, USA. */ #include @@ -44,14 +45,14 @@ output_file_create (name) char *name; { if (name[0] == '-' && name[1] == '\0') - { - as_fatal (_("Can't open a bfd on stdout %s "), name); - } + as_fatal (_("Can't open a bfd on stdout %s "), name); + else if (!(stdoutput = bfd_openw (name, TARGET_FORMAT))) { as_perror (_("FATAL: Can't create %s"), name); exit (EXIT_FAILURE); } + bfd_set_format (stdoutput, bfd_object); #ifdef BFD_ASSEMBLER bfd_set_arch_mach (stdoutput, TARGET_ARCH, TARGET_MACH); @@ -73,14 +74,14 @@ output_file_close (filename) exit (EXIT_FAILURE); } #else - /* Close the bfd without getting bfd to write out anything by itself */ + /* Close the bfd without getting bfd to write out anything by itself. */ if (bfd_close_all_done (stdoutput) == 0) { as_perror (_("FATAL: Can't close %s\n"), filename); exit (EXIT_FAILURE); } #endif - stdoutput = NULL; /* Trust nobody! */ + stdoutput = NULL; /* Trust nobody! */ } #ifndef BFD_ASSEMBLER @@ -108,11 +109,12 @@ output_file_create (name) return; } - stdoutput = fopen (name, "wb"); + stdoutput = fopen (name, FOPEN_WB); /* Some systems don't grok "b" in fopen modes. */ + /* XXX - is this still necessary now that we use FOPEN_WB ? */ if (stdoutput == NULL) - stdoutput = fopen (name, "w"); + stdoutput = fopen (name, FOPEN_W); if (stdoutput == NULL) { @@ -130,18 +132,21 @@ output_file_close (filename) as_perror (_("FATAL: Can't close %s"), filename); exit (EXIT_FAILURE); } - stdoutput = NULL; /* Trust nobody! */ + + /* Trust nobody! */ + stdoutput = NULL; } void output_file_append (where, length, filename) - char *where; - long length; - char *filename; + char * where; + long length; + char * filename; { for (; length; length--, where++) { (void) putc (*where, stdoutput); + if (ferror (stdoutput)) /* if ( EOF == (putc( *where, stdoutput )) ) */ { @@ -153,4 +158,3 @@ output_file_append (where, length, filen #endif -/* end of output-file.c */ diff -uprN binutils-2.11.90.0.23/gas/read.c binutils-2.11.90.0.24/gas/read.c --- binutils-2.11.90.0.23/gas/read.c Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/read.c Sat Jul 14 10:29:15 2001 @@ -1,6 +1,6 @@ /* read.c - read a source file - Copyright 1986, 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, - 1998, 1999, 2000 Free Software Foundation, Inc. + 1998, 1999, 2000, 2001 Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -349,6 +349,7 @@ static const pseudo_typeS potable[] = { {"ifne", s_if, (int) O_ne}, {"ifnes", s_ifeqs, 1}, {"ifnotdef", s_ifdef, 1}, + {"incbin", s_incbin, 0}, {"include", s_include, 0}, {"int", cons, 4}, {"irp", s_irp, 0}, @@ -4900,6 +4901,123 @@ equals (sym_name, reassign) } } +/* .incbin -- include a file verbatim at the current location. */ + +void +s_incbin (x) + int x ATTRIBUTE_UNUSED; +{ + FILE * binfile; + char * path; + char * filename; + char * binfrag; + long skip = 0; + long count = 0; + long bytes; + int len; + +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + SKIP_WHITESPACE (); + filename = demand_copy_string (& len); + if (filename == NULL) + return; + + SKIP_WHITESPACE (); + + /* Look for optional skip and count. */ + if (* input_line_pointer == ',') + { + ++ input_line_pointer; + skip = get_absolute_expression (); + + SKIP_WHITESPACE (); + + if (* input_line_pointer == ',') + { + ++ input_line_pointer; + + count = get_absolute_expression (); + if (count == 0) + as_warn (_(".incbin count zero, ignoring `%s'"), filename); + + SKIP_WHITESPACE (); + } + } + + demand_empty_rest_of_line (); + + /* Try opening absolute path first, then try include dirs. */ + binfile = fopen (filename, FOPEN_RB); + if (binfile == NULL) + { + int i; + + path = xmalloc ((unsigned long) len + include_dir_maxlen + 5); + + for (i = 0; i < include_dir_count; i++) + { + sprintf (path, "%s/%s", include_dirs[i], filename); + + binfile = fopen (path, FOPEN_RB); + if (binfile != NULL) + break; + } + + if (binfile == NULL) + as_bad (_("file not found: %s"), filename); + } + else + path = xstrdup (filename); + + if (binfile) + { + long file_len; + + register_dependency (path); + + /* Compute the length of the file. */ + if (fseek (binfile, 0, SEEK_END) != 0) + { + as_bad (_("seek to end of .incbin file failed `%s'"), path); + goto done; + } + file_len = ftell (binfile); + + /* If a count was not specified use the size of the file. */ + if (count == 0) + count = file_len; + + if (skip + count > file_len) + { + as_bad (_("skip (%ld) + count (%ld) larger than file size (%ld)"), + skip, count, file_len); + goto done; + } + + if (fseek (binfile, skip, SEEK_SET) != 0) + { + as_bad (_("could not skip to %ld in file `%s'"), skip, path); + goto done; + } + + /* Allocate frag space and store file contents in it. */ + binfrag = frag_more (count); + + bytes = fread (binfrag, 1, count, binfile); + if (bytes < count) + as_warn (_("truncated file `%s', %ld of %ld bytes read"), + path, bytes, count); + } +done: + if (binfile != NULL) + fclose (binfile); + if (path) + free (path); +} + /* .include -- include a file at this point. */ void @@ -4948,7 +5066,7 @@ s_include (arg) strcpy (path, include_dirs[i]); strcat (path, "/"); strcat (path, filename); - if (0 != (try = fopen (path, "r"))) + if (0 != (try = fopen (path, FOPEN_RT))) { fclose (try); goto gotit; diff -uprN binutils-2.11.90.0.23/gas/read.h binutils-2.11.90.0.24/gas/read.h --- binutils-2.11.90.0.23/gas/read.h Fri Mar 9 11:16:45 2001 +++ binutils-2.11.90.0.24/gas/read.h Sat Jul 14 10:29:15 2001 @@ -181,3 +181,4 @@ extern void s_text PARAMS ((int)); extern void stringer PARAMS ((int append_zero)); extern void s_xstab PARAMS ((int what)); extern void s_rva PARAMS ((int)); +extern void s_incbin PARAMS ((int)); diff -uprN binutils-2.11.90.0.23/gas/testsuite/ChangeLog binutils-2.11.90.0.24/gas/testsuite/ChangeLog --- binutils-2.11.90.0.23/gas/testsuite/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/gas/testsuite/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,16 @@ +2001-07-14 Nick Clifton + + Fix failures in MIPS testsuite: + * gas/mips/abs.d: Pass -march=4000 to gas. + * gas/mips/eld_rel2.d: Do not expect zeroes at end of dump. + * gas/mips/elf_e_flags1.d: Pass -march=4000 to gas. + +2001-07-08 Anders Norlander + + * gas/all/gas.exp: Run incbin test. + * gas/all/incbin.s: New file. + * gas/all/incbin.d: New file. + 2001-07-04 Richard Sandiford * gas/mips/elf-rel3.s: Add zero word to end of file. diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/all/gas.exp binutils-2.11.90.0.24/gas/testsuite/gas/all/gas.exp --- binutils-2.11.90.0.23/gas/testsuite/gas/all/gas.exp Thu Jul 20 15:39:11 2000 +++ binutils-2.11.90.0.24/gas/testsuite/gas/all/gas.exp Sat Jul 14 10:29:15 2001 @@ -156,6 +156,8 @@ if ![istarget *c54x*-*-*] then { test_cond } +run_dump_test incbin + # FIXME: this is here cause of a bug in DejaGnu 1.1.1. When it is no longer # in use, then this can be removed. if [info exists errorInfo] then { diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/all/incbin.d binutils-2.11.90.0.24/gas/testsuite/gas/all/incbin.d --- binutils-2.11.90.0.23/gas/testsuite/gas/all/incbin.d Wed Dec 31 16:00:00 1969 +++ binutils-2.11.90.0.24/gas/testsuite/gas/all/incbin.d Sat Jul 14 10:29:15 2001 @@ -0,0 +1,17 @@ +#as: -I$srcdir/$subdir +#objdump: -s -j .data +#name: incbin + +# Test the incbin pseudo-op + +.*: .* + +Contents of section .data: + 0000 2e646174 610a2e69 6e636269 6e202269 .data..incbin "i + 0010 6e636269 6e2e7322 0a2e696e 6362696e ncbin.s"..incbin + 0020 2022696e 6362696e 2e73222c 302c3238 "incbin.s",0,28 + 0030 0a2e696e 6362696e 2022696e 6362696e ..incbin "incbin + 0040 2e73222c 31352c39 0a2e7032 616c6967 .s",15,9..p2alig + 0050 6e20340a 2e646174 610a2e69 6e636269 n 4..data..incbi + 0060 6e202269 6e636269 6e2e7322 0a2e696e n "incbin.s"..in + 0070 696e6362 696e2e73 22000000 00000000 incbin.s"....... diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/all/incbin.s binutils-2.11.90.0.24/gas/testsuite/gas/all/incbin.s --- binutils-2.11.90.0.23/gas/testsuite/gas/all/incbin.s Wed Dec 31 16:00:00 1969 +++ binutils-2.11.90.0.24/gas/testsuite/gas/all/incbin.s Sat Jul 14 10:29:15 2001 @@ -0,0 +1,5 @@ +.data +.incbin "incbin.s" +.incbin "incbin.s",0,28 +.incbin "incbin.s",15,9 +.p2align 4 diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/mips/abs.d binutils-2.11.90.0.24/gas/testsuite/gas/mips/abs.d --- binutils-2.11.90.0.23/gas/testsuite/gas/mips/abs.d Mon May 21 10:36:59 2001 +++ binutils-2.11.90.0.24/gas/testsuite/gas/mips/abs.d Sat Jul 14 10:29:15 2001 @@ -1,4 +1,5 @@ #objdump: -dr --prefix-addresses -mmips:4000 +#as:-march=4000 #name: MIPS abs # Test the abs macro. diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/mips/elf-rel2.d binutils-2.11.90.0.24/gas/testsuite/gas/mips/elf-rel2.d --- binutils-2.11.90.0.23/gas/testsuite/gas/mips/elf-rel2.d Fri May 25 12:06:58 2001 +++ binutils-2.11.90.0.24/gas/testsuite/gas/mips/elf-rel2.d Sat Jul 14 10:29:15 2001 @@ -23,4 +23,4 @@ OFFSET TYPE VALUE Contents of section \.text: 0000 d782c000 d782c008 d782c010 c782c000 .* 0010 c782c004 c782c008 8f82c000 8f82c004 .* - 0020 8f82c008 00000000 00000000 00000000 .* + 0020 8f82c008 .* diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/mips/elf_e_flags1.d binutils-2.11.90.0.24/gas/testsuite/gas/mips/elf_e_flags1.d --- binutils-2.11.90.0.23/gas/testsuite/gas/mips/elf_e_flags1.d Mon May 21 10:37:00 2001 +++ binutils-2.11.90.0.24/gas/testsuite/gas/mips/elf_e_flags1.d Sat Jul 14 10:29:15 2001 @@ -1,6 +1,7 @@ # name: ELF e_flags: nothing special # source: elf_e_flags.s # objdump: -fd +# as: -march=4000 .*:.*file format.*mips.* architecture: mips:.*, flags 0x00000011: diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/sparc/unalign.d binutils-2.11.90.0.24/gas/testsuite/gas/sparc/unalign.d --- binutils-2.11.90.0.23/gas/testsuite/gas/sparc/unalign.d Sat Jun 2 23:18:03 2001 +++ binutils-2.11.90.0.24/gas/testsuite/gas/sparc/unalign.d Sat Jul 14 10:29:15 2001 @@ -12,7 +12,4 @@ Disassembly of section .data: 1: R_SPARC_UA32 fred 4: 00 02 00 00 (unimp|illtrap) 0x20000 6: R_SPARC_UA16 jim - 8: 03 00 00 00 sethi %hi\(0\), %g1 - 9: R_SPARC_UA64 harry - c: 00 00 00 00 (unimp|illtrap) 0 - 10: 00 04 05 06 (unimp|illtrap) 0x40506 + 8: 03 04 05 06 sethi %hi\(0x10141800\), %g1 diff -uprN binutils-2.11.90.0.23/gas/testsuite/gas/sparc/unalign.s binutils-2.11.90.0.24/gas/testsuite/gas/sparc/unalign.s --- binutils-2.11.90.0.23/gas/testsuite/gas/sparc/unalign.s Fri Jun 1 20:50:16 2001 +++ binutils-2.11.90.0.24/gas/testsuite/gas/sparc/unalign.s Sat Jul 14 10:29:15 2001 @@ -7,9 +7,6 @@ foo: .byte 0x2 .uahalf jim .byte 0x3 - .uaxword harry .byte 0x4 .byte 0x5 .byte 0x6 - - diff -uprN binutils-2.11.90.0.23/include/ChangeLog binutils-2.11.90.0.24/include/ChangeLog --- binutils-2.11.90.0.23/include/ChangeLog Tue Jun 19 11:57:58 2001 +++ binutils-2.11.90.0.24/include/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,13 @@ +2001-07-12 Jeff Johnston + + * opcode/cgen.h (CGEN_INSN): Add regex support. + (build_insn_regex): Declare. + +2001-07-10 Jeff Johnston + + * xregex.h: New file to support libiberty regex. + * xregex2.h: Ditto. + 2001-06-15 Hans-Peter Nilsson * bfdlink.h (struct bfd_link_info): New member export_dynamic. diff -uprN binutils-2.11.90.0.23/include/opcode/ChangeLog binutils-2.11.90.0.24/include/opcode/ChangeLog --- binutils-2.11.90.0.23/include/opcode/ChangeLog Tue Jun 19 11:57:58 2001 +++ binutils-2.11.90.0.24/include/opcode/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,12 @@ +2001-07-11 Frank Ch. Eigler + + * cgen.h (CGEN_MACH): Add insn_chunk_bitsize field. + (cgen_cpu_desc): Ditto. + +2001-07-07 Ben Elliston + + * m88k.h: Clean up and reformat. Remove unused code. + 2001-06-14 Geoffrey Keating * cgen.h (cgen_keyword): Add nonalpha_chars field. diff -uprN binutils-2.11.90.0.23/include/opcode/cgen.h binutils-2.11.90.0.24/include/opcode/cgen.h --- binutils-2.11.90.0.23/include/opcode/cgen.h Tue Jun 19 11:57:58 2001 +++ binutils-2.11.90.0.24/include/opcode/cgen.h Sat Jul 14 10:29:15 2001 @@ -199,6 +199,8 @@ typedef struct { const char *bfd_name; /* one of enum mach_attr */ int num; + /* parameter from mach->cpu */ + unsigned int insn_chunk_bitsize; } CGEN_MACH; /* Parse result (also extraction result). @@ -1007,6 +1009,11 @@ struct cgen_insn const CGEN_IBASE *base; const CGEN_OPCODE *opcode; const CGEN_OPINST *opinst; + + /* Regex to disambiguate overloaded opcodes */ + void *rx; +#define CGEN_INSN_RX(insn) ((insn)->rx) +#define CGEN_MAX_RX_ELEMENTS (CGEN_MAX_SYNTAX_ELEMENTS * 5) }; /* Instruction lists. @@ -1166,6 +1173,10 @@ typedef struct cgen_cpu_desc lazily fetch the data from there. */ unsigned int word_bitsize; + /* Instruction chunk size (in bits), for purposes of endianness + conversion. */ + unsigned int insn_chunk_bitsize; + /* Indicator if sizes are unknown. This is used by default_insn_bitsize,base_insn_bitsize if there is a difference between the selected isa's. */ @@ -1357,6 +1368,11 @@ extern void CGEN_SYM (cpu_close) PARAMS Called by init_asm/init_dis. */ extern void CGEN_SYM (init_opcode_table) PARAMS ((CGEN_CPU_DESC cd_)); + +/* build the insn selection regex. + called by init_opcode_table */ + +extern char * CGEN_SYM(build_insn_regex) PARAMS ((CGEN_INSN *insn_)); /* Initialize the ibld table for use. Called by init_asm/init_dis. */ diff -uprN binutils-2.11.90.0.23/include/opcode/m88k.h binutils-2.11.90.0.24/include/opcode/m88k.h --- binutils-2.11.90.0.23/include/opcode/m88k.h Thu Mar 15 14:53:15 2001 +++ binutils-2.11.90.0.24/include/opcode/m88k.h Sat Jul 14 10:29:15 2001 @@ -1,5 +1,5 @@ -/* Table of opcodes for the motorola 88k family. - Copyright 1989, 1990, 1991, 1993 Free Software Foundation, Inc. +/* Table of opcodes for the Motorola M88k family. + Copyright 1989, 1990, 1991, 1993, 2001 Free Software Foundation, Inc. This file is part of GDB and GAS. @@ -34,22 +34,6 @@ Foundation, Inc., 59 Temple Place - Suit * pointer to the next instruction in the linked list. These pointers * are initialized by init_disasm(). * - * Structure Format - * - * struct INSTAB { - * UPINT opcode; - * char *mnemonic; - * struct OPSPEC op1,op2,op3; - * struct SIM_FLAGS flgs; - * struct INSTAB *next; - * } - * - * struct OPSPEC { - * UPINT offset:5; - * UPINT width:6; - * UPINT type:5; - * } - * * Revision History * * Revision 1.0 11/08/85 Creation date @@ -61,54 +45,52 @@ Foundation, Inc., 59 Temple Place - Suit #include +/* Define the number of bits in the primary opcode field of the instruction, + the destination field, the source 1 and source 2 fields. */ -/* - * This file contains the structures and constants needed to build the M88000 - * simulator. It is the main include file, containing all the - * structures, macros and definitions except for the floating point - * instruction set. - */ +/* Size of opcode field. */ +#define OP 8 -/* - * The following flag informs the Simulator as to what type of byte ordering - * will be used. For instance, a BOFLAG = 1 indicates a DEC VAX and IBM type - * of ordering shall be used. -*/ +/* Size of destination. */ +#define DEST 6 -/* # define BOFLAG 1 */ /* BYTE ORDERING FLAG */ +/* Size of source1. */ +#define SOURCE1 6 -/* define the number of bits in the primary opcode field of the instruction, - * the destination field, the source 1 and source 2 fields. - */ -# define OP 8 /* size of opcode field */ -# define DEST 6 /* size of destination */ -# define SOURCE1 6 /* size of source1 */ -# define SOURCE2 6 /* size of source2 */ - -# define REGs 32 /* number of registers */ - -# define WORD long -# define FLAG unsigned -# define STATE short +/* Size of source2. */ +#define SOURCE2 6 + +/* Number of registers. */ +#define REGs 32 -# define TRUE 1 -# define FALSE 0 +/* Type definitions. */ -# define READ 0 -# define WRITE 1 +typedef unsigned int UINT; +#define WORD long +#define FLAG unsigned +#define STATE short /* The next four equates define the priorities that the various classes * of instructions have regarding writing results back into registers and - * signalling exceptions. - */ + * signalling exceptions. */ + /* PMEM is also defined in on Delta 88's. Sigh! */ #undef PMEM -# define PINT 0 /* Integer Priority */ -# define PFLT 1 /* Floating Point Priority */ -# define PMEM 2 /* Memory Priority */ -# define NA 3 /* Not Applicable, instruction doesnt write to regs */ -# define HIPRI 3 /* highest of these priorities */ +/* Integer priority. */ +#define PINT 0 + +/* Floating point priority. */ +#define PFLT 1 + +/* Memory priority. */ +#define PMEM 2 + +/* Not applicable, instruction doesn't write to regs. */ +#define NA 3 + +/* Highest of these priorities. */ +#define HIPRI 3 /* The instruction registers are an artificial mechanism to speed up * simulator execution. In the real processor, an instruction register @@ -122,231 +104,267 @@ Foundation, Inc., 59 Temple Place - Suit * Yes this wastes memory, but it executes much quicker. */ -struct IR_FIELDS { - unsigned op:OP, - dest: DEST, - src1: SOURCE1, - src2: SOURCE2; - int ltncy, - extime, - wb_pri; /* writeback priority */ - unsigned imm_flags:2,/* immediate size */ - rs1_used:1, /* register source 1 used */ - rs2_used:1, /* register source 2 used */ - rsd_used:1, /* register source/dest. used */ - c_flag:1, /* complement */ - u_flag:1, /* upper half word */ - n_flag:1, /* execute next */ - wb_flag:1, /* uses writeback slot */ - dest_64:1, /* dest size */ - s1_64:1, /* source 1 size */ - s2_64:1, /* source 2 size */ - scale_flag:1, /* scaled register */ - brk_flg:1; - }; - -struct mem_segs { - struct mem_wrd *seg; /* pointer (returned by calloc) to segment */ - unsigned long baseaddr; /* base load address from file headers */ - unsigned long endaddr; /* Ending address of segment */ - int flags; /* segment control flags (none defined 12/5/86) */ +struct IR_FIELDS +{ + unsigned op:OP, + dest: DEST, + src1: SOURCE1, + src2: SOURCE2; + int ltncy, + extime, + /* Writeback priority. */ + wb_pri; + /* Immediate size. */ + unsigned imm_flags:2, + /* Register source 1 used. */ + rs1_used:1, + /* Register source 2 used. */ + rs2_used:1, + /* Register source/dest. used. */ + rsd_used:1, + /* Complement. */ + c_flag:1, + /* Upper half word. */ + u_flag:1, + /* Execute next. */ + n_flag:1, + /* Uses writeback slot. */ + wb_flag:1, + /* Dest size. */ + dest_64:1, + /* Source 1 size. */ + s1_64:1, + /* Source 2 size. */ + s2_64:1, + scale_flag:1, + /* Scaled register. */ + brk_flg:1; +}; + +struct mem_segs +{ + /* Pointer (returned by calloc) to segment. */ + struct mem_wrd *seg; + + /* Base load address from file headers. */ + unsigned long baseaddr; + + /* Ending address of segment. */ + unsigned long endaddr; + + /* Segment control flags (none defined). */ + int flags; }; #define MAXSEGS (10) /* max number of segment allowed */ #define MEMSEGSIZE (sizeof(struct mem_segs))/* size of mem_segs structure */ - +#if 0 #define BRK_RD (0x01) /* break on memory read */ #define BRK_WR (0x02) /* break on memory write */ #define BRK_EXEC (0x04) /* break on execution */ #define BRK_CNT (0x08) /* break on terminal count */ +#endif - -struct mem_wrd { - struct IR_FIELDS opcode; /* simulator instruction break down */ - union { - unsigned long l; /* memory element break down */ - unsigned short s[2]; - unsigned char c[4]; - } mem; +struct mem_wrd +{ + /* Simulator instruction break down. */ + struct IR_FIELDS opcode; + union { + /* Memory element break down. */ + unsigned long l; + unsigned short s[2]; + unsigned char c[4]; + } mem; }; -#define MEMWRDSIZE (sizeof(struct mem_wrd)) /* size of each 32 bit memory model */ +/* Size of each 32 bit memory model. */ +#define MEMWRDSIZE (sizeof (struct mem_wrd)) + +extern struct mem_segs memory[]; +extern struct PROCESSOR m78000; -/* External declarations */ +struct PROCESSOR +{ + unsigned WORD + /* Execute instruction pointer. */ + ip, + /* Vector base register. */ + vbr, + /* Processor status register. */ + psr; + + /* Source 1. */ + WORD S1bus, + /* Source 2. */ + S2bus, + /* Destination. */ + Dbus, + /* Data address bus. */ + DAbus, + ALU, + /* Data registers. */ + Regs[REGs], + /* Max clocks before reg is available. */ + time_left[REGs], + /* Writeback priority of reg. */ + wb_pri[REGs], + /* Integer unit control regs. */ + SFU0_regs[REGs], + /* Floating point control regs. */ + SFU1_regs[REGs], + Scoreboard[REGs], + Vbr; + unsigned WORD scoreboard, + Psw, + Tpsw; + /* Waiting for a jump instruction. */ + FLAG jump_pending:1; +}; + +/* Size of immediate field. */ -extern struct mem_segs memory[]; -extern struct PROCESSOR m78000; +#define i26bit 1 +#define i16bit 2 +#define i10bit 3 + +/* Definitions for fields in psr. */ + +#define mode 31 +#define rbo 30 +#define ser 29 +#define carry 28 +#define sf7m 11 +#define sf6m 10 +#define sf5m 9 +#define sf4m 8 +#define sf3m 7 +#define sf2m 6 +#define sf1m 5 +#define mam 4 +#define inm 3 +#define exm 2 +#define trm 1 +#define ovfm 0 + +/* The 1 clock operations. */ + +#define ADDU 1 +#define ADDC 2 +#define ADDUC 3 +#define ADD 4 + +#define SUBU ADD+1 +#define SUBB ADD+2 +#define SUBUB ADD+3 +#define SUB ADD+4 + +#define AND_ ADD+5 +#define OR ADD+6 +#define XOR ADD+7 +#define CMP ADD+8 + +/* Loads. */ + +#define LDAB CMP+1 +#define LDAH CMP+2 +#define LDA CMP+3 +#define LDAD CMP+4 + +#define LDB LDAD+1 +#define LDH LDAD+2 +#define LD LDAD+3 +#define LDD LDAD+4 +#define LDBU LDAD+5 +#define LDHU LDAD+6 + +/* Stores. */ + +#define STB LDHU+1 +#define STH LDHU+2 +#define ST LDHU+3 +#define STD LDHU+4 + +/* Exchange. */ + +#define XMEMBU LDHU+5 +#define XMEM LDHU+6 + +/* Branches. */ + +#define JSR STD+1 +#define BSR STD+2 +#define BR STD+3 +#define JMP STD+4 +#define BB1 STD+5 +#define BB0 STD+6 +#define RTN STD+7 +#define BCND STD+8 + +/* Traps. */ + +#define TB1 BCND+1 +#define TB0 BCND+2 +#define TCND BCND+3 +#define RTE BCND+4 +#define TBND BCND+5 + +/* Misc. */ + +#define MUL TBND + 1 +#define DIV MUL +2 +#define DIVU MUL +3 +#define MASK MUL +4 +#define FF0 MUL +5 +#define FF1 MUL +6 +#define CLR MUL +7 +#define SET MUL +8 +#define EXT MUL +9 +#define EXTU MUL +10 +#define MAK MUL +11 +#define ROT MUL +12 + +/* Control register manipulations. */ + +#define LDCR ROT +1 +#define STCR ROT +2 +#define XCR ROT +3 + +#define FLDCR ROT +4 +#define FSTCR ROT +5 +#define FXCR ROT +6 + +#define NOP XCR +1 + +/* Floating point instructions. */ + +#define FADD NOP +1 +#define FSUB NOP +2 +#define FMUL NOP +3 +#define FDIV NOP +4 +#define FSQRT NOP +5 +#define FCMP NOP +6 +#define FIP NOP +7 +#define FLT NOP +8 +#define INT NOP +9 +#define NINT NOP +10 +#define TRNC NOP +11 +#define FLDC NOP +12 +#define FSTC NOP +13 +#define FXC NOP +14 + +#define UEXT(src,off,wid) \ + ((((unsigned int)(src)) >> (off)) & ((1 << (wid)) - 1)) + +#define SEXT(src,off,wid) \ + (((((int)(src))<<(32 - ((off) + (wid)))) >>(32 - (wid))) ) -struct PROCESSOR { - unsigned WORD - ip, /* execute instruction pointer */ - vbr, /* vector base register */ - psr; /* processor status register */ - - WORD S1bus, /* source 1 */ - S2bus, /* source 2 */ - Dbus, /* destination */ - DAbus, /* data address bus */ - ALU, - Regs[REGs], /* data registers */ - time_left[REGs], /* max clocks before reg is available */ - wb_pri[REGs], /* writeback priority of reg */ - SFU0_regs[REGs], /* integer unit control regs */ - SFU1_regs[REGs], /* floating point control regs */ - Scoreboard[REGs], - Vbr; - unsigned WORD scoreboard, - Psw, - Tpsw; - FLAG jump_pending:1; /* waiting for a jump instr. */ - }; - -# define i26bit 1 /* size of immediate field */ -# define i16bit 2 -# define i10bit 3 - -/* Definitions for fields in psr */ - -# define mode 31 -# define rbo 30 -# define ser 29 -# define carry 28 -# define sf7m 11 -# define sf6m 10 -# define sf5m 9 -# define sf4m 8 -# define sf3m 7 -# define sf2m 6 -# define sf1m 5 -# define mam 4 -# define inm 3 -# define exm 2 -# define trm 1 -# define ovfm 0 - -#define MODEMASK (1<<(mode-1)) -# define SILENT 0 /* simulate without output to crt */ -# define VERBOSE 1 /* simulate in verbose mode */ -# define PR_INSTR 2 /* only print instructions */ - -# define RESET 16 /* reset phase */ - -# define PHASE1 0 /* data path phases */ -# define PHASE2 1 - -/* the 1 clock operations */ - -# define ADDU 1 -# define ADDC 2 -# define ADDUC 3 -# define ADD 4 - -# define SUBU ADD+1 -# define SUBB ADD+2 -# define SUBUB ADD+3 -# define SUB ADD+4 - -# define AND_ ADD+5 -# define OR ADD+6 -# define XOR ADD+7 -# define CMP ADD+8 - -/* the LOADS */ - -# define LDAB CMP+1 -# define LDAH CMP+2 -# define LDA CMP+3 -# define LDAD CMP+4 - -# define LDB LDAD+1 -# define LDH LDAD+2 -# define LD LDAD+3 -# define LDD LDAD+4 -# define LDBU LDAD+5 -# define LDHU LDAD+6 - -/* the STORES */ - -# define STB LDHU+1 -# define STH LDHU+2 -# define ST LDHU+3 -# define STD LDHU+4 - -/* the exchange */ - -# define XMEMBU LDHU+5 -# define XMEM LDHU+6 - -/* the branches */ -# define JSR STD+1 -# define BSR STD+2 -# define BR STD+3 -# define JMP STD+4 -# define BB1 STD+5 -# define BB0 STD+6 -# define RTN STD+7 -# define BCND STD+8 - -/* the TRAPS */ -# define TB1 BCND+1 -# define TB0 BCND+2 -# define TCND BCND+3 -# define RTE BCND+4 -# define TBND BCND+5 - -/* the MISC instructions */ -# define MUL TBND + 1 -# define DIV MUL +2 -# define DIVU MUL +3 -# define MASK MUL +4 -# define FF0 MUL +5 -# define FF1 MUL +6 -# define CLR MUL +7 -# define SET MUL +8 -# define EXT MUL +9 -# define EXTU MUL +10 -# define MAK MUL +11 -# define ROT MUL +12 - -/* control register manipulations */ - -# define LDCR ROT +1 -# define STCR ROT +2 -# define XCR ROT +3 - -# define FLDCR ROT +4 -# define FSTCR ROT +5 -# define FXCR ROT +6 - - -# define NOP XCR +1 - -/* floating point instructions */ - -# define FADD NOP +1 -# define FSUB NOP +2 -# define FMUL NOP +3 -# define FDIV NOP +4 -# define FSQRT NOP +5 -# define FCMP NOP +6 -# define FIP NOP +7 -# define FLT NOP +8 -# define INT NOP +9 -# define NINT NOP +10 -# define TRNC NOP +11 -# define FLDC NOP +12 -# define FSTC NOP +13 -# define FXC NOP +14 - -# define UEXT(src,off,wid) ((((unsigned int)(src))>>(off)) & ((1<<(wid)) - 1)) -# define SEXT(src,off,wid) (((((int)(src))<<(32-((off)+(wid)))) >>(32-(wid))) ) -# define MAKE(src,off,wid) \ - ((((unsigned int)(src)) & ((1<<(wid)) - 1)) << (off)) +#define MAKE(src,off,wid) \ + ((((unsigned int)(src)) & ((1 << (wid)) - 1)) << (off)) -# define opword(n) (unsigned long) (memaddr->mem.l) +#define opword(n) (unsigned long) (memaddr->mem.l) -/* Constants and Masks */ +/* Constants and masks. */ #define SFU0 0x80000000 #define SFU1 0x84000000 @@ -360,72 +378,68 @@ struct PROCESSOR { #define CTRL 0x0000f000 #define CTRLMASK 0xfc00f800 -/* Operands types */ +/* Operands types. */ -enum operand_type { +enum operand_type +{ HEX = 1, REG = 2, CONT = 3, IND = 3, BF = 4, - REGSC = 5 /* scaled register */, - CRREG = 6 /* control register */, - FCRREG = 7 /* floating point control register */, + /* Scaled register. */ + REGSC = 5, + /* Control register. */ + CRREG = 6, + /* Floating point control register. */ + FCRREG = 7, PCREL = 8, CONDMASK = 9, - XREG = 10, /* extended register */ - DEC = 11, /* decimal */ + /* Extended register. */ + XREG = 10, + /* Decimal. */ + DEC = 11 }; -/* Hashing Specification */ +/* Hashing specification. */ #define HASHVAL 79 -/* Type definitions */ - -typedef unsigned int UINT; - -/* Structure templates */ - -#if never -typedef struct { - unsigned int offset:5; - unsigned int width:6; - unsigned int type:5; -} OPSPEC; -#endif +/* Structure templates. */ -typedef struct { - unsigned int offset; - unsigned int width; - enum operand_type type; +typedef struct +{ + unsigned int offset; + unsigned int width; + enum operand_type type; } OPSPEC; - struct SIM_FLAGS { - int ltncy, /* latency (max number of clocks needed to execute) */ - extime, /* execution time (min number of clocks needed to execute) */ - wb_pri; /* writeback slot priority */ - unsigned op:OP, /* simulator version of opcode */ - imm_flags:2, /* 10,16 or 26 bit immediate flags */ - rs1_used:1, /* register source 1 used */ - rs2_used:1, /* register source 2 used */ - rsd_used:1, /* register source/dest used */ - c_flag:1, /* complement */ - u_flag:1, /* upper half word */ - n_flag:1, /* execute next */ - wb_flag:1, /* uses writeback slot */ - dest_64:1, /* double precision dest */ - s1_64:1, /* double precision source 1 */ - s2_64:1, /* double precision source 2 */ - scale_flag:1; /* register is scaled */ +struct SIM_FLAGS +{ + int ltncy, /* latency (max number of clocks needed to execute). */ + extime, /* execution time (min number of clocks needed to execute). */ + wb_pri; /* writeback slot priority. */ + unsigned op:OP, /* simulator version of opcode. */ + imm_flags:2, /* 10,16 or 26 bit immediate flags. */ + rs1_used:1, /* register source 1 used. */ + rs2_used:1, /* register source 2 used. */ + rsd_used:1, /* register source/dest used. */ + c_flag:1, /* complement. */ + u_flag:1, /* upper half word. */ + n_flag:1, /* execute next. */ + wb_flag:1, /* uses writeback slot. */ + dest_64:1, /* double precision dest. */ + s1_64:1, /* double precision source 1. */ + s2_64:1, /* double precision source 2. */ + scale_flag:1; /* register is scaled. */ }; typedef struct INSTRUCTAB { - unsigned int opcode; - char *mnemonic; - OPSPEC op1,op2,op3; - struct SIM_FLAGS flgs; - struct INSTRUCTAB *next; + unsigned int opcode; + char *mnemonic; + OPSPEC op1,op2,op3; + struct SIM_FLAGS flgs; + struct INSTRUCTAB *next; } INSTAB; @@ -623,7 +637,7 @@ static INSTAB instructions[] = { {0x80008800,"fstcr ",{16,5,REG} ,{5,6,FCRREG} ,NO_OPERAND , {1,1,PFLT,FSTC ,0,0,1,1,0,0,0,1,0,0,0,0} ,NULL }, {0x8000c800,"fxcr ",{21,5,REG} ,{16,5,REG} ,{5,6,FCRREG} , {1,1,PFLT,FXC ,0,0,1,1,0,0,0,1,0,0,0,0} ,NULL }, -/* The following are new for the 88110. */ +/* The following are new for the 88110. */ {0x8400aaa0,"fadd.ddd ",{21,5,XREG} ,{16,5,XREG} ,{0,5,XREG}, {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}, NULL }, {0x8400aa80,"fadd.dds ",{21,5,XREG} ,{16,5,XREG} ,{0,5,XREG}, {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}, NULL }, diff -uprN binutils-2.11.90.0.23/include/xregex.h binutils-2.11.90.0.24/include/xregex.h --- binutils-2.11.90.0.23/include/xregex.h Wed Dec 31 16:00:00 1969 +++ binutils-2.11.90.0.24/include/xregex.h Sat Jul 14 10:29:15 2001 @@ -0,0 +1,24 @@ +/* This file redefines all regex external names before including + a renamed copy of glibc's regex.h. */ + +#ifndef _XREGEX_H +#define _XREGEX_H 1 + +# define regfree xregfree +# define regexec xregexec +# define regcomp xregcomp +# define regerror xregerror +# define re_set_registers xre_set_registers +# define re_match_2 xre_match_2 +# define re_match xre_match +# define re_search xre_search +# define re_compile_pattern xre_compile_pattern +# define re_set_syntax xre_set_syntax +# define re_search_2 xre_search_2 +# define re_compile_fastmap xre_compile_fastmap +# define re_syntax_options xre_syntax_options +# define re_max_failures xre_max_failures + +#include "xregex2.h" + +#endif /* xregex.h */ diff -uprN binutils-2.11.90.0.23/include/xregex2.h binutils-2.11.90.0.24/include/xregex2.h --- binutils-2.11.90.0.23/include/xregex2.h Wed Dec 31 16:00:00 1969 +++ binutils-2.11.90.0.24/include/xregex2.h Sat Jul 14 10:29:15 2001 @@ -0,0 +1,566 @@ +/* Definitions for data structures and routines for the regular + expression library, version 0.12. + Copyright (C) 1985,1989-1993,1995-1998, 2000 Free Software Foundation, Inc. + + This file is part of the GNU C Library. Its master source is NOT part of + the C library, however. The master source lives in /gd/gnu/lib. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +#ifndef _REGEX_H +#define _REGEX_H 1 + +/* Allow the use in C++ code. */ +#ifdef __cplusplus +extern "C" { +#endif + +/* POSIX says that must be included (by the caller) before + . */ + +#if !defined _POSIX_C_SOURCE && !defined _POSIX_SOURCE && defined VMS +/* VMS doesn't have `size_t' in , even though POSIX says it + should be there. */ +# include +#endif + +/* The following two types have to be signed and unsigned integer type + wide enough to hold a value of a pointer. For most ANSI compilers + ptrdiff_t and size_t should be likely OK. Still size of these two + types is 2 for Microsoft C. Ugh... */ +typedef long int s_reg_t; +typedef unsigned long int active_reg_t; + +/* The following bits are used to determine the regexp syntax we + recognize. The set/not-set meanings are chosen so that Emacs syntax + remains the value 0. The bits are given in alphabetical order, and + the definitions shifted by one from the previous bit; thus, when we + add or remove a bit, only one other definition need change. */ +typedef unsigned long int reg_syntax_t; + +/* If this bit is not set, then \ inside a bracket expression is literal. + If set, then such a \ quotes the following character. */ +#define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1) + +/* If this bit is not set, then + and ? are operators, and \+ and \? are + literals. + If set, then \+ and \? are operators and + and ? are literals. */ +#define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1) + +/* If this bit is set, then character classes are supported. They are: + [:alpha:], [:upper:], [:lower:], [:digit:], [:alnum:], [:xdigit:], + [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:]. + If not set, then character classes are not supported. */ +#define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1) + +/* If this bit is set, then ^ and $ are always anchors (outside bracket + expressions, of course). + If this bit is not set, then it depends: + ^ is an anchor if it is at the beginning of a regular + expression or after an open-group or an alternation operator; + $ is an anchor if it is at the end of a regular expression, or + before a close-group or an alternation operator. + + This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because + POSIX draft 11.2 says that * etc. in leading positions is undefined. + We already implemented a previous draft which made those constructs + invalid, though, so we haven't changed the code back. */ +#define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1) + +/* If this bit is set, then special characters are always special + regardless of where they are in the pattern. + If this bit is not set, then special characters are special only in + some contexts; otherwise they are ordinary. Specifically, + * + ? and intervals are only special when not after the beginning, + open-group, or alternation operator. */ +#define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1) + +/* If this bit is set, then *, +, ?, and { cannot be first in an re or + immediately after an alternation or begin-group operator. */ +#define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1) + +/* If this bit is set, then . matches newline. + If not set, then it doesn't. */ +#define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1) + +/* If this bit is set, then . doesn't match NUL. + If not set, then it does. */ +#define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1) + +/* If this bit is set, nonmatching lists [^...] do not match newline. + If not set, they do. */ +#define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1) + +/* If this bit is set, either \{...\} or {...} defines an + interval, depending on RE_NO_BK_BRACES. + If not set, \{, \}, {, and } are literals. */ +#define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1) + +/* If this bit is set, +, ? and | aren't recognized as operators. + If not set, they are. */ +#define RE_LIMITED_OPS (RE_INTERVALS << 1) + +/* If this bit is set, newline is an alternation operator. + If not set, newline is literal. */ +#define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1) + +/* If this bit is set, then `{...}' defines an interval, and \{ and \} + are literals. + If not set, then `\{...\}' defines an interval. */ +#define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1) + +/* If this bit is set, (...) defines a group, and \( and \) are literals. + If not set, \(...\) defines a group, and ( and ) are literals. */ +#define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1) + +/* If this bit is set, then \ matches . + If not set, then \ is a back-reference. */ +#define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1) + +/* If this bit is set, then | is an alternation operator, and \| is literal. + If not set, then \| is an alternation operator, and | is literal. */ +#define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1) + +/* If this bit is set, then an ending range point collating higher + than the starting range point, as in [z-a], is invalid. + If not set, then when ending range point collates higher than the + starting range point, the range is ignored. */ +#define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1) + +/* If this bit is set, then an unmatched ) is ordinary. + If not set, then an unmatched ) is invalid. */ +#define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1) + +/* If this bit is set, succeed as soon as we match the whole pattern, + without further backtracking. */ +#define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1) + +/* If this bit is set, do not process the GNU regex operators. + If not set, then the GNU regex operators are recognized. */ +#define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1) + +/* If this bit is set, turn on internal regex debugging. + If not set, and debugging was on, turn it off. + This only works if regex.c is compiled -DDEBUG. + We define this bit always, so that all that's needed to turn on + debugging is to recompile regex.c; the calling code can always have + this bit set, and it won't affect anything in the normal case. */ +#define RE_DEBUG (RE_NO_GNU_OPS << 1) + +/* If this bit is set, a syntactically invalid interval is treated as + a string of ordinary characters. For example, the ERE 'a{1' is + treated as 'a\{1'. */ +#define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1) + +/* This global variable defines the particular regexp syntax to use (for + some interfaces). When a regexp is compiled, the syntax used is + stored in the pattern buffer, so changing this does not affect + already-compiled regexps. */ +extern reg_syntax_t re_syntax_options; + +/* Define combinations of the above bits for the standard possibilities. + (The [[[ comments delimit what gets put into the Texinfo file, so + don't delete them!) */ +/* [[[begin syntaxes]]] */ +#define RE_SYNTAX_EMACS 0 + +#define RE_SYNTAX_AWK \ + (RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL \ + | RE_NO_BK_PARENS | RE_NO_BK_REFS \ + | RE_NO_BK_VBAR | RE_NO_EMPTY_RANGES \ + | RE_DOT_NEWLINE | RE_CONTEXT_INDEP_ANCHORS \ + | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS) + +#define RE_SYNTAX_GNU_AWK \ + ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG) \ + & ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS)) + +#define RE_SYNTAX_POSIX_AWK \ + (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \ + | RE_INTERVALS | RE_NO_GNU_OPS) + +#define RE_SYNTAX_GREP \ + (RE_BK_PLUS_QM | RE_CHAR_CLASSES \ + | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS \ + | RE_NEWLINE_ALT) + +#define RE_SYNTAX_EGREP \ + (RE_CHAR_CLASSES | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE \ + | RE_NEWLINE_ALT | RE_NO_BK_PARENS \ + | RE_NO_BK_VBAR) + +#define RE_SYNTAX_POSIX_EGREP \ + (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES \ + | RE_INVALID_INTERVAL_ORD) + +/* P1003.2/D11.2, section 4.20.7.1, lines 5078ff. */ +#define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC + +#define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC + +/* Syntax bits common to both basic and extended POSIX regex syntax. */ +#define _RE_SYNTAX_POSIX_COMMON \ + (RE_CHAR_CLASSES | RE_DOT_NEWLINE | RE_DOT_NOT_NULL \ + | RE_INTERVALS | RE_NO_EMPTY_RANGES) + +#define RE_SYNTAX_POSIX_BASIC \ + (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM) + +/* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes + RE_LIMITED_OPS, i.e., \? \+ \| are not recognized. Actually, this + isn't minimal, since other operators, such as \`, aren't disabled. */ +#define RE_SYNTAX_POSIX_MINIMAL_BASIC \ + (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS) + +#define RE_SYNTAX_POSIX_EXTENDED \ + (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INDEP_OPS | RE_NO_BK_BRACES \ + | RE_NO_BK_PARENS | RE_NO_BK_VBAR \ + | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD) + +/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is + removed and RE_NO_BK_REFS is added. */ +#define RE_SYNTAX_POSIX_MINIMAL_EXTENDED \ + (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES \ + | RE_NO_BK_PARENS | RE_NO_BK_REFS \ + | RE_NO_BK_VBAR | RE_UNMATCHED_RIGHT_PAREN_ORD) +/* [[[end syntaxes]]] */ + +/* Maximum number of duplicates an interval can allow. Some systems + (erroneously) define this in other header files, but we want our + value, so remove any previous define. */ +#ifdef RE_DUP_MAX +# undef RE_DUP_MAX +#endif +/* If sizeof(int) == 2, then ((1 << 15) - 1) overflows. */ +#define RE_DUP_MAX (0x7fff) + + +/* POSIX `cflags' bits (i.e., information for `regcomp'). */ + +/* If this bit is set, then use extended regular expression syntax. + If not set, then use basic regular expression syntax. */ +#define REG_EXTENDED 1 + +/* If this bit is set, then ignore case when matching. + If not set, then case is significant. */ +#define REG_ICASE (REG_EXTENDED << 1) + +/* If this bit is set, then anchors do not match at newline + characters in the string. + If not set, then anchors do match at newlines. */ +#define REG_NEWLINE (REG_ICASE << 1) + +/* If this bit is set, then report only success or fail in regexec. + If not set, then returns differ between not matching and errors. */ +#define REG_NOSUB (REG_NEWLINE << 1) + + +/* POSIX `eflags' bits (i.e., information for regexec). */ + +/* If this bit is set, then the beginning-of-line operator doesn't match + the beginning of the string (presumably because it's not the + beginning of a line). + If not set, then the beginning-of-line operator does match the + beginning of the string. */ +#define REG_NOTBOL 1 + +/* Like REG_NOTBOL, except for the end-of-line. */ +#define REG_NOTEOL (1 << 1) + + +/* If any error codes are removed, changed, or added, update the + `re_error_msg' table in regex.c. */ +typedef enum +{ +#ifdef _XOPEN_SOURCE + REG_ENOSYS = -1, /* This will never happen for this implementation. */ +#endif + + REG_NOERROR = 0, /* Success. */ + REG_NOMATCH, /* Didn't find a match (for regexec). */ + + /* POSIX regcomp return error codes. (In the order listed in the + standard.) */ + REG_BADPAT, /* Invalid pattern. */ + REG_ECOLLATE, /* Not implemented. */ + REG_ECTYPE, /* Invalid character class name. */ + REG_EESCAPE, /* Trailing backslash. */ + REG_ESUBREG, /* Invalid back reference. */ + REG_EBRACK, /* Unmatched left bracket. */ + REG_EPAREN, /* Parenthesis imbalance. */ + REG_EBRACE, /* Unmatched \{. */ + REG_BADBR, /* Invalid contents of \{\}. */ + REG_ERANGE, /* Invalid range end. */ + REG_ESPACE, /* Ran out of memory. */ + REG_BADRPT, /* No preceding re for repetition op. */ + + /* Error codes we've added. */ + REG_EEND, /* Premature end. */ + REG_ESIZE, /* Compiled pattern bigger than 2^16 bytes. */ + REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */ +} reg_errcode_t; + +/* This data structure represents a compiled pattern. Before calling + the pattern compiler, the fields `buffer', `allocated', `fastmap', + `translate', and `no_sub' can be set. After the pattern has been + compiled, the `re_nsub' field is available. All other fields are + private to the regex routines. */ + +#ifndef RE_TRANSLATE_TYPE +# define RE_TRANSLATE_TYPE char * +#endif + +struct re_pattern_buffer +{ +/* [[[begin pattern_buffer]]] */ + /* Space that holds the compiled pattern. It is declared as + `unsigned char *' because its elements are + sometimes used as array indexes. */ + unsigned char *buffer; + + /* Number of bytes to which `buffer' points. */ + unsigned long int allocated; + + /* Number of bytes actually used in `buffer'. */ + unsigned long int used; + + /* Syntax setting with which the pattern was compiled. */ + reg_syntax_t syntax; + + /* Pointer to a fastmap, if any, otherwise zero. re_search uses + the fastmap, if there is one, to skip over impossible + starting points for matches. */ + char *fastmap; + + /* Either a translate table to apply to all characters before + comparing them, or zero for no translation. The translation + is applied to a pattern when it is compiled and to a string + when it is matched. */ + RE_TRANSLATE_TYPE translate; + + /* Number of subexpressions found by the compiler. */ + size_t re_nsub; + + /* Zero if this pattern cannot match the empty string, one else. + Well, in truth it's used only in `re_search_2', to see + whether or not we should use the fastmap, so we don't set + this absolutely perfectly; see `re_compile_fastmap' (the + `duplicate' case). */ + unsigned can_be_null : 1; + + /* If REGS_UNALLOCATED, allocate space in the `regs' structure + for `max (RE_NREGS, re_nsub + 1)' groups. + If REGS_REALLOCATE, reallocate space if necessary. + If REGS_FIXED, use what's there. */ +#define REGS_UNALLOCATED 0 +#define REGS_REALLOCATE 1 +#define REGS_FIXED 2 + unsigned regs_allocated : 2; + + /* Set to zero when `regex_compile' compiles a pattern; set to one + by `re_compile_fastmap' if it updates the fastmap. */ + unsigned fastmap_accurate : 1; + + /* If set, `re_match_2' does not return information about + subexpressions. */ + unsigned no_sub : 1; + + /* If set, a beginning-of-line anchor doesn't match at the + beginning of the string. */ + unsigned not_bol : 1; + + /* Similarly for an end-of-line anchor. */ + unsigned not_eol : 1; + + /* If true, an anchor at a newline matches. */ + unsigned newline_anchor : 1; + +/* [[[end pattern_buffer]]] */ +}; + +typedef struct re_pattern_buffer regex_t; + +/* Type for byte offsets within the string. POSIX mandates this. */ +typedef int regoff_t; + + +/* This is the structure we store register match data in. See + regex.texinfo for a full description of what registers match. */ +struct re_registers +{ + unsigned num_regs; + regoff_t *start; + regoff_t *end; +}; + + +/* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer, + `re_match_2' returns information about at least this many registers + the first time a `regs' structure is passed. */ +#ifndef RE_NREGS +# define RE_NREGS 30 +#endif + + +/* POSIX specification for registers. Aside from the different names than + `re_registers', POSIX uses an array of structures, instead of a + structure of arrays. */ +typedef struct +{ + regoff_t rm_so; /* Byte offset from string's start to substring's start. */ + regoff_t rm_eo; /* Byte offset from string's start to substring's end. */ +} regmatch_t; + +/* Declarations for routines. */ + +/* To avoid duplicating every routine declaration -- once with a + prototype (if we are ANSI), and once without (if we aren't) -- we + use the following macro to declare argument types. This + unfortunately clutters up the declarations a bit, but I think it's + worth it. */ + +#if __STDC__ + +# define _RE_ARGS(args) args + +#else /* not __STDC__ */ + +# define _RE_ARGS(args) () + +#endif /* not __STDC__ */ + +/* Sets the current default syntax to SYNTAX, and return the old syntax. + You can also simply assign to the `re_syntax_options' variable. */ +extern reg_syntax_t re_set_syntax _RE_ARGS ((reg_syntax_t syntax)); + +/* Compile the regular expression PATTERN, with length LENGTH + and syntax given by the global `re_syntax_options', into the buffer + BUFFER. Return NULL if successful, and an error string if not. */ +extern const char *re_compile_pattern + _RE_ARGS ((const char *pattern, size_t length, + struct re_pattern_buffer *buffer)); + + +/* Compile a fastmap for the compiled pattern in BUFFER; used to + accelerate searches. Return 0 if successful and -2 if was an + internal error. */ +extern int re_compile_fastmap _RE_ARGS ((struct re_pattern_buffer *buffer)); + + +/* Search in the string STRING (with length LENGTH) for the pattern + compiled into BUFFER. Start searching at position START, for RANGE + characters. Return the starting position of the match, -1 for no + match, or -2 for an internal error. Also return register + information in REGS (if REGS and BUFFER->no_sub are nonzero). */ +extern int re_search + _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, + int length, int start, int range, struct re_registers *regs)); + + +/* Like `re_search', but search in the concatenation of STRING1 and + STRING2. Also, stop searching at index START + STOP. */ +extern int re_search_2 + _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1, + int length1, const char *string2, int length2, + int start, int range, struct re_registers *regs, int stop)); + + +/* Like `re_search', but return how many characters in STRING the regexp + in BUFFER matched, starting at position START. */ +extern int re_match + _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, + int length, int start, struct re_registers *regs)); + + +/* Relates to `re_match' as `re_search_2' relates to `re_search'. */ +extern int re_match_2 + _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1, + int length1, const char *string2, int length2, + int start, struct re_registers *regs, int stop)); + + +/* Set REGS to hold NUM_REGS registers, storing them in STARTS and + ENDS. Subsequent matches using BUFFER and REGS will use this memory + for recording register information. STARTS and ENDS must be + allocated with malloc, and must each be at least `NUM_REGS * sizeof + (regoff_t)' bytes long. + + If NUM_REGS == 0, then subsequent matches should allocate their own + register data. + + Unless this function is called, the first search or match using + PATTERN_BUFFER will allocate its own register data, without + freeing the old data. */ +extern void re_set_registers + _RE_ARGS ((struct re_pattern_buffer *buffer, struct re_registers *regs, + unsigned num_regs, regoff_t *starts, regoff_t *ends)); + +#if defined _REGEX_RE_COMP || defined _LIBC +# ifndef _CRAY +/* 4.2 bsd compatibility. */ +extern char *re_comp _RE_ARGS ((const char *)); +extern int re_exec _RE_ARGS ((const char *)); +# endif +#endif + +/* GCC 2.95 and later have "__restrict"; C99 compilers have + "restrict", and "configure" may have defined "restrict". */ +#ifndef __restrict +# if ! (2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__)) +# if defined restrict || 199901L <= __STDC_VERSION__ +# define __restrict restrict +# else +# define __restrict +# endif +# endif +#endif +/* For now unconditionally define __restrict_arr to expand to nothing. + Ideally we would have a test for the compiler which allows defining + it to restrict. */ +#define __restrict_arr + +/* POSIX compatibility. */ +extern int regcomp _RE_ARGS ((regex_t *__restrict __preg, + const char *__restrict __pattern, + int __cflags)); + +extern int regexec _RE_ARGS ((const regex_t *__restrict __preg, + const char *__restrict __string, size_t __nmatch, + regmatch_t __pmatch[__restrict_arr], + int __eflags)); + +extern size_t regerror _RE_ARGS ((int __errcode, const regex_t *__preg, + char *__errbuf, size_t __errbuf_size)); + +extern void regfree _RE_ARGS ((regex_t *__preg)); + + +#ifdef __cplusplus +} +#endif /* C++ */ + +#endif /* regex.h */ + +/* +Local variables: +make-backup-files: t +version-control: t +trim-versions-without-asking: nil +End: +*/ diff -uprN binutils-2.11.90.0.23/ld/ChangeLog binutils-2.11.90.0.24/ld/ChangeLog --- binutils-2.11.90.0.23/ld/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,44 @@ +2001-07-14 H.J. Lu + + * emultempl/elf32.em (output_prev_sec_find): Never return + bfd_abs_section_ptr, bfd_com_section_ptr nor + bfd_und_section_ptr. + +2001-07-14 Nick Clifton + + * Makefile.am (em32relf.c): Chnage dependencu from generic.em to + elf32.em. + * Makefile.in: Regenerate. + +2001-07-14 matthew green + + * configure.tgt (i386-*-netbsdelf*): New target. + (i386-*-netbsd*): Set targ_extra_emuls to `elf_i386' + +2001-07-13 Jakub Jelinek + + * emultempl/elf32.em (output_prev_sec_find): New. + (place_orphan): Use it. + +2001-07-11 H.J. Lu + + * ldmain.c (main): Fix typos in the last change. + +2001-07-11 Jakub Jelinek + + * ldmain.c (main): Disallow -F and -f without -shared. + +2001-07-11 Nick Clifton + + * emultempl/pe.em (after_open): Check for the output_bfd not + having any coff_data structure allocated to it. + +2001-07-09 David O'Brien + + * emultempl/elf32.em: Do not assuming that contents of the buffer + returned from basename function will remain unchanged accross other + function calls. + 2001-07-03 H.J. Lu * scripttempl/elf.sc (DYNAMIC_PAD): Revert the change made on diff -uprN binutils-2.11.90.0.23/ld/Makefile.am binutils-2.11.90.0.24/ld/Makefile.am --- binutils-2.11.90.0.23/ld/Makefile.am Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/Makefile.am Sat Jul 14 10:29:15 2001 @@ -439,7 +439,7 @@ eelf32mcore.c: $(srcdir)/emulparams/elf3 $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} ${GENSCRIPTS} elf32mcore "$(tdir_mcore)" em32relf.c: $(srcdir)/emulparams/m32relf.sh \ - $(srcdir)/emultempl/generic.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} + $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} ${GENSCRIPTS} m32relf "$(tdir_m32r)" eelf32_sparc.c: $(srcdir)/emulparams/elf32_sparc.sh \ $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} diff -uprN binutils-2.11.90.0.23/ld/Makefile.in binutils-2.11.90.0.24/ld/Makefile.in --- binutils-2.11.90.0.23/ld/Makefile.in Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/Makefile.in Sat Jul 14 10:29:15 2001 @@ -476,7 +476,7 @@ deffilep.c ldgram.c ldlex.c DISTFILES = $(DIST_COMMON) $(SOURCES) $(HEADERS) $(TEXINFOS) $(EXTRA_DIST) -TAR = tar +TAR = gtar GZIP_ENV = --best SOURCES = $(ld_new_SOURCES) $(EXTRA_ld_new_SOURCES) OBJECTS = $(ld_new_OBJECTS) @@ -1153,7 +1153,7 @@ eelf32mcore.c: $(srcdir)/emulparams/elf3 $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} ${GENSCRIPTS} elf32mcore "$(tdir_mcore)" em32relf.c: $(srcdir)/emulparams/m32relf.sh \ - $(srcdir)/emultempl/generic.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} + $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} ${GENSCRIPTS} m32relf "$(tdir_m32r)" eelf32_sparc.c: $(srcdir)/emulparams/elf32_sparc.sh \ $(srcdir)/emultempl/elf32.em $(srcdir)/scripttempl/elf.sc ${GEN_DEPENDS} diff -uprN binutils-2.11.90.0.23/ld/configure.tgt binutils-2.11.90.0.24/ld/configure.tgt --- binutils-2.11.90.0.23/ld/configure.tgt Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/configure.tgt Sat Jul 14 10:29:15 2001 @@ -150,7 +150,12 @@ i[3456]86-*-unixware) targ_emul=elf_i386 i[3456]86-*-solaris*) targ_emul=elf_i386_ldso targ_extra_emuls="elf_i386" ;; -i[3456]86-*-netbsd*) targ_emul=i386nbsd ;; +i[3456]86-*-netbsdelf*) targ_emul=elf_i386 + targ_extra_emuls=i386nbsd + ;; +i[3456]86-*-netbsd*) targ_emul=i386nbsd + targ_extra_emuls=elf_i386 + ;; i[3456]86-*-netware) targ_emul=i386nw ;; i[3456]86-*-elf*) targ_emul=elf_i386 ;; i[3456]86-*-freebsdaout* | i[3456]86-*-freebsd[12].* | i[3456]86-*-freebsd[12]) diff -uprN binutils-2.11.90.0.23/ld/emultempl/elf32.em binutils-2.11.90.0.24/ld/emultempl/elf32.em --- binutils-2.11.90.0.23/ld/emultempl/elf32.em Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/emultempl/elf32.em Sat Jul 14 10:29:15 2001 @@ -360,6 +360,9 @@ cat >>e${EMULATION_NAME}.c <filename); + /* Tell the ELF backend that the output file needs a DT_NEEDED entry for this file if it is used to resolve the reference in a regular object. */ @@ -1019,6 +1022,38 @@ EOF if test x"$LDEMUL_PLACE_ORPHAN" != xgld"$EMULATION_NAME"_place_orphan; then cat >>e${EMULATION_NAME}.c <next) + { + lookup = &u->output_section_statement; + if (lookup == os) + break; + if (lookup->bfd_section != NULL + && lookup->bfd_section != bfd_abs_section_ptr + && lookup->bfd_section != bfd_com_section_ptr + && lookup->bfd_section != bfd_und_section_ptr) + s = lookup->bfd_section; + } + + if (u == NULL) + return NULL; + + return s; +} + + /* Place an orphan section. We use this to put random SHF_ALLOC sections in the right segment. */ @@ -1205,12 +1240,16 @@ gld${EMULATION_NAME}_place_orphan (file, if (place != NULL) { - asection *snew, **pps; + asection *snew, **pps, *bfd_section; snew = os->bfd_section; + bfd_section = place->os->bfd_section; + if (place->section == NULL && bfd_section == NULL) + bfd_section = output_prev_sec_find (place->os); + if (place->section != NULL - || (place->os->bfd_section != NULL - && place->os->bfd_section != snew)) + || (bfd_section != NULL + && bfd_section != snew)) { /* Shuffle the section to make the output file look neater. This is really only cosmetic. */ @@ -1219,15 +1258,15 @@ gld${EMULATION_NAME}_place_orphan (file, #if 0 /* Finding the end of the list is a little tricky. We make a wild stab at it by comparing section flags. */ - flagword first_flags = place->os->bfd_section->flags; - for (pps = &place->os->bfd_section->next; + flagword first_flags = bfd_section->flags; + for (pps = &bfd_section->next; *pps != NULL && (*pps)->flags == first_flags; pps = &(*pps)->next) ; place->section = pps; #else /* Put orphans after the first section on the list. */ - place->section = &place->os->bfd_section->next; + place->section = &bfd_section->next; #endif } diff -uprN binutils-2.11.90.0.23/ld/emultempl/pe.em binutils-2.11.90.0.24/ld/emultempl/pe.em --- binutils-2.11.90.0.23/ld/emultempl/pe.em Sat Jun 9 00:13:01 2001 +++ binutils-2.11.90.0.24/ld/emultempl/pe.em Sat Jul 14 10:29:15 2001 @@ -800,7 +800,7 @@ gld_${EMULATION_NAME}_after_open () FIXME: This should be done via a function, rather than by including an internal BFD header. */ - if (!coff_data (output_bfd)->pe) + if (coff_data (output_bfd) == NULL || coff_data (output_bfd)->pe == NULL) einfo (_("%F%P: PE operations on non PE file.\n")); pe_data (output_bfd)->pe_opthdr = pe; diff -uprN binutils-2.11.90.0.23/ld/ldmain.c binutils-2.11.90.0.24/ld/ldmain.c --- binutils-2.11.90.0.23/ld/ldmain.c Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/ld/ldmain.c Sat Jul 14 10:29:15 2001 @@ -274,6 +274,14 @@ main (argc, argv) einfo (_("%P%F: -r and -shared may not be used together\n")); } + if (! link_info.shared) + { + if (command_line.filter_shlib) + einfo (_("%P%F: -F may not be used without -shared\n")); + if (command_line.auxiliary_filters) + einfo (_("%P%F: -f may not be used without -shared\n")); + } + /* Treat ld -r -s as ld -r -S -x (i.e., strip all local symbols). I don't see how else this can be handled, since in this case we must preserve all externally visible symbols. */ diff -uprN binutils-2.11.90.0.23/ld/testsuite/ChangeLog binutils-2.11.90.0.24/ld/testsuite/ChangeLog --- binutils-2.11.90.0.23/ld/testsuite/ChangeLog Tue Jun 19 11:57:59 2001 +++ binutils-2.11.90.0.24/ld/testsuite/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,8 @@ +2001-07-12 H.J. Lu + + * ld-selective/selective.exp: Mark selective1, selective2, + selective4 and selective5 xfail on alpha*-*. + 2001-06-13 Hans-Peter Nilsson * config/default.exp (AS, GASP, OBJDUMP, NM, NMFLAGS, OBJCOPY, diff -uprN binutils-2.11.90.0.23/ld/testsuite/ld-selective/selective.exp binutils-2.11.90.0.24/ld/testsuite/ld-selective/selective.exp --- binutils-2.11.90.0.23/ld/testsuite/ld-selective/selective.exp Sat Jun 9 00:13:02 2001 +++ binutils-2.11.90.0.24/ld/testsuite/ld-selective/selective.exp Sat Jul 14 10:29:15 2001 @@ -38,11 +38,11 @@ if {[istarget "*-*-pe"]} { # # FIXME: Instead of table, read settings from each source-file. set seltests { - {selective1 C 1.c {} {} {dropme1 dropme2} {}} - {selective2 C 2.c {} {} {foo} {mips*-*}} + {selective1 C 1.c {} {} {dropme1 dropme2} {alpha*-*}} + {selective2 C 2.c {} {} {foo} {alpha*-* mips*-*}} {selective3 C 2.c {-u foo} {foo} {{foo 0}} {mips*-*}} - {selective4 C++ 3.cc {} {start foo__1A foo__1B} {bar__1A} {mips*-*}} - {selective5 C++ 4.cc {} {} {foo__1B foo__1A} {mips*-*}} + {selective4 C++ 3.cc {} {start foo__1A foo__1B} {bar__1A} {alpha*-* mips*-*}} + {selective5 C++ 4.cc {} {} {foo__1B foo__1A} {alpha*-* mips*-*}} {selective6 C++ 5.cc {} {} {foo__1B foo__1A dropme1__Fv dropme2__Fv} {*-*-*}} } diff -uprN binutils-2.11.90.0.23/libiberty/ChangeLog binutils-2.11.90.0.24/libiberty/ChangeLog --- binutils-2.11.90.0.23/libiberty/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/libiberty/ChangeLog Sat Jul 14 10:29:15 2001 @@ -1,3 +1,8 @@ +2001-07-10 Jeff Johnston + + * Makefile.in: Add support for regex code. + * regex.c: New file. + 2001-07-05 Mark Klein * Makefile.in: Add ffs.c dependency. diff -uprN binutils-2.11.90.0.23/libiberty/Makefile.in binutils-2.11.90.0.24/libiberty/Makefile.in --- binutils-2.11.90.0.23/libiberty/Makefile.in Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/libiberty/Makefile.in Sat Jul 14 10:29:15 2001 @@ -128,7 +128,8 @@ CFILES = asprintf.c alloca.c argv.c atex floatformat.c hashtab.c hex.c index.c insque.c lbasename.c \ md5.c make-temp-file.c memchr.c \ memcmp.c memcpy.c memmove.c memset.c mkstemps.c objalloc.c obstack.c \ - partition.c pexecute.c putenv.c random.c rename.c rindex.c setenv.c \ + partition.c pexecute.c putenv.c random.c \ + regex.c rename.c rindex.c setenv.c \ sigsetmask.c safe-ctype.c sort.c spaces.c splay-tree.c strcasecmp.c \ strncasecmp.c strchr.c strdup.c strerror.c strncmp.c strrchr.c \ strsignal.c strstr.c strtod.c strtol.c strtoul.c tmpnam.c vasprintf.c \ @@ -140,7 +141,7 @@ REQUIRED_OFILES = argv.o alloca.o choose cp-demangle.o dyn-string.o fdmatch.o fnmatch.o getopt.o getopt1.o \ getpwd.o getruntime.o hashtab.o hex.o floatformat.o lbasename.o \ md5.o make-temp-file.o objalloc.o \ - obstack.o partition.o pexecute.o safe-ctype.o sort.o spaces.o \ + obstack.o partition.o pexecute.o regex.o safe-ctype.o sort.o spaces.o \ splay-tree.o strerror.o strsignal.o xatexit.o xexit.o xmalloc.o \ xmemdup.o xstrdup.o xstrerror.o ternary.o @@ -282,6 +283,7 @@ objalloc.o: config.h $(INCDIR)/objalloc. obstack.o: config.h $(INCDIR)/obstack.h partition.o: config.h $(INCDIR)/partition.h pexecute.o: config.h $(INCDIR)/libiberty.h +regex.o: $(INCDIR)/xregex.h $(INCDIR)/xregex2.h rename.o: config.h setenv.o: config.h sort.o: config.h $(INCDIR)/sort.h $(INCDIR)/ansidecl.h diff -uprN binutils-2.11.90.0.23/libiberty/regex.c binutils-2.11.90.0.24/libiberty/regex.c --- binutils-2.11.90.0.23/libiberty/regex.c Wed Dec 31 16:00:00 1969 +++ binutils-2.11.90.0.24/libiberty/regex.c Sat Jul 14 10:29:16 2001 @@ -0,0 +1,8245 @@ +/* Extended regular expression matching and search library, + version 0.12. + (Implements POSIX draft P1003.2/D11.2, except for some of the + internationalization features.) + Copyright (C) 1993-1999, 2000, 2001 Free Software Foundation, Inc. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +/* This file has been modified for usage in libiberty. It includes "xregex.h" + instead of . The "xregex.h" header file renames all external + routines with an "x" prefix so they do not collide with the native regex + routines or with other components regex routines. */ +#if defined _AIX && !defined REGEX_MALLOC + #pragma alloca +#endif + +#undef _GNU_SOURCE +#define _GNU_SOURCE + +#ifdef HAVE_CONFIG_H +# include +#endif + +#ifndef PARAMS +# if defined __GNUC__ || (defined __STDC__ && __STDC__) +# define PARAMS(args) args +# else +# define PARAMS(args) () +# endif /* GCC. */ +#endif /* Not PARAMS. */ + +#ifndef INSIDE_RECURSION + +# if defined STDC_HEADERS && !defined emacs +# include +# else +/* We need this for `regex.h', and perhaps for the Emacs include files. */ +# include +# endif + +# define WIDE_CHAR_SUPPORT (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC) + +/* For platform which support the ISO C amendement 1 functionality we + support user defined character classes. */ +# if defined _LIBC || WIDE_CHAR_SUPPORT +/* Solaris 2.5 has a bug: must be included before . */ +# include +# include +# endif + +# ifdef _LIBC +/* We have to keep the namespace clean. */ +# define regfree(preg) __regfree (preg) +# define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) +# define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) +# define regerror(errcode, preg, errbuf, errbuf_size) \ + __regerror(errcode, preg, errbuf, errbuf_size) +# define re_set_registers(bu, re, nu, st, en) \ + __re_set_registers (bu, re, nu, st, en) +# define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ + __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) +# define re_match(bufp, string, size, pos, regs) \ + __re_match (bufp, string, size, pos, regs) +# define re_search(bufp, string, size, startpos, range, regs) \ + __re_search (bufp, string, size, startpos, range, regs) +# define re_compile_pattern(pattern, length, bufp) \ + __re_compile_pattern (pattern, length, bufp) +# define re_set_syntax(syntax) __re_set_syntax (syntax) +# define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ + __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) +# define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) + +# define btowc __btowc + +/* We are also using some library internals. */ +# include +# include +# include +# include +# endif + +/* This is for other GNU distributions with internationalized messages. */ +# if HAVE_LIBINTL_H || defined _LIBC +# include +# ifdef _LIBC +# undef gettext +# define gettext(msgid) __dcgettext ("libc", msgid, LC_MESSAGES) +# endif +# else +# define gettext(msgid) (msgid) +# endif + +# ifndef gettext_noop +/* This define is so xgettext can find the internationalizable + strings. */ +# define gettext_noop(String) String +# endif + +/* The `emacs' switch turns on certain matching commands + that make sense only in Emacs. */ +# ifdef emacs + +# include "lisp.h" +# include "buffer.h" +# include "syntax.h" + +# else /* not emacs */ + +/* If we are not linking with Emacs proper, + we can't use the relocating allocator + even if config.h says that we can. */ +# undef REL_ALLOC + +# if defined STDC_HEADERS || defined _LIBC +# include +# else +char *malloc (); +char *realloc (); +# endif + +/* When used in Emacs's lib-src, we need to get bzero and bcopy somehow. + If nothing else has been done, use the method below. */ +# ifdef INHIBIT_STRING_HEADER +# if !(defined HAVE_BZERO && defined HAVE_BCOPY) +# if !defined bzero && !defined bcopy +# undef INHIBIT_STRING_HEADER +# endif +# endif +# endif + +/* This is the normal way of making sure we have a bcopy and a bzero. + This is used in most programs--a few other programs avoid this + by defining INHIBIT_STRING_HEADER. */ +# ifndef INHIBIT_STRING_HEADER +# if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC +# include +# ifndef bzero +# ifndef _LIBC +# define bzero(s, n) (memset (s, '\0', n), (s)) +# else +# define bzero(s, n) __bzero (s, n) +# endif +# endif +# else +# include +# ifndef memcmp +# define memcmp(s1, s2, n) bcmp (s1, s2, n) +# endif +# ifndef memcpy +# define memcpy(d, s, n) (bcopy (s, d, n), (d)) +# endif +# endif +# endif + +/* Define the syntax stuff for \<, \>, etc. */ + +/* This must be nonzero for the wordchar and notwordchar pattern + commands in re_match_2. */ +# ifndef Sword +# define Sword 1 +# endif + +# ifdef SWITCH_ENUM_BUG +# define SWITCH_ENUM_CAST(x) ((int)(x)) +# else +# define SWITCH_ENUM_CAST(x) (x) +# endif + +# endif /* not emacs */ + +# if defined _LIBC || HAVE_LIMITS_H +# include +# endif + +# ifndef MB_LEN_MAX +# define MB_LEN_MAX 1 +# endif + +/* Get the interface, including the syntax bits. */ +# include "xregex.h" /* change for libiberty */ + +/* isalpha etc. are used for the character classes. */ +# include + +/* Jim Meyering writes: + + "... Some ctype macros are valid only for character codes that + isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when + using /bin/cc or gcc but without giving an ansi option). So, all + ctype uses should be through macros like ISPRINT... If + STDC_HEADERS is defined, then autoconf has verified that the ctype + macros don't need to be guarded with references to isascii. ... + Defining isascii to 1 should let any compiler worth its salt + eliminate the && through constant folding." + Solaris defines some of these symbols so we must undefine them first. */ + +# undef ISASCII +# if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII) +# define ISASCII(c) 1 +# else +# define ISASCII(c) isascii(c) +# endif + +# ifdef isblank +# define ISBLANK(c) (ISASCII (c) && isblank (c)) +# else +# define ISBLANK(c) ((c) == ' ' || (c) == '\t') +# endif +# ifdef isgraph +# define ISGRAPH(c) (ISASCII (c) && isgraph (c)) +# else +# define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c)) +# endif + +# undef ISPRINT +# define ISPRINT(c) (ISASCII (c) && isprint (c)) +# define ISDIGIT(c) (ISASCII (c) && isdigit (c)) +# define ISALNUM(c) (ISASCII (c) && isalnum (c)) +# define ISALPHA(c) (ISASCII (c) && isalpha (c)) +# define ISCNTRL(c) (ISASCII (c) && iscntrl (c)) +# define ISLOWER(c) (ISASCII (c) && islower (c)) +# define ISPUNCT(c) (ISASCII (c) && ispunct (c)) +# define ISSPACE(c) (ISASCII (c) && isspace (c)) +# define ISUPPER(c) (ISASCII (c) && isupper (c)) +# define ISXDIGIT(c) (ISASCII (c) && isxdigit (c)) + +# ifdef _tolower +# define TOLOWER(c) _tolower(c) +# else +# define TOLOWER(c) tolower(c) +# endif + +# ifndef NULL +# define NULL (void *)0 +# endif + +/* We remove any previous definition of `SIGN_EXTEND_CHAR', + since ours (we hope) works properly with all combinations of + machines, compilers, `char' and `unsigned char' argument types. + (Per Bothner suggested the basic approach.) */ +# undef SIGN_EXTEND_CHAR +# if __STDC__ +# define SIGN_EXTEND_CHAR(c) ((signed char) (c)) +# else /* not __STDC__ */ +/* As in Harbison and Steele. */ +# define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128) +# endif + +# ifndef emacs +/* How many characters in the character set. */ +# define CHAR_SET_SIZE 256 + +# ifdef SYNTAX_TABLE + +extern char *re_syntax_table; + +# else /* not SYNTAX_TABLE */ + +static char re_syntax_table[CHAR_SET_SIZE]; + +static void init_syntax_once PARAMS ((void)); + +static void +init_syntax_once () +{ + register int c; + static int done = 0; + + if (done) + return; + bzero (re_syntax_table, sizeof re_syntax_table); + + for (c = 0; c < CHAR_SET_SIZE; ++c) + if (ISALNUM (c)) + re_syntax_table[c] = Sword; + + re_syntax_table['_'] = Sword; + + done = 1; +} + +# endif /* not SYNTAX_TABLE */ + +# define SYNTAX(c) re_syntax_table[(unsigned char) (c)] + +# endif /* emacs */ + +/* Integer type for pointers. */ +# if !defined _LIBC +typedef unsigned long int uintptr_t; +# endif + +/* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we + use `alloca' instead of `malloc'. This is because using malloc in + re_search* or re_match* could cause memory leaks when C-g is used in + Emacs; also, malloc is slower and causes storage fragmentation. On + the other hand, malloc is more portable, and easier to debug. + + Because we sometimes use alloca, some routines have to be macros, + not functions -- `alloca'-allocated space disappears at the end of the + function it is called in. */ + +# ifdef REGEX_MALLOC + +# define REGEX_ALLOCATE malloc +# define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) +# define REGEX_FREE free + +# else /* not REGEX_MALLOC */ + +/* Emacs already defines alloca, sometimes. */ +# ifndef alloca + +/* Make alloca work the best possible way. */ +# ifdef __GNUC__ +# define alloca __builtin_alloca +# else /* not __GNUC__ */ +# if HAVE_ALLOCA_H +# include +# endif /* HAVE_ALLOCA_H */ +# endif /* not __GNUC__ */ + +# endif /* not alloca */ + +# define REGEX_ALLOCATE alloca + +/* Assumes a `char *destination' variable. */ +# define REGEX_REALLOCATE(source, osize, nsize) \ + (destination = (char *) alloca (nsize), \ + memcpy (destination, source, osize)) + +/* No need to do anything to free, after alloca. */ +# define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ + +# endif /* not REGEX_MALLOC */ + +/* Define how to allocate the failure stack. */ + +# if defined REL_ALLOC && defined REGEX_MALLOC + +# define REGEX_ALLOCATE_STACK(size) \ + r_alloc (&failure_stack_ptr, (size)) +# define REGEX_REALLOCATE_STACK(source, osize, nsize) \ + r_re_alloc (&failure_stack_ptr, (nsize)) +# define REGEX_FREE_STACK(ptr) \ + r_alloc_free (&failure_stack_ptr) + +# else /* not using relocating allocator */ + +# ifdef REGEX_MALLOC + +# define REGEX_ALLOCATE_STACK malloc +# define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) +# define REGEX_FREE_STACK free + +# else /* not REGEX_MALLOC */ + +# define REGEX_ALLOCATE_STACK alloca + +# define REGEX_REALLOCATE_STACK(source, osize, nsize) \ + REGEX_REALLOCATE (source, osize, nsize) +/* No need to explicitly free anything. */ +# define REGEX_FREE_STACK(arg) + +# endif /* not REGEX_MALLOC */ +# endif /* not using relocating allocator */ + + +/* True if `size1' is non-NULL and PTR is pointing anywhere inside + `string1' or just past its end. This works if PTR is NULL, which is + a good thing. */ +# define FIRST_STRING_P(ptr) \ + (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) + +/* (Re)Allocate N items of type T using malloc, or fail. */ +# define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) +# define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) +# define RETALLOC_IF(addr, n, t) \ + if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) +# define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) + +# define BYTEWIDTH 8 /* In bits. */ + +# define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) + +# undef MAX +# undef MIN +# define MAX(a, b) ((a) > (b) ? (a) : (b)) +# define MIN(a, b) ((a) < (b) ? (a) : (b)) + +typedef char boolean; +# define false 0 +# define true 1 + +static reg_errcode_t byte_regex_compile _RE_ARGS ((const char *pattern, size_t size, + reg_syntax_t syntax, + struct re_pattern_buffer *bufp)); +static reg_errcode_t wcs_regex_compile _RE_ARGS ((const char *pattern, size_t size, + reg_syntax_t syntax, + struct re_pattern_buffer *bufp)); + +static int byte_re_match_2_internal PARAMS ((struct re_pattern_buffer *bufp, + const char *string1, int size1, + const char *string2, int size2, + int pos, + struct re_registers *regs, + int stop)); +static int wcs_re_match_2_internal PARAMS ((struct re_pattern_buffer *bufp, + const char *cstring1, int csize1, + const char *cstring2, int csize2, + int pos, + struct re_registers *regs, + int stop, + wchar_t *string1, int size1, + wchar_t *string2, int size2, + int *mbs_offset1, int *mbs_offset2)); +static int byte_re_search_2 PARAMS ((struct re_pattern_buffer *bufp, + const char *string1, int size1, + const char *string2, int size2, + int startpos, int range, + struct re_registers *regs, int stop)); +static int wcs_re_search_2 PARAMS ((struct re_pattern_buffer *bufp, + const char *string1, int size1, + const char *string2, int size2, + int startpos, int range, + struct re_registers *regs, int stop)); +static int byte_re_compile_fastmap PARAMS ((struct re_pattern_buffer *bufp)); +static int wcs_re_compile_fastmap PARAMS ((struct re_pattern_buffer *bufp)); + + +/* These are the command codes that appear in compiled regular + expressions. Some opcodes are followed by argument bytes. A + command code can specify any interpretation whatsoever for its + arguments. Zero bytes may appear in the compiled regular expression. */ + +typedef enum +{ + no_op = 0, + + /* Succeed right away--no more backtracking. */ + succeed, + + /* Followed by one byte giving n, then by n literal bytes. */ + exactn, + +# ifdef MBS_SUPPORT + /* Same as exactn, but contains binary data. */ + exactn_bin, +# endif + + /* Matches any (more or less) character. */ + anychar, + + /* Matches any one char belonging to specified set. First + following byte is number of bitmap bytes. Then come bytes + for a bitmap saying which chars are in. Bits in each byte + are ordered low-bit-first. A character is in the set if its + bit is 1. A character too large to have a bit in the map is + automatically not in the set. */ + /* ifdef MBS_SUPPORT, following element is length of character + classes, length of collating symbols, length of equivalence + classes, length of character ranges, and length of characters. + Next, character class element, collating symbols elements, + equivalence class elements, range elements, and character + elements follow. + See regex_compile function. */ + charset, + + /* Same parameters as charset, but match any character that is + not one of those specified. */ + charset_not, + + /* Start remembering the text that is matched, for storing in a + register. Followed by one byte with the register number, in + the range 0 to one less than the pattern buffer's re_nsub + field. Then followed by one byte with the number of groups + inner to this one. (This last has to be part of the + start_memory only because we need it in the on_failure_jump + of re_match_2.) */ + start_memory, + + /* Stop remembering the text that is matched and store it in a + memory register. Followed by one byte with the register + number, in the range 0 to one less than `re_nsub' in the + pattern buffer, and one byte with the number of inner groups, + just like `start_memory'. (We need the number of inner + groups here because we don't have any easy way of finding the + corresponding start_memory when we're at a stop_memory.) */ + stop_memory, + + /* Match a duplicate of something remembered. Followed by one + byte containing the register number. */ + duplicate, + + /* Fail unless at beginning of line. */ + begline, + + /* Fail unless at end of line. */ + endline, + + /* Succeeds if at beginning of buffer (if emacs) or at beginning + of string to be matched (if not). */ + begbuf, + + /* Analogously, for end of buffer/string. */ + endbuf, + + /* Followed by two byte relative address to which to jump. */ + jump, + + /* Same as jump, but marks the end of an alternative. */ + jump_past_alt, + + /* Followed by two-byte relative address of place to resume at + in case of failure. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + on_failure_jump, + + /* Like on_failure_jump, but pushes a placeholder instead of the + current string position when executed. */ + on_failure_keep_string_jump, + + /* Throw away latest failure point and then jump to following + two-byte relative address. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + pop_failure_jump, + + /* Change to pop_failure_jump if know won't have to backtrack to + match; otherwise change to jump. This is used to jump + back to the beginning of a repeat. If what follows this jump + clearly won't match what the repeat does, such that we can be + sure that there is no use backtracking out of repetitions + already matched, then we change it to a pop_failure_jump. + Followed by two-byte address. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + maybe_pop_jump, + + /* Jump to following two-byte address, and push a dummy failure + point. This failure point will be thrown away if an attempt + is made to use it for a failure. A `+' construct makes this + before the first repeat. Also used as an intermediary kind + of jump when compiling an alternative. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + dummy_failure_jump, + + /* Push a dummy failure point and continue. Used at the end of + alternatives. */ + push_dummy_failure, + + /* Followed by two-byte relative address and two-byte number n. + After matching N times, jump to the address upon failure. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + succeed_n, + + /* Followed by two-byte relative address, and two-byte number n. + Jump to the address N times, then fail. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + jump_n, + + /* Set the following two-byte relative address to the + subsequent two-byte number. The address *includes* the two + bytes of number. */ + /* ifdef MBS_SUPPORT, the size of address is 1. */ + set_number_at, + + wordchar, /* Matches any word-constituent character. */ + notwordchar, /* Matches any char that is not a word-constituent. */ + + wordbeg, /* Succeeds if at word beginning. */ + wordend, /* Succeeds if at word end. */ + + wordbound, /* Succeeds if at a word boundary. */ + notwordbound /* Succeeds if not at a word boundary. */ + +# ifdef emacs + ,before_dot, /* Succeeds if before point. */ + at_dot, /* Succeeds if at point. */ + after_dot, /* Succeeds if after point. */ + + /* Matches any character whose syntax is specified. Followed by + a byte which contains a syntax code, e.g., Sword. */ + syntaxspec, + + /* Matches any character whose syntax is not that specified. */ + notsyntaxspec +# endif /* emacs */ +} re_opcode_t; +#endif /* not INSIDE_RECURSION */ + + +#ifdef BYTE +# define CHAR_T char +# define UCHAR_T unsigned char +# define COMPILED_BUFFER_VAR bufp->buffer +# define OFFSET_ADDRESS_SIZE 2 +# define PREFIX(name) byte_##name +# define ARG_PREFIX(name) name +# define PUT_CHAR(c) putchar (c) +#elif defined WCHAR +# define CHAR_T wchar_t +# define UCHAR_T wchar_t +# define COMPILED_BUFFER_VAR wc_buffer +# define OFFSET_ADDRESS_SIZE 1 /* the size which STORE_NUMBER macro use */ +# define CHAR_CLASS_SIZE ((__alignof__(wctype_t)+sizeof(wctype_t))/sizeof(CHAR_T)+1) +# define PREFIX(name) wcs_##name +# define ARG_PREFIX(name) c##name +/* Should we use wide stream?? */ +# define PUT_CHAR(c) printf ("%C", c); +# define TRUE 1 +# define FALSE 0 +#else +# ifdef MBS_SUPPORT +# define WCHAR +# define INSIDE_RECURSION +# include "regex.c" +# undef INSIDE_RECURSION +# endif +# define BYTE +# define INSIDE_RECURSION +# include "regex.c" +# undef INSIDE_RECURSION +#endif + +#ifdef INSIDE_RECURSION +/* Common operations on the compiled pattern. */ + +/* Store NUMBER in two contiguous bytes starting at DESTINATION. */ +/* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ + +# ifdef WCHAR +# define STORE_NUMBER(destination, number) \ + do { \ + *(destination) = (UCHAR_T)(number); \ + } while (0) +# else /* BYTE */ +# define STORE_NUMBER(destination, number) \ + do { \ + (destination)[0] = (number) & 0377; \ + (destination)[1] = (number) >> 8; \ + } while (0) +# endif /* WCHAR */ + +/* Same as STORE_NUMBER, except increment DESTINATION to + the byte after where the number is stored. Therefore, DESTINATION + must be an lvalue. */ +/* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ + +# define STORE_NUMBER_AND_INCR(destination, number) \ + do { \ + STORE_NUMBER (destination, number); \ + (destination) += OFFSET_ADDRESS_SIZE; \ + } while (0) + +/* Put into DESTINATION a number stored in two contiguous bytes starting + at SOURCE. */ +/* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ + +# ifdef WCHAR +# define EXTRACT_NUMBER(destination, source) \ + do { \ + (destination) = *(source); \ + } while (0) +# else /* BYTE */ +# define EXTRACT_NUMBER(destination, source) \ + do { \ + (destination) = *(source) & 0377; \ + (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ + } while (0) +# endif + +# ifdef DEBUG +static void PREFIX(extract_number) _RE_ARGS ((int *dest, UCHAR_T *source)); +static void +PREFIX(extract_number) (dest, source) + int *dest; + UCHAR_T *source; +{ +# ifdef WCHAR + *dest = *source; +# else /* BYTE */ + int temp = SIGN_EXTEND_CHAR (*(source + 1)); + *dest = *source & 0377; + *dest += temp << 8; +# endif +} + +# ifndef EXTRACT_MACROS /* To debug the macros. */ +# undef EXTRACT_NUMBER +# define EXTRACT_NUMBER(dest, src) PREFIX(extract_number) (&dest, src) +# endif /* not EXTRACT_MACROS */ + +# endif /* DEBUG */ + +/* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. + SOURCE must be an lvalue. */ + +# define EXTRACT_NUMBER_AND_INCR(destination, source) \ + do { \ + EXTRACT_NUMBER (destination, source); \ + (source) += OFFSET_ADDRESS_SIZE; \ + } while (0) + +# ifdef DEBUG +static void PREFIX(extract_number_and_incr) _RE_ARGS ((int *destination, + UCHAR_T **source)); +static void +PREFIX(extract_number_and_incr) (destination, source) + int *destination; + UCHAR_T **source; +{ + PREFIX(extract_number) (destination, *source); + *source += OFFSET_ADDRESS_SIZE; +} + +# ifndef EXTRACT_MACROS +# undef EXTRACT_NUMBER_AND_INCR +# define EXTRACT_NUMBER_AND_INCR(dest, src) \ + PREFIX(extract_number_and_incr) (&dest, &src) +# endif /* not EXTRACT_MACROS */ + +# endif /* DEBUG */ + + + +/* If DEBUG is defined, Regex prints many voluminous messages about what + it is doing (if the variable `debug' is nonzero). If linked with the + main program in `iregex.c', you can enter patterns and strings + interactively. And if linked with the main program in `main.c' and + the other test files, you can run the already-written tests. */ + +# ifdef DEBUG + +# ifndef DEFINED_ONCE + +/* We use standard I/O for debugging. */ +# include + +/* It is useful to test things that ``must'' be true when debugging. */ +# include + +static int debug; + +# define DEBUG_STATEMENT(e) e +# define DEBUG_PRINT1(x) if (debug) printf (x) +# define DEBUG_PRINT2(x1, x2) if (debug) printf (x1, x2) +# define DEBUG_PRINT3(x1, x2, x3) if (debug) printf (x1, x2, x3) +# define DEBUG_PRINT4(x1, x2, x3, x4) if (debug) printf (x1, x2, x3, x4) +# endif /* not DEFINED_ONCE */ + +# define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ + if (debug) PREFIX(print_partial_compiled_pattern) (s, e) +# define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ + if (debug) PREFIX(print_double_string) (w, s1, sz1, s2, sz2) + + +/* Print the fastmap in human-readable form. */ + +# ifndef DEFINED_ONCE +void +print_fastmap (fastmap) + char *fastmap; +{ + unsigned was_a_range = 0; + unsigned i = 0; + + while (i < (1 << BYTEWIDTH)) + { + if (fastmap[i++]) + { + was_a_range = 0; + putchar (i - 1); + while (i < (1 << BYTEWIDTH) && fastmap[i]) + { + was_a_range = 1; + i++; + } + if (was_a_range) + { + printf ("-"); + putchar (i - 1); + } + } + } + putchar ('\n'); +} +# endif /* not DEFINED_ONCE */ + + +/* Print a compiled pattern string in human-readable form, starting at + the START pointer into it and ending just before the pointer END. */ + +void +PREFIX(print_partial_compiled_pattern) (start, end) + UCHAR_T *start; + UCHAR_T *end; +{ + int mcnt, mcnt2; + UCHAR_T *p1; + UCHAR_T *p = start; + UCHAR_T *pend = end; + + if (start == NULL) + { + printf ("(null)\n"); + return; + } + + /* Loop over pattern commands. */ + while (p < pend) + { +# ifdef _LIBC + printf ("%td:\t", p - start); +# else + printf ("%ld:\t", (long int) (p - start)); +# endif + + switch ((re_opcode_t) *p++) + { + case no_op: + printf ("/no_op"); + break; + + case exactn: + mcnt = *p++; + printf ("/exactn/%d", mcnt); + do + { + putchar ('/'); + PUT_CHAR (*p++); + } + while (--mcnt); + break; + +# ifdef MBS_SUPPORT + case exactn_bin: + mcnt = *p++; + printf ("/exactn_bin/%d", mcnt); + do + { + printf("/%lx", (long int) *p++); + } + while (--mcnt); + break; +# endif /* MBS_SUPPORT */ + + case start_memory: + mcnt = *p++; + printf ("/start_memory/%d/%ld", mcnt, (long int) *p++); + break; + + case stop_memory: + mcnt = *p++; + printf ("/stop_memory/%d/%ld", mcnt, (long int) *p++); + break; + + case duplicate: + printf ("/duplicate/%ld", (long int) *p++); + break; + + case anychar: + printf ("/anychar"); + break; + + case charset: + case charset_not: + { +# ifdef WCHAR + int i, length; + wchar_t *workp = p; + printf ("/charset [%s", + (re_opcode_t) *(workp - 1) == charset_not ? "^" : ""); + p += 5; + length = *workp++; /* the length of char_classes */ + for (i=0 ; ibuffer; + + PREFIX(print_partial_compiled_pattern) (buffer, buffer + + bufp->used / sizeof(UCHAR_T)); + printf ("%ld bytes used/%ld bytes allocated.\n", + bufp->used, bufp->allocated); + + if (bufp->fastmap_accurate && bufp->fastmap) + { + printf ("fastmap: "); + print_fastmap (bufp->fastmap); + } + +# ifdef _LIBC + printf ("re_nsub: %Zd\t", bufp->re_nsub); +# else + printf ("re_nsub: %ld\t", (long int) bufp->re_nsub); +# endif + printf ("regs_alloc: %d\t", bufp->regs_allocated); + printf ("can_be_null: %d\t", bufp->can_be_null); + printf ("newline_anchor: %d\n", bufp->newline_anchor); + printf ("no_sub: %d\t", bufp->no_sub); + printf ("not_bol: %d\t", bufp->not_bol); + printf ("not_eol: %d\t", bufp->not_eol); + printf ("syntax: %lx\n", bufp->syntax); + /* Perhaps we should print the translate table? */ +} + + +void +PREFIX(print_double_string) (where, string1, size1, string2, size2) + const CHAR_T *where; + const CHAR_T *string1; + const CHAR_T *string2; + int size1; + int size2; +{ + int this_char; + + if (where == NULL) + printf ("(null)"); + else + { + if (FIRST_STRING_P (where)) + { + for (this_char = where - string1; this_char < size1; this_char++) + PUT_CHAR (string1[this_char]); + + where = string2; + } + + for (this_char = where - string2; this_char < size2; this_char++) + PUT_CHAR (string2[this_char]); + } +} + +# ifndef DEFINED_ONCE +void +printchar (c) + int c; +{ + putc (c, stderr); +} +# endif + +# else /* not DEBUG */ + +# ifndef DEFINED_ONCE +# undef assert +# define assert(e) + +# define DEBUG_STATEMENT(e) +# define DEBUG_PRINT1(x) +# define DEBUG_PRINT2(x1, x2) +# define DEBUG_PRINT3(x1, x2, x3) +# define DEBUG_PRINT4(x1, x2, x3, x4) +# endif /* not DEFINED_ONCE */ +# define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) +# define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) + +# endif /* not DEBUG */ + + + +# ifdef WCHAR +/* This convert a multibyte string to a wide character string. + And write their correspondances to offset_buffer(see below) + and write whether each wchar_t is binary data to is_binary. + This assume invalid multibyte sequences as binary data. + We assume offset_buffer and is_binary is already allocated + enough space. */ + +static size_t convert_mbs_to_wcs (CHAR_T *dest, const unsigned char* src, + size_t len, int *offset_buffer, + char *is_binary); +static size_t +convert_mbs_to_wcs (dest, src, len, offset_buffer, is_binary) + CHAR_T *dest; + const unsigned char* src; + size_t len; /* the length of multibyte string. */ + + /* It hold correspondances between src(char string) and + dest(wchar_t string) for optimization. + e.g. src = "xxxyzz" + dest = {'X', 'Y', 'Z'} + (each "xxx", "y" and "zz" represent one multibyte character + corresponding to 'X', 'Y' and 'Z'.) + offset_buffer = {0, 0+3("xxx"), 0+3+1("y"), 0+3+1+2("zz")} + = {0, 3, 4, 6} + */ + int *offset_buffer; + char *is_binary; +{ + wchar_t *pdest = dest; + const unsigned char *psrc = src; + size_t wc_count = 0; + + mbstate_t mbs; + int i, consumed; + size_t mb_remain = len; + size_t mb_count = 0; + + /* Initialize the conversion state. */ + memset (&mbs, 0, sizeof (mbstate_t)); + + offset_buffer[0] = 0; + for( ; mb_remain > 0 ; ++wc_count, ++pdest, mb_remain -= consumed, + psrc += consumed) + { + consumed = mbrtowc (pdest, psrc, mb_remain, &mbs); + + if (consumed <= 0) + /* failed to convert. maybe src contains binary data. + So we consume 1 byte manualy. */ + { + *pdest = *psrc; + consumed = 1; + is_binary[wc_count] = TRUE; + } + else + is_binary[wc_count] = FALSE; + /* In sjis encoding, we use yen sign as escape character in + place of reverse solidus. So we convert 0x5c(yen sign in + sjis) to not 0xa5(yen sign in UCS2) but 0x5c(reverse + solidus in UCS2). */ + if (consumed == 1 && (int) *psrc == 0x5c && (int) *pdest == 0xa5) + *pdest = (wchar_t) *psrc; + + offset_buffer[wc_count + 1] = mb_count += consumed; + } + + /* Fill remain of the buffer with sentinel. */ + for (i = wc_count + 1 ; i <= len ; i++) + offset_buffer[i] = mb_count + 1; + + return wc_count; +} + +# endif /* WCHAR */ + +#else /* not INSIDE_RECURSION */ + +/* Set by `re_set_syntax' to the current regexp syntax to recognize. Can + also be assigned to arbitrarily: each pattern buffer stores its own + syntax, so it can be changed between regex compilations. */ +/* This has no initializer because initialized variables in Emacs + become read-only after dumping. */ +reg_syntax_t re_syntax_options; + + +/* Specify the precise syntax of regexps for compilation. This provides + for compatibility for various utilities which historically have + different, incompatible syntaxes. + + The argument SYNTAX is a bit mask comprised of the various bits + defined in regex.h. We return the old syntax. */ + +reg_syntax_t +re_set_syntax (syntax) + reg_syntax_t syntax; +{ + reg_syntax_t ret = re_syntax_options; + + re_syntax_options = syntax; +# ifdef DEBUG + if (syntax & RE_DEBUG) + debug = 1; + else if (debug) /* was on but now is not */ + debug = 0; +# endif /* DEBUG */ + return ret; +} +# ifdef _LIBC +weak_alias (__re_set_syntax, re_set_syntax) +# endif + +/* This table gives an error message for each of the error codes listed + in regex.h. Obviously the order here has to be same as there. + POSIX doesn't require that we do anything for REG_NOERROR, + but why not be nice? */ + +static const char re_error_msgid[] = + { +# define REG_NOERROR_IDX 0 + gettext_noop ("Success") /* REG_NOERROR */ + "\0" +# define REG_NOMATCH_IDX (REG_NOERROR_IDX + sizeof "Success") + gettext_noop ("No match") /* REG_NOMATCH */ + "\0" +# define REG_BADPAT_IDX (REG_NOMATCH_IDX + sizeof "No match") + gettext_noop ("Invalid regular expression") /* REG_BADPAT */ + "\0" +# define REG_ECOLLATE_IDX (REG_BADPAT_IDX + sizeof "Invalid regular expression") + gettext_noop ("Invalid collation character") /* REG_ECOLLATE */ + "\0" +# define REG_ECTYPE_IDX (REG_ECOLLATE_IDX + sizeof "Invalid collation character") + gettext_noop ("Invalid character class name") /* REG_ECTYPE */ + "\0" +# define REG_EESCAPE_IDX (REG_ECTYPE_IDX + sizeof "Invalid character class name") + gettext_noop ("Trailing backslash") /* REG_EESCAPE */ + "\0" +# define REG_ESUBREG_IDX (REG_EESCAPE_IDX + sizeof "Trailing backslash") + gettext_noop ("Invalid back reference") /* REG_ESUBREG */ + "\0" +# define REG_EBRACK_IDX (REG_ESUBREG_IDX + sizeof "Invalid back reference") + gettext_noop ("Unmatched [ or [^") /* REG_EBRACK */ + "\0" +# define REG_EPAREN_IDX (REG_EBRACK_IDX + sizeof "Unmatched [ or [^") + gettext_noop ("Unmatched ( or \\(") /* REG_EPAREN */ + "\0" +# define REG_EBRACE_IDX (REG_EPAREN_IDX + sizeof "Unmatched ( or \\(") + gettext_noop ("Unmatched \\{") /* REG_EBRACE */ + "\0" +# define REG_BADBR_IDX (REG_EBRACE_IDX + sizeof "Unmatched \\{") + gettext_noop ("Invalid content of \\{\\}") /* REG_BADBR */ + "\0" +# define REG_ERANGE_IDX (REG_BADBR_IDX + sizeof "Invalid content of \\{\\}") + gettext_noop ("Invalid range end") /* REG_ERANGE */ + "\0" +# define REG_ESPACE_IDX (REG_ERANGE_IDX + sizeof "Invalid range end") + gettext_noop ("Memory exhausted") /* REG_ESPACE */ + "\0" +# define REG_BADRPT_IDX (REG_ESPACE_IDX + sizeof "Memory exhausted") + gettext_noop ("Invalid preceding regular expression") /* REG_BADRPT */ + "\0" +# define REG_EEND_IDX (REG_BADRPT_IDX + sizeof "Invalid preceding regular expression") + gettext_noop ("Premature end of regular expression") /* REG_EEND */ + "\0" +# define REG_ESIZE_IDX (REG_EEND_IDX + sizeof "Premature end of regular expression") + gettext_noop ("Regular expression too big") /* REG_ESIZE */ + "\0" +# define REG_ERPAREN_IDX (REG_ESIZE_IDX + sizeof "Regular expression too big") + gettext_noop ("Unmatched ) or \\)") /* REG_ERPAREN */ + }; + +static const size_t re_error_msgid_idx[] = + { + REG_NOERROR_IDX, + REG_NOMATCH_IDX, + REG_BADPAT_IDX, + REG_ECOLLATE_IDX, + REG_ECTYPE_IDX, + REG_EESCAPE_IDX, + REG_ESUBREG_IDX, + REG_EBRACK_IDX, + REG_EPAREN_IDX, + REG_EBRACE_IDX, + REG_BADBR_IDX, + REG_ERANGE_IDX, + REG_ESPACE_IDX, + REG_BADRPT_IDX, + REG_EEND_IDX, + REG_ESIZE_IDX, + REG_ERPAREN_IDX + }; + +#endif /* INSIDE_RECURSION */ + +#ifndef DEFINED_ONCE +/* Avoiding alloca during matching, to placate r_alloc. */ + +/* Define MATCH_MAY_ALLOCATE unless we need to make sure that the + searching and matching functions should not call alloca. On some + systems, alloca is implemented in terms of malloc, and if we're + using the relocating allocator routines, then malloc could cause a + relocation, which might (if the strings being searched are in the + ralloc heap) shift the data out from underneath the regexp + routines. + + Here's another reason to avoid allocation: Emacs + processes input from X in a signal handler; processing X input may + call malloc; if input arrives while a matching routine is calling + malloc, then we're scrod. But Emacs can't just block input while + calling matching routines; then we don't notice interrupts when + they come in. So, Emacs blocks input around all regexp calls + except the matching calls, which it leaves unprotected, in the + faith that they will not malloc. */ + +/* Normally, this is fine. */ +# define MATCH_MAY_ALLOCATE + +/* When using GNU C, we are not REALLY using the C alloca, no matter + what config.h may say. So don't take precautions for it. */ +# ifdef __GNUC__ +# undef C_ALLOCA +# endif + +/* The match routines may not allocate if (1) they would do it with malloc + and (2) it's not safe for them to use malloc. + Note that if REL_ALLOC is defined, matching would not use malloc for the + failure stack, but we would still use it for the register vectors; + so REL_ALLOC should not affect this. */ +# if (defined C_ALLOCA || defined REGEX_MALLOC) && defined emacs +# undef MATCH_MAY_ALLOCATE +# endif +#endif /* not DEFINED_ONCE */ + +#ifdef INSIDE_RECURSION +/* Failure stack declarations and macros; both re_compile_fastmap and + re_match_2 use a failure stack. These have to be macros because of + REGEX_ALLOCATE_STACK. */ + + +/* Number of failure points for which to initially allocate space + when matching. If this number is exceeded, we allocate more + space, so it is not a hard limit. */ +# ifndef INIT_FAILURE_ALLOC +# define INIT_FAILURE_ALLOC 5 +# endif + +/* Roughly the maximum number of failure points on the stack. Would be + exactly that if always used MAX_FAILURE_ITEMS items each time we failed. + This is a variable only so users of regex can assign to it; we never + change it ourselves. */ + +# ifdef INT_IS_16BIT + +# ifndef DEFINED_ONCE +# if defined MATCH_MAY_ALLOCATE +/* 4400 was enough to cause a crash on Alpha OSF/1, + whose default stack limit is 2mb. */ +long int re_max_failures = 4000; +# else +long int re_max_failures = 2000; +# endif +# endif + +union PREFIX(fail_stack_elt) +{ + UCHAR_T *pointer; + long int integer; +}; + +typedef union PREFIX(fail_stack_elt) PREFIX(fail_stack_elt_t); + +typedef struct +{ + PREFIX(fail_stack_elt_t) *stack; + unsigned long int size; + unsigned long int avail; /* Offset of next open position. */ +} PREFIX(fail_stack_type); + +# else /* not INT_IS_16BIT */ + +# ifndef DEFINED_ONCE +# if defined MATCH_MAY_ALLOCATE +/* 4400 was enough to cause a crash on Alpha OSF/1, + whose default stack limit is 2mb. */ +int re_max_failures = 4000; +# else +int re_max_failures = 2000; +# endif +# endif + +union PREFIX(fail_stack_elt) +{ + UCHAR_T *pointer; + int integer; +}; + +typedef union PREFIX(fail_stack_elt) PREFIX(fail_stack_elt_t); + +typedef struct +{ + PREFIX(fail_stack_elt_t) *stack; + unsigned size; + unsigned avail; /* Offset of next open position. */ +} PREFIX(fail_stack_type); + +# endif /* INT_IS_16BIT */ + +# ifndef DEFINED_ONCE +# define FAIL_STACK_EMPTY() (fail_stack.avail == 0) +# define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0) +# define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size) +# endif + + +/* Define macros to initialize and free the failure stack. + Do `return -2' if the alloc fails. */ + +# ifdef MATCH_MAY_ALLOCATE +# define INIT_FAIL_STACK() \ + do { \ + fail_stack.stack = (PREFIX(fail_stack_elt_t) *) \ + REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * sizeof (PREFIX(fail_stack_elt_t))); \ + \ + if (fail_stack.stack == NULL) \ + return -2; \ + \ + fail_stack.size = INIT_FAILURE_ALLOC; \ + fail_stack.avail = 0; \ + } while (0) + +# define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack) +# else +# define INIT_FAIL_STACK() \ + do { \ + fail_stack.avail = 0; \ + } while (0) + +# define RESET_FAIL_STACK() +# endif + + +/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items. + + Return 1 if succeeds, and 0 if either ran out of memory + allocating space for it or it was already too large. + + REGEX_REALLOCATE_STACK requires `destination' be declared. */ + +# define DOUBLE_FAIL_STACK(fail_stack) \ + ((fail_stack).size > (unsigned) (re_max_failures * MAX_FAILURE_ITEMS) \ + ? 0 \ + : ((fail_stack).stack = (PREFIX(fail_stack_elt_t) *) \ + REGEX_REALLOCATE_STACK ((fail_stack).stack, \ + (fail_stack).size * sizeof (PREFIX(fail_stack_elt_t)), \ + ((fail_stack).size << 1) * sizeof (PREFIX(fail_stack_elt_t))),\ + \ + (fail_stack).stack == NULL \ + ? 0 \ + : ((fail_stack).size <<= 1, \ + 1))) + + +/* Push pointer POINTER on FAIL_STACK. + Return 1 if was able to do so and 0 if ran out of memory allocating + space to do so. */ +# define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \ + ((FAIL_STACK_FULL () \ + && !DOUBLE_FAIL_STACK (FAIL_STACK)) \ + ? 0 \ + : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \ + 1)) + +/* Push a pointer value onto the failure stack. + Assumes the variable `fail_stack'. Probably should only + be called from within `PUSH_FAILURE_POINT'. */ +# define PUSH_FAILURE_POINTER(item) \ + fail_stack.stack[fail_stack.avail++].pointer = (UCHAR_T *) (item) + +/* This pushes an integer-valued item onto the failure stack. + Assumes the variable `fail_stack'. Probably should only + be called from within `PUSH_FAILURE_POINT'. */ +# define PUSH_FAILURE_INT(item) \ + fail_stack.stack[fail_stack.avail++].integer = (item) + +/* Push a fail_stack_elt_t value onto the failure stack. + Assumes the variable `fail_stack'. Probably should only + be called from within `PUSH_FAILURE_POINT'. */ +# define PUSH_FAILURE_ELT(item) \ + fail_stack.stack[fail_stack.avail++] = (item) + +/* These three POP... operations complement the three PUSH... operations. + All assume that `fail_stack' is nonempty. */ +# define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer +# define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer +# define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail] + +/* Used to omit pushing failure point id's when we're not debugging. */ +# ifdef DEBUG +# define DEBUG_PUSH PUSH_FAILURE_INT +# define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_INT () +# else +# define DEBUG_PUSH(item) +# define DEBUG_POP(item_addr) +# endif + + +/* Push the information about the state we will need + if we ever fail back to it. + + Requires variables fail_stack, regstart, regend, reg_info, and + num_regs_pushed be declared. DOUBLE_FAIL_STACK requires `destination' + be declared. + + Does `return FAILURE_CODE' if runs out of memory. */ + +# define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \ + do { \ + char *destination; \ + /* Must be int, so when we don't save any registers, the arithmetic \ + of 0 + -1 isn't done as unsigned. */ \ + /* Can't be int, since there is not a shred of a guarantee that int \ + is wide enough to hold a value of something to which pointer can \ + be assigned */ \ + active_reg_t this_reg; \ + \ + DEBUG_STATEMENT (failure_id++); \ + DEBUG_STATEMENT (nfailure_points_pushed++); \ + DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \ + DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\ + DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\ + \ + DEBUG_PRINT2 (" slots needed: %ld\n", NUM_FAILURE_ITEMS); \ + DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \ + \ + /* Ensure we have enough space allocated for what we will push. */ \ + while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \ + { \ + if (!DOUBLE_FAIL_STACK (fail_stack)) \ + return failure_code; \ + \ + DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \ + (fail_stack).size); \ + DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\ + } \ + \ + /* Push the info, starting with the registers. */ \ + DEBUG_PRINT1 ("\n"); \ + \ + if (1) \ + for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \ + this_reg++) \ + { \ + DEBUG_PRINT2 (" Pushing reg: %lu\n", this_reg); \ + DEBUG_STATEMENT (num_regs_pushed++); \ + \ + DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \ + PUSH_FAILURE_POINTER (regstart[this_reg]); \ + \ + DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \ + PUSH_FAILURE_POINTER (regend[this_reg]); \ + \ + DEBUG_PRINT2 (" info: %p\n ", \ + reg_info[this_reg].word.pointer); \ + DEBUG_PRINT2 (" match_null=%d", \ + REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \ + DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \ + DEBUG_PRINT2 (" matched_something=%d", \ + MATCHED_SOMETHING (reg_info[this_reg])); \ + DEBUG_PRINT2 (" ever_matched=%d", \ + EVER_MATCHED_SOMETHING (reg_info[this_reg])); \ + DEBUG_PRINT1 ("\n"); \ + PUSH_FAILURE_ELT (reg_info[this_reg].word); \ + } \ + \ + DEBUG_PRINT2 (" Pushing low active reg: %ld\n", lowest_active_reg);\ + PUSH_FAILURE_INT (lowest_active_reg); \ + \ + DEBUG_PRINT2 (" Pushing high active reg: %ld\n", highest_active_reg);\ + PUSH_FAILURE_INT (highest_active_reg); \ + \ + DEBUG_PRINT2 (" Pushing pattern %p:\n", pattern_place); \ + DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \ + PUSH_FAILURE_POINTER (pattern_place); \ + \ + DEBUG_PRINT2 (" Pushing string %p: `", string_place); \ + DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \ + size2); \ + DEBUG_PRINT1 ("'\n"); \ + PUSH_FAILURE_POINTER (string_place); \ + \ + DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \ + DEBUG_PUSH (failure_id); \ + } while (0) + +# ifndef DEFINED_ONCE +/* This is the number of items that are pushed and popped on the stack + for each register. */ +# define NUM_REG_ITEMS 3 + +/* Individual items aside from the registers. */ +# ifdef DEBUG +# define NUM_NONREG_ITEMS 5 /* Includes failure point id. */ +# else +# define NUM_NONREG_ITEMS 4 +# endif + +/* We push at most this many items on the stack. */ +/* We used to use (num_regs - 1), which is the number of registers + this regexp will save; but that was changed to 5 + to avoid stack overflow for a regexp with lots of parens. */ +# define MAX_FAILURE_ITEMS (5 * NUM_REG_ITEMS + NUM_NONREG_ITEMS) + +/* We actually push this many items. */ +# define NUM_FAILURE_ITEMS \ + (((0 \ + ? 0 : highest_active_reg - lowest_active_reg + 1) \ + * NUM_REG_ITEMS) \ + + NUM_NONREG_ITEMS) + +/* How many items can still be added to the stack without overflowing it. */ +# define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) +# endif /* not DEFINED_ONCE */ + + +/* Pops what PUSH_FAIL_STACK pushes. + + We restore into the parameters, all of which should be lvalues: + STR -- the saved data position. + PAT -- the saved pattern position. + LOW_REG, HIGH_REG -- the highest and lowest active registers. + REGSTART, REGEND -- arrays of string positions. + REG_INFO -- array of information about each subexpression. + + Also assumes the variables `fail_stack' and (if debugging), `bufp', + `pend', `string1', `size1', `string2', and `size2'. */ +# define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\ +{ \ + DEBUG_STATEMENT (unsigned failure_id;) \ + active_reg_t this_reg; \ + const UCHAR_T *string_temp; \ + \ + assert (!FAIL_STACK_EMPTY ()); \ + \ + /* Remove failure points and point to how many regs pushed. */ \ + DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \ + DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \ + DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \ + \ + assert (fail_stack.avail >= NUM_NONREG_ITEMS); \ + \ + DEBUG_POP (&failure_id); \ + DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \ + \ + /* If the saved string location is NULL, it came from an \ + on_failure_keep_string_jump opcode, and we want to throw away the \ + saved NULL, thus retaining our current position in the string. */ \ + string_temp = POP_FAILURE_POINTER (); \ + if (string_temp != NULL) \ + str = (const CHAR_T *) string_temp; \ + \ + DEBUG_PRINT2 (" Popping string %p: `", str); \ + DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ + DEBUG_PRINT1 ("'\n"); \ + \ + pat = (UCHAR_T *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" Popping pattern %p:\n", pat); \ + DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ + \ + /* Restore register info. */ \ + high_reg = (active_reg_t) POP_FAILURE_INT (); \ + DEBUG_PRINT2 (" Popping high active reg: %ld\n", high_reg); \ + \ + low_reg = (active_reg_t) POP_FAILURE_INT (); \ + DEBUG_PRINT2 (" Popping low active reg: %ld\n", low_reg); \ + \ + if (1) \ + for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \ + { \ + DEBUG_PRINT2 (" Popping reg: %ld\n", this_reg); \ + \ + reg_info[this_reg].word = POP_FAILURE_ELT (); \ + DEBUG_PRINT2 (" info: %p\n", \ + reg_info[this_reg].word.pointer); \ + \ + regend[this_reg] = (const CHAR_T *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \ + \ + regstart[this_reg] = (const CHAR_T *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \ + } \ + else \ + { \ + for (this_reg = highest_active_reg; this_reg > high_reg; this_reg--) \ + { \ + reg_info[this_reg].word.integer = 0; \ + regend[this_reg] = 0; \ + regstart[this_reg] = 0; \ + } \ + highest_active_reg = high_reg; \ + } \ + \ + set_regs_matched_done = 0; \ + DEBUG_STATEMENT (nfailure_points_popped++); \ +} /* POP_FAILURE_POINT */ + +/* Structure for per-register (a.k.a. per-group) information. + Other register information, such as the + starting and ending positions (which are addresses), and the list of + inner groups (which is a bits list) are maintained in separate + variables. + + We are making a (strictly speaking) nonportable assumption here: that + the compiler will pack our bit fields into something that fits into + the type of `word', i.e., is something that fits into one item on the + failure stack. */ + + +/* Declarations and macros for re_match_2. */ + +typedef union +{ + PREFIX(fail_stack_elt_t) word; + struct + { + /* This field is one if this group can match the empty string, + zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */ +# define MATCH_NULL_UNSET_VALUE 3 + unsigned match_null_string_p : 2; + unsigned is_active : 1; + unsigned matched_something : 1; + unsigned ever_matched_something : 1; + } bits; +} PREFIX(register_info_type); + +# ifndef DEFINED_ONCE +# define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p) +# define IS_ACTIVE(R) ((R).bits.is_active) +# define MATCHED_SOMETHING(R) ((R).bits.matched_something) +# define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something) + + +/* Call this when have matched a real character; it sets `matched' flags + for the subexpressions which we are currently inside. Also records + that those subexprs have matched. */ +# define SET_REGS_MATCHED() \ + do \ + { \ + if (!set_regs_matched_done) \ + { \ + active_reg_t r; \ + set_regs_matched_done = 1; \ + for (r = lowest_active_reg; r <= highest_active_reg; r++) \ + { \ + MATCHED_SOMETHING (reg_info[r]) \ + = EVER_MATCHED_SOMETHING (reg_info[r]) \ + = 1; \ + } \ + } \ + } \ + while (0) +# endif /* not DEFINED_ONCE */ + +/* Registers are set to a sentinel when they haven't yet matched. */ +static CHAR_T PREFIX(reg_unset_dummy); +# define REG_UNSET_VALUE (&PREFIX(reg_unset_dummy)) +# define REG_UNSET(e) ((e) == REG_UNSET_VALUE) + +/* Subroutine declarations and macros for regex_compile. */ +static void PREFIX(store_op1) _RE_ARGS ((re_opcode_t op, UCHAR_T *loc, int arg)); +static void PREFIX(store_op2) _RE_ARGS ((re_opcode_t op, UCHAR_T *loc, + int arg1, int arg2)); +static void PREFIX(insert_op1) _RE_ARGS ((re_opcode_t op, UCHAR_T *loc, + int arg, UCHAR_T *end)); +static void PREFIX(insert_op2) _RE_ARGS ((re_opcode_t op, UCHAR_T *loc, + int arg1, int arg2, UCHAR_T *end)); +static boolean PREFIX(at_begline_loc_p) _RE_ARGS ((const CHAR_T *pattern, + const CHAR_T *p, + reg_syntax_t syntax)); +static boolean PREFIX(at_endline_loc_p) _RE_ARGS ((const CHAR_T *p, + const CHAR_T *pend, + reg_syntax_t syntax)); +# ifdef WCHAR +static reg_errcode_t wcs_compile_range _RE_ARGS ((CHAR_T range_start, + const CHAR_T **p_ptr, + const CHAR_T *pend, + char *translate, + reg_syntax_t syntax, + UCHAR_T *b, + CHAR_T *char_set)); +static void insert_space _RE_ARGS ((int num, CHAR_T *loc, CHAR_T *end)); +# else /* BYTE */ +static reg_errcode_t byte_compile_range _RE_ARGS ((unsigned int range_start, + const char **p_ptr, + const char *pend, + char *translate, + reg_syntax_t syntax, + unsigned char *b)); +# endif /* WCHAR */ + +/* Fetch the next character in the uncompiled pattern---translating it + if necessary. Also cast from a signed character in the constant + string passed to us by the user to an unsigned char that we can use + as an array index (in, e.g., `translate'). */ +/* ifdef MBS_SUPPORT, we translate only if character <= 0xff, + because it is impossible to allocate 4GB array for some encodings + which have 4 byte character_set like UCS4. */ +# ifndef PATFETCH +# ifdef WCHAR +# define PATFETCH(c) \ + do {if (p == pend) return REG_EEND; \ + c = (UCHAR_T) *p++; \ + if (translate && (c <= 0xff)) c = (UCHAR_T) translate[c]; \ + } while (0) +# else /* BYTE */ +# define PATFETCH(c) \ + do {if (p == pend) return REG_EEND; \ + c = (unsigned char) *p++; \ + if (translate) c = (unsigned char) translate[c]; \ + } while (0) +# endif /* WCHAR */ +# endif + +/* Fetch the next character in the uncompiled pattern, with no + translation. */ +# define PATFETCH_RAW(c) \ + do {if (p == pend) return REG_EEND; \ + c = (UCHAR_T) *p++; \ + } while (0) + +/* Go backwards one character in the pattern. */ +# define PATUNFETCH p-- + + +/* If `translate' is non-null, return translate[D], else just D. We + cast the subscript to translate because some data is declared as + `char *', to avoid warnings when a string constant is passed. But + when we use a character as a subscript we must make it unsigned. */ +/* ifdef MBS_SUPPORT, we translate only if character <= 0xff, + because it is impossible to allocate 4GB array for some encodings + which have 4 byte character_set like UCS4. */ + +# ifndef TRANSLATE +# ifdef WCHAR +# define TRANSLATE(d) \ + ((translate && ((UCHAR_T) (d)) <= 0xff) \ + ? (char) translate[(unsigned char) (d)] : (d)) +# else /* BYTE */ +# define TRANSLATE(d) \ + (translate ? (char) translate[(unsigned char) (d)] : (d)) +# endif /* WCHAR */ +# endif + + +/* Macros for outputting the compiled pattern into `buffer'. */ + +/* If the buffer isn't allocated when it comes in, use this. */ +# define INIT_BUF_SIZE (32 * sizeof(UCHAR_T)) + +/* Make sure we have at least N more bytes of space in buffer. */ +# ifdef WCHAR +# define GET_BUFFER_SPACE(n) \ + while (((unsigned long)b - (unsigned long)COMPILED_BUFFER_VAR \ + + (n)*sizeof(CHAR_T)) > bufp->allocated) \ + EXTEND_BUFFER () +# else /* BYTE */ +# define GET_BUFFER_SPACE(n) \ + while ((unsigned long) (b - bufp->buffer + (n)) > bufp->allocated) \ + EXTEND_BUFFER () +# endif /* WCHAR */ + +/* Make sure we have one more byte of buffer space and then add C to it. */ +# define BUF_PUSH(c) \ + do { \ + GET_BUFFER_SPACE (1); \ + *b++ = (UCHAR_T) (c); \ + } while (0) + + +/* Ensure we have two more bytes of buffer space and then append C1 and C2. */ +# define BUF_PUSH_2(c1, c2) \ + do { \ + GET_BUFFER_SPACE (2); \ + *b++ = (UCHAR_T) (c1); \ + *b++ = (UCHAR_T) (c2); \ + } while (0) + + +/* As with BUF_PUSH_2, except for three bytes. */ +# define BUF_PUSH_3(c1, c2, c3) \ + do { \ + GET_BUFFER_SPACE (3); \ + *b++ = (UCHAR_T) (c1); \ + *b++ = (UCHAR_T) (c2); \ + *b++ = (UCHAR_T) (c3); \ + } while (0) + +/* Store a jump with opcode OP at LOC to location TO. We store a + relative address offset by the three bytes the jump itself occupies. */ +# define STORE_JUMP(op, loc, to) \ + PREFIX(store_op1) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE))) + +/* Likewise, for a two-argument jump. */ +# define STORE_JUMP2(op, loc, to, arg) \ + PREFIX(store_op2) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)), arg) + +/* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ +# define INSERT_JUMP(op, loc, to) \ + PREFIX(insert_op1) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)), b) + +/* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ +# define INSERT_JUMP2(op, loc, to, arg) \ + PREFIX(insert_op2) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)),\ + arg, b) + +/* This is not an arbitrary limit: the arguments which represent offsets + into the pattern are two bytes long. So if 2^16 bytes turns out to + be too small, many things would have to change. */ +/* Any other compiler which, like MSC, has allocation limit below 2^16 + bytes will have to use approach similar to what was done below for + MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up + reallocating to 0 bytes. Such thing is not going to work too well. + You have been warned!! */ +# ifndef DEFINED_ONCE +# if defined _MSC_VER && !defined WIN32 +/* Microsoft C 16-bit versions limit malloc to approx 65512 bytes. + The REALLOC define eliminates a flurry of conversion warnings, + but is not required. */ +# define MAX_BUF_SIZE 65500L +# define REALLOC(p,s) realloc ((p), (size_t) (s)) +# else +# define MAX_BUF_SIZE (1L << 16) +# define REALLOC(p,s) realloc ((p), (s)) +# endif + +/* Extend the buffer by twice its current size via realloc and + reset the pointers that pointed into the old block to point to the + correct places in the new one. If extending the buffer results in it + being larger than MAX_BUF_SIZE, then flag memory exhausted. */ +# if __BOUNDED_POINTERS__ +# define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) +# define MOVE_BUFFER_POINTER(P) \ + (__ptrlow (P) += incr, SET_HIGH_BOUND (P), __ptrvalue (P) += incr) +# define ELSE_EXTEND_BUFFER_HIGH_BOUND \ + else \ + { \ + SET_HIGH_BOUND (b); \ + SET_HIGH_BOUND (begalt); \ + if (fixup_alt_jump) \ + SET_HIGH_BOUND (fixup_alt_jump); \ + if (laststart) \ + SET_HIGH_BOUND (laststart); \ + if (pending_exact) \ + SET_HIGH_BOUND (pending_exact); \ + } +# else +# define MOVE_BUFFER_POINTER(P) (P) += incr +# define ELSE_EXTEND_BUFFER_HIGH_BOUND +# endif +# endif /* not DEFINED_ONCE */ + +# ifdef WCHAR +# define EXTEND_BUFFER() \ + do { \ + UCHAR_T *old_buffer = COMPILED_BUFFER_VAR; \ + int wchar_count; \ + if (bufp->allocated + sizeof(UCHAR_T) > MAX_BUF_SIZE) \ + return REG_ESIZE; \ + bufp->allocated <<= 1; \ + if (bufp->allocated > MAX_BUF_SIZE) \ + bufp->allocated = MAX_BUF_SIZE; \ + /* How many characters the new buffer can have? */ \ + wchar_count = bufp->allocated / sizeof(UCHAR_T); \ + if (wchar_count == 0) wchar_count = 1; \ + /* Truncate the buffer to CHAR_T align. */ \ + bufp->allocated = wchar_count * sizeof(UCHAR_T); \ + RETALLOC (COMPILED_BUFFER_VAR, wchar_count, UCHAR_T); \ + bufp->buffer = (char*)COMPILED_BUFFER_VAR; \ + if (COMPILED_BUFFER_VAR == NULL) \ + return REG_ESPACE; \ + /* If the buffer moved, move all the pointers into it. */ \ + if (old_buffer != COMPILED_BUFFER_VAR) \ + { \ + int incr = COMPILED_BUFFER_VAR - old_buffer; \ + MOVE_BUFFER_POINTER (b); \ + MOVE_BUFFER_POINTER (begalt); \ + if (fixup_alt_jump) \ + MOVE_BUFFER_POINTER (fixup_alt_jump); \ + if (laststart) \ + MOVE_BUFFER_POINTER (laststart); \ + if (pending_exact) \ + MOVE_BUFFER_POINTER (pending_exact); \ + } \ + ELSE_EXTEND_BUFFER_HIGH_BOUND \ + } while (0) +# else /* BYTE */ +# define EXTEND_BUFFER() \ + do { \ + UCHAR_T *old_buffer = COMPILED_BUFFER_VAR; \ + if (bufp->allocated == MAX_BUF_SIZE) \ + return REG_ESIZE; \ + bufp->allocated <<= 1; \ + if (bufp->allocated > MAX_BUF_SIZE) \ + bufp->allocated = MAX_BUF_SIZE; \ + bufp->buffer = (UCHAR_T *) REALLOC (COMPILED_BUFFER_VAR, \ + bufp->allocated); \ + if (COMPILED_BUFFER_VAR == NULL) \ + return REG_ESPACE; \ + /* If the buffer moved, move all the pointers into it. */ \ + if (old_buffer != COMPILED_BUFFER_VAR) \ + { \ + int incr = COMPILED_BUFFER_VAR - old_buffer; \ + MOVE_BUFFER_POINTER (b); \ + MOVE_BUFFER_POINTER (begalt); \ + if (fixup_alt_jump) \ + MOVE_BUFFER_POINTER (fixup_alt_jump); \ + if (laststart) \ + MOVE_BUFFER_POINTER (laststart); \ + if (pending_exact) \ + MOVE_BUFFER_POINTER (pending_exact); \ + } \ + ELSE_EXTEND_BUFFER_HIGH_BOUND \ + } while (0) +# endif /* WCHAR */ + +# ifndef DEFINED_ONCE +/* Since we have one byte reserved for the register number argument to + {start,stop}_memory, the maximum number of groups we can report + things about is what fits in that byte. */ +# define MAX_REGNUM 255 + +/* But patterns can have more than `MAX_REGNUM' registers. We just + ignore the excess. */ +typedef unsigned regnum_t; + + +/* Macros for the compile stack. */ + +/* Since offsets can go either forwards or backwards, this type needs to + be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ +/* int may be not enough when sizeof(int) == 2. */ +typedef long pattern_offset_t; + +typedef struct +{ + pattern_offset_t begalt_offset; + pattern_offset_t fixup_alt_jump; + pattern_offset_t inner_group_offset; + pattern_offset_t laststart_offset; + regnum_t regnum; +} compile_stack_elt_t; + + +typedef struct +{ + compile_stack_elt_t *stack; + unsigned size; + unsigned avail; /* Offset of next open position. */ +} compile_stack_type; + + +# define INIT_COMPILE_STACK_SIZE 32 + +# define COMPILE_STACK_EMPTY (compile_stack.avail == 0) +# define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) + +/* The next available element. */ +# define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) + +# endif /* not DEFINED_ONCE */ + +/* Set the bit for character C in a list. */ +# ifndef DEFINED_ONCE +# define SET_LIST_BIT(c) \ + (b[((unsigned char) (c)) / BYTEWIDTH] \ + |= 1 << (((unsigned char) c) % BYTEWIDTH)) +# endif /* DEFINED_ONCE */ + +/* Get the next unsigned number in the uncompiled pattern. */ +# define GET_UNSIGNED_NUMBER(num) \ + { \ + while (p != pend) \ + { \ + PATFETCH (c); \ + if (c < '0' || c > '9') \ + break; \ + if (num <= RE_DUP_MAX) \ + { \ + if (num < 0) \ + num = 0; \ + num = num * 10 + c - '0'; \ + } \ + } \ + } + +# ifndef DEFINED_ONCE +# if defined _LIBC || WIDE_CHAR_SUPPORT +/* The GNU C library provides support for user-defined character classes + and the functions from ISO C amendement 1. */ +# ifdef CHARCLASS_NAME_MAX +# define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX +# else +/* This shouldn't happen but some implementation might still have this + problem. Use a reasonable default value. */ +# define CHAR_CLASS_MAX_LENGTH 256 +# endif + +# ifdef _LIBC +# define IS_CHAR_CLASS(string) __wctype (string) +# else +# define IS_CHAR_CLASS(string) wctype (string) +# endif +# else +# define CHAR_CLASS_MAX_LENGTH 6 /* Namely, `xdigit'. */ + +# define IS_CHAR_CLASS(string) \ + (STREQ (string, "alpha") || STREQ (string, "upper") \ + || STREQ (string, "lower") || STREQ (string, "digit") \ + || STREQ (string, "alnum") || STREQ (string, "xdigit") \ + || STREQ (string, "space") || STREQ (string, "print") \ + || STREQ (string, "punct") || STREQ (string, "graph") \ + || STREQ (string, "cntrl") || STREQ (string, "blank")) +# endif +# endif /* DEFINED_ONCE */ + +# ifndef MATCH_MAY_ALLOCATE + +/* If we cannot allocate large objects within re_match_2_internal, + we make the fail stack and register vectors global. + The fail stack, we grow to the maximum size when a regexp + is compiled. + The register vectors, we adjust in size each time we + compile a regexp, according to the number of registers it needs. */ + +static PREFIX(fail_stack_type) fail_stack; + +/* Size with which the following vectors are currently allocated. + That is so we can make them bigger as needed, + but never make them smaller. */ +# ifdef DEFINED_ONCE +static int regs_allocated_size; + +static const char ** regstart, ** regend; +static const char ** old_regstart, ** old_regend; +static const char **best_regstart, **best_regend; +static const char **reg_dummy; +# endif /* DEFINED_ONCE */ + +static PREFIX(register_info_type) *PREFIX(reg_info); +static PREFIX(register_info_type) *PREFIX(reg_info_dummy); + +/* Make the register vectors big enough for NUM_REGS registers, + but don't make them smaller. */ + +static void +PREFIX(regex_grow_registers) (num_regs) + int num_regs; +{ + if (num_regs > regs_allocated_size) + { + RETALLOC_IF (regstart, num_regs, const char *); + RETALLOC_IF (regend, num_regs, const char *); + RETALLOC_IF (old_regstart, num_regs, const char *); + RETALLOC_IF (old_regend, num_regs, const char *); + RETALLOC_IF (best_regstart, num_regs, const char *); + RETALLOC_IF (best_regend, num_regs, const char *); + RETALLOC_IF (PREFIX(reg_info), num_regs, PREFIX(register_info_type)); + RETALLOC_IF (reg_dummy, num_regs, const char *); + RETALLOC_IF (PREFIX(reg_info_dummy), num_regs, PREFIX(register_info_type)); + + regs_allocated_size = num_regs; + } +} + +# endif /* not MATCH_MAY_ALLOCATE */ + +# ifndef DEFINED_ONCE +static boolean group_in_compile_stack _RE_ARGS ((compile_stack_type + compile_stack, + regnum_t regnum)); +# endif /* not DEFINED_ONCE */ + +/* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. + Returns one of error codes defined in `regex.h', or zero for success. + + Assumes the `allocated' (and perhaps `buffer') and `translate' + fields are set in BUFP on entry. + + If it succeeds, results are put in BUFP (if it returns an error, the + contents of BUFP are undefined): + `buffer' is the compiled pattern; + `syntax' is set to SYNTAX; + `used' is set to the length of the compiled pattern; + `fastmap_accurate' is zero; + `re_nsub' is the number of subexpressions in PATTERN; + `not_bol' and `not_eol' are zero; + + The `fastmap' and `newline_anchor' fields are neither + examined nor set. */ + +/* Return, freeing storage we allocated. */ +# ifdef WCHAR +# define FREE_STACK_RETURN(value) \ + return (free(pattern), free(mbs_offset), free(is_binary), free (compile_stack.stack), value) +# else +# define FREE_STACK_RETURN(value) \ + return (free (compile_stack.stack), value) +# endif /* WCHAR */ + +static reg_errcode_t +PREFIX(regex_compile) (ARG_PREFIX(pattern), ARG_PREFIX(size), syntax, bufp) + const char *ARG_PREFIX(pattern); + size_t ARG_PREFIX(size); + reg_syntax_t syntax; + struct re_pattern_buffer *bufp; +{ + /* We fetch characters from PATTERN here. Even though PATTERN is + `char *' (i.e., signed), we declare these variables as unsigned, so + they can be reliably used as array indices. */ + register UCHAR_T c, c1; + +#ifdef WCHAR + /* A temporary space to keep wchar_t pattern and compiled pattern. */ + CHAR_T *pattern, *COMPILED_BUFFER_VAR; + size_t size; + /* offset buffer for optimization. See convert_mbs_to_wc. */ + int *mbs_offset = NULL; + /* It hold whether each wchar_t is binary data or not. */ + char *is_binary = NULL; + /* A flag whether exactn is handling binary data or not. */ + char is_exactn_bin = FALSE; +#endif /* WCHAR */ + + /* A random temporary spot in PATTERN. */ + const CHAR_T *p1; + + /* Points to the end of the buffer, where we should append. */ + register UCHAR_T *b; + + /* Keeps track of unclosed groups. */ + compile_stack_type compile_stack; + + /* Points to the current (ending) position in the pattern. */ +#ifdef WCHAR + const CHAR_T *p; + const CHAR_T *pend; +#else /* BYTE */ + const CHAR_T *p = pattern; + const CHAR_T *pend = pattern + size; +#endif /* WCHAR */ + + /* How to translate the characters in the pattern. */ + RE_TRANSLATE_TYPE translate = bufp->translate; + + /* Address of the count-byte of the most recently inserted `exactn' + command. This makes it possible to tell if a new exact-match + character can be added to that command or if the character requires + a new `exactn' command. */ + UCHAR_T *pending_exact = 0; + + /* Address of start of the most recently finished expression. + This tells, e.g., postfix * where to find the start of its + operand. Reset at the beginning of groups and alternatives. */ + UCHAR_T *laststart = 0; + + /* Address of beginning of regexp, or inside of last group. */ + UCHAR_T *begalt; + + /* Address of the place where a forward jump should go to the end of + the containing expression. Each alternative of an `or' -- except the + last -- ends with a forward jump of this sort. */ + UCHAR_T *fixup_alt_jump = 0; + + /* Counts open-groups as they are encountered. Remembered for the + matching close-group on the compile stack, so the same register + number is put in the stop_memory as the start_memory. */ + regnum_t regnum = 0; + +#ifdef WCHAR + /* Initialize the wchar_t PATTERN and offset_buffer. */ + p = pend = pattern = TALLOC(csize + 1, CHAR_T); + mbs_offset = TALLOC(csize + 1, int); + is_binary = TALLOC(csize + 1, char); + if (pattern == NULL || mbs_offset == NULL || is_binary == NULL) + { + free(pattern); + free(mbs_offset); + free(is_binary); + return REG_ESPACE; + } + pattern[csize] = L'\0'; /* sentinel */ + size = convert_mbs_to_wcs(pattern, cpattern, csize, mbs_offset, is_binary); + pend = p + size; + if (size < 0) + { + free(pattern); + free(mbs_offset); + free(is_binary); + return REG_BADPAT; + } +#endif + +#ifdef DEBUG + DEBUG_PRINT1 ("\nCompiling pattern: "); + if (debug) + { + unsigned debug_count; + + for (debug_count = 0; debug_count < size; debug_count++) + PUT_CHAR (pattern[debug_count]); + putchar ('\n'); + } +#endif /* DEBUG */ + + /* Initialize the compile stack. */ + compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); + if (compile_stack.stack == NULL) + { +#ifdef WCHAR + free(pattern); + free(mbs_offset); + free(is_binary); +#endif + return REG_ESPACE; + } + + compile_stack.size = INIT_COMPILE_STACK_SIZE; + compile_stack.avail = 0; + + /* Initialize the pattern buffer. */ + bufp->syntax = syntax; + bufp->fastmap_accurate = 0; + bufp->not_bol = bufp->not_eol = 0; + + /* Set `used' to zero, so that if we return an error, the pattern + printer (for debugging) will think there's no pattern. We reset it + at the end. */ + bufp->used = 0; + + /* Always count groups, whether or not bufp->no_sub is set. */ + bufp->re_nsub = 0; + +#if !defined emacs && !defined SYNTAX_TABLE + /* Initialize the syntax table. */ + init_syntax_once (); +#endif + + if (bufp->allocated == 0) + { + if (bufp->buffer) + { /* If zero allocated, but buffer is non-null, try to realloc + enough space. This loses if buffer's address is bogus, but + that is the user's responsibility. */ +#ifdef WCHAR + /* Free bufp->buffer and allocate an array for wchar_t pattern + buffer. */ + free(bufp->buffer); + COMPILED_BUFFER_VAR = TALLOC (INIT_BUF_SIZE/sizeof(UCHAR_T), + UCHAR_T); +#else + RETALLOC (COMPILED_BUFFER_VAR, INIT_BUF_SIZE, UCHAR_T); +#endif /* WCHAR */ + } + else + { /* Caller did not allocate a buffer. Do it for them. */ + COMPILED_BUFFER_VAR = TALLOC (INIT_BUF_SIZE / sizeof(UCHAR_T), + UCHAR_T); + } + + if (!COMPILED_BUFFER_VAR) FREE_STACK_RETURN (REG_ESPACE); +#ifdef WCHAR + bufp->buffer = (char*)COMPILED_BUFFER_VAR; +#endif /* WCHAR */ + bufp->allocated = INIT_BUF_SIZE; + } +#ifdef WCHAR + else + COMPILED_BUFFER_VAR = (UCHAR_T*) bufp->buffer; +#endif + + begalt = b = COMPILED_BUFFER_VAR; + + /* Loop through the uncompiled pattern until we're at the end. */ + while (p != pend) + { + PATFETCH (c); + + switch (c) + { + case '^': + { + if ( /* If at start of pattern, it's an operator. */ + p == pattern + 1 + /* If context independent, it's an operator. */ + || syntax & RE_CONTEXT_INDEP_ANCHORS + /* Otherwise, depends on what's come before. */ + || PREFIX(at_begline_loc_p) (pattern, p, syntax)) + BUF_PUSH (begline); + else + goto normal_char; + } + break; + + + case '$': + { + if ( /* If at end of pattern, it's an operator. */ + p == pend + /* If context independent, it's an operator. */ + || syntax & RE_CONTEXT_INDEP_ANCHORS + /* Otherwise, depends on what's next. */ + || PREFIX(at_endline_loc_p) (p, pend, syntax)) + BUF_PUSH (endline); + else + goto normal_char; + } + break; + + + case '+': + case '?': + if ((syntax & RE_BK_PLUS_QM) + || (syntax & RE_LIMITED_OPS)) + goto normal_char; + handle_plus: + case '*': + /* If there is no previous pattern... */ + if (!laststart) + { + if (syntax & RE_CONTEXT_INVALID_OPS) + FREE_STACK_RETURN (REG_BADRPT); + else if (!(syntax & RE_CONTEXT_INDEP_OPS)) + goto normal_char; + } + + { + /* Are we optimizing this jump? */ + boolean keep_string_p = false; + + /* 1 means zero (many) matches is allowed. */ + char zero_times_ok = 0, many_times_ok = 0; + + /* If there is a sequence of repetition chars, collapse it + down to just one (the right one). We can't combine + interval operators with these because of, e.g., `a{2}*', + which should only match an even number of `a's. */ + + for (;;) + { + zero_times_ok |= c != '+'; + many_times_ok |= c != '?'; + + if (p == pend) + break; + + PATFETCH (c); + + if (c == '*' + || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?'))) + ; + + else if (syntax & RE_BK_PLUS_QM && c == '\\') + { + if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); + + PATFETCH (c1); + if (!(c1 == '+' || c1 == '?')) + { + PATUNFETCH; + PATUNFETCH; + break; + } + + c = c1; + } + else + { + PATUNFETCH; + break; + } + + /* If we get here, we found another repeat character. */ + } + + /* Star, etc. applied to an empty pattern is equivalent + to an empty pattern. */ + if (!laststart) + break; + + /* Now we know whether or not zero matches is allowed + and also whether or not two or more matches is allowed. */ + if (many_times_ok) + { /* More than one repetition is allowed, so put in at the + end a backward relative jump from `b' to before the next + jump we're going to put in below (which jumps from + laststart to after this jump). + + But if we are at the `*' in the exact sequence `.*\n', + insert an unconditional jump backwards to the ., + instead of the beginning of the loop. This way we only + push a failure point once, instead of every time + through the loop. */ + assert (p - 1 > pattern); + + /* Allocate the space for the jump. */ + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + + /* We know we are not at the first character of the pattern, + because laststart was nonzero. And we've already + incremented `p', by the way, to be the character after + the `*'. Do we have to do something analogous here + for null bytes, because of RE_DOT_NOT_NULL? */ + if (TRANSLATE (*(p - 2)) == TRANSLATE ('.') + && zero_times_ok + && p < pend && TRANSLATE (*p) == TRANSLATE ('\n') + && !(syntax & RE_DOT_NEWLINE)) + { /* We have .*\n. */ + STORE_JUMP (jump, b, laststart); + keep_string_p = true; + } + else + /* Anything else. */ + STORE_JUMP (maybe_pop_jump, b, laststart - + (1 + OFFSET_ADDRESS_SIZE)); + + /* We've added more stuff to the buffer. */ + b += 1 + OFFSET_ADDRESS_SIZE; + } + + /* On failure, jump from laststart to b + 3, which will be the + end of the buffer after this jump is inserted. */ + /* ifdef WCHAR, 'b + 1 + OFFSET_ADDRESS_SIZE' instead of + 'b + 3'. */ + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump + : on_failure_jump, + laststart, b + 1 + OFFSET_ADDRESS_SIZE); + pending_exact = 0; + b += 1 + OFFSET_ADDRESS_SIZE; + + if (!zero_times_ok) + { + /* At least one repetition is required, so insert a + `dummy_failure_jump' before the initial + `on_failure_jump' instruction of the loop. This + effects a skip over that instruction the first time + we hit that loop. */ + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + INSERT_JUMP (dummy_failure_jump, laststart, laststart + + 2 + 2 * OFFSET_ADDRESS_SIZE); + b += 1 + OFFSET_ADDRESS_SIZE; + } + } + break; + + + case '.': + laststart = b; + BUF_PUSH (anychar); + break; + + + case '[': + { + boolean had_char_class = false; +#ifdef WCHAR + CHAR_T range_start = 0xffffffff; +#else + unsigned int range_start = 0xffffffff; +#endif + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + +#ifdef WCHAR + /* We assume a charset(_not) structure as a wchar_t array. + charset[0] = (re_opcode_t) charset(_not) + charset[1] = l (= length of char_classes) + charset[2] = m (= length of collating_symbols) + charset[3] = n (= length of equivalence_classes) + charset[4] = o (= length of char_ranges) + charset[5] = p (= length of chars) + + charset[6] = char_class (wctype_t) + charset[6+CHAR_CLASS_SIZE] = char_class (wctype_t) + ... + charset[l+5] = char_class (wctype_t) + + charset[l+6] = collating_symbol (wchar_t) + ... + charset[l+m+5] = collating_symbol (wchar_t) + ifdef _LIBC we use the index if + _NL_COLLATE_SYMB_EXTRAMB instead of + wchar_t string. + + charset[l+m+6] = equivalence_classes (wchar_t) + ... + charset[l+m+n+5] = equivalence_classes (wchar_t) + ifdef _LIBC we use the index in + _NL_COLLATE_WEIGHT instead of + wchar_t string. + + charset[l+m+n+6] = range_start + charset[l+m+n+7] = range_end + ... + charset[l+m+n+2o+4] = range_start + charset[l+m+n+2o+5] = range_end + ifdef _LIBC we use the value looked up + in _NL_COLLATE_COLLSEQ instead of + wchar_t character. + + charset[l+m+n+2o+6] = char + ... + charset[l+m+n+2o+p+5] = char + + */ + + /* We need at least 6 spaces: the opcode, the length of + char_classes, the length of collating_symbols, the length of + equivalence_classes, the length of char_ranges, the length of + chars. */ + GET_BUFFER_SPACE (6); + + /* Save b as laststart. And We use laststart as the pointer + to the first element of the charset here. + In other words, laststart[i] indicates charset[i]. */ + laststart = b; + + /* We test `*p == '^' twice, instead of using an if + statement, so we only need one BUF_PUSH. */ + BUF_PUSH (*p == '^' ? charset_not : charset); + if (*p == '^') + p++; + + /* Push the length of char_classes, the length of + collating_symbols, the length of equivalence_classes, the + length of char_ranges and the length of chars. */ + BUF_PUSH_3 (0, 0, 0); + BUF_PUSH_2 (0, 0); + + /* Remember the first position in the bracket expression. */ + p1 = p; + + /* charset_not matches newline according to a syntax bit. */ + if ((re_opcode_t) b[-6] == charset_not + && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) + { + BUF_PUSH('\n'); + laststart[5]++; /* Update the length of characters */ + } + + /* Read in characters and ranges, setting map bits. */ + for (;;) + { + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + PATFETCH (c); + + /* \ might escape characters inside [...] and [^...]. */ + if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') + { + if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); + + PATFETCH (c1); + BUF_PUSH(c1); + laststart[5]++; /* Update the length of chars */ + range_start = c1; + continue; + } + + /* Could be the end of the bracket expression. If it's + not (i.e., when the bracket expression is `[]' so + far), the ']' character bit gets set way below. */ + if (c == ']' && p != p1 + 1) + break; + + /* Look ahead to see if it's a range when the last thing + was a character class. */ + if (had_char_class && c == '-' && *p != ']') + FREE_STACK_RETURN (REG_ERANGE); + + /* Look ahead to see if it's a range when the last thing + was a character: if this is a hyphen not at the + beginning or the end of a list, then it's the range + operator. */ + if (c == '-' + && !(p - 2 >= pattern && p[-2] == '[') + && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') + && *p != ']') + { + reg_errcode_t ret; + /* Allocate the space for range_start and range_end. */ + GET_BUFFER_SPACE (2); + /* Update the pointer to indicate end of buffer. */ + b += 2; + ret = wcs_compile_range (range_start, &p, pend, translate, + syntax, b, laststart); + if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); + range_start = 0xffffffff; + } + else if (p[0] == '-' && p[1] != ']') + { /* This handles ranges made up of characters only. */ + reg_errcode_t ret; + + /* Move past the `-'. */ + PATFETCH (c1); + /* Allocate the space for range_start and range_end. */ + GET_BUFFER_SPACE (2); + /* Update the pointer to indicate end of buffer. */ + b += 2; + ret = wcs_compile_range (c, &p, pend, translate, syntax, b, + laststart); + if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); + range_start = 0xffffffff; + } + + /* See if we're at the beginning of a possible character + class. */ + else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') + { /* Leave room for the null. */ + char str[CHAR_CLASS_MAX_LENGTH + 1]; + + PATFETCH (c); + c1 = 0; + + /* If pattern is `[[:'. */ + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (;;) + { + PATFETCH (c); + if ((c == ':' && *p == ']') || p == pend) + break; + if (c1 < CHAR_CLASS_MAX_LENGTH) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[c1] = '\0'; + + /* If isn't a word bracketed by `[:' and `:]': + undo the ending character, the letters, and leave + the leading `:' and `[' (but store them as character). */ + if (c == ':' && *p == ']') + { + wctype_t wt; + uintptr_t alignedp; + + /* Query the character class as wctype_t. */ + wt = IS_CHAR_CLASS (str); + if (wt == 0) + FREE_STACK_RETURN (REG_ECTYPE); + + /* Throw away the ] at the end of the character + class. */ + PATFETCH (c); + + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + /* Allocate the space for character class. */ + GET_BUFFER_SPACE(CHAR_CLASS_SIZE); + /* Update the pointer to indicate end of buffer. */ + b += CHAR_CLASS_SIZE; + /* Move data which follow character classes + not to violate the data. */ + insert_space(CHAR_CLASS_SIZE, + laststart + 6 + laststart[1], + b - 1); + alignedp = ((uintptr_t)(laststart + 6 + laststart[1]) + + __alignof__(wctype_t) - 1) + & ~(uintptr_t)(__alignof__(wctype_t) - 1); + /* Store the character class. */ + *((wctype_t*)alignedp) = wt; + /* Update length of char_classes */ + laststart[1] += CHAR_CLASS_SIZE; + + had_char_class = true; + } + else + { + c1++; + while (c1--) + PATUNFETCH; + BUF_PUSH ('['); + BUF_PUSH (':'); + laststart[5] += 2; /* Update the length of characters */ + range_start = ':'; + had_char_class = false; + } + } + else if (syntax & RE_CHAR_CLASSES && c == '[' && (*p == '=' + || *p == '.')) + { + CHAR_T str[128]; /* Should be large enough. */ + CHAR_T delim = *p; /* '=' or '.' */ +# ifdef _LIBC + uint32_t nrules = + _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); +# endif + PATFETCH (c); + c1 = 0; + + /* If pattern is `[[=' or '[[.'. */ + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (;;) + { + PATFETCH (c); + if ((c == delim && *p == ']') || p == pend) + break; + if (c1 < sizeof (str) - 1) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[c1] = '\0'; + + if (c == delim && *p == ']' && str[0] != '\0') + { + unsigned int i, offset; + /* If we have no collation data we use the default + collation in which each character is in a class + by itself. It also means that ASCII is the + character set and therefore we cannot have character + with more than one byte in the multibyte + representation. */ + + /* If not defined _LIBC, we push the name and + `\0' for the sake of matching performance. */ + int datasize = c1 + 1; + +# ifdef _LIBC + int32_t idx = 0; + if (nrules == 0) +# endif + { + if (c1 != 1) + FREE_STACK_RETURN (REG_ECOLLATE); + } +# ifdef _LIBC + else + { + const int32_t *table; + const int32_t *weights; + const int32_t *extra; + const int32_t *indirect; + wint_t *cp; + + /* This #include defines a local function! */ +# include + + if(delim == '=') + { + /* We push the index for equivalence class. */ + cp = (wint_t*)str; + + table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_TABLEWC); + weights = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_WEIGHTWC); + extra = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_EXTRAWC); + indirect = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_INDIRECTWC); + + idx = findidx ((const wint_t**)&cp); + if (idx == 0 || cp < (wint_t*) str + c1) + /* This is no valid character. */ + FREE_STACK_RETURN (REG_ECOLLATE); + + str[0] = (wchar_t)idx; + } + else /* delim == '.' */ + { + /* We push collation sequence value + for collating symbol. */ + int32_t table_size; + const int32_t *symb_table; + const unsigned char *extra; + int32_t idx; + int32_t elem; + int32_t second; + int32_t hash; + char char_str[c1]; + + /* We have to convert the name to a single-byte + string. This is possible since the names + consist of ASCII characters and the internal + representation is UCS4. */ + for (i = 0; i < c1; ++i) + char_str[i] = str[i]; + + table_size = + _NL_CURRENT_WORD (LC_COLLATE, + _NL_COLLATE_SYMB_HASH_SIZEMB); + symb_table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_TABLEMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_EXTRAMB); + + /* Locate the character in the hashing table. */ + hash = elem_hash (char_str, c1); + + idx = 0; + elem = hash % table_size; + second = hash % (table_size - 2); + while (symb_table[2 * elem] != 0) + { + /* First compare the hashing value. */ + if (symb_table[2 * elem] == hash + && c1 == extra[symb_table[2 * elem + 1]] + && memcmp (str, + &extra[symb_table[2 * elem + 1] + + 1], c1) == 0) + { + /* Yep, this is the entry. */ + idx = symb_table[2 * elem + 1]; + idx += 1 + extra[idx]; + break; + } + + /* Next entry. */ + elem += second; + } + + if (symb_table[2 * elem] != 0) + { + /* Compute the index of the byte sequence + in the table. */ + idx += 1 + extra[idx]; + /* Adjust for the alignment. */ + idx = (idx + 3) & ~4; + + str[0] = (wchar_t) idx + 4; + } + else if (symb_table[2 * elem] == 0 && c1 == 1) + { + /* No valid character. Match it as a + single byte character. */ + had_char_class = false; + BUF_PUSH(str[0]); + /* Update the length of characters */ + laststart[5]++; + range_start = str[0]; + + /* Throw away the ] at the end of the + collating symbol. */ + PATFETCH (c); + /* exit from the switch block. */ + continue; + } + else + FREE_STACK_RETURN (REG_ECOLLATE); + } + datasize = 1; + } +# endif + /* Throw away the ] at the end of the equivalence + class (or collating symbol). */ + PATFETCH (c); + + /* Allocate the space for the equivalence class + (or collating symbol) (and '\0' if needed). */ + GET_BUFFER_SPACE(datasize); + /* Update the pointer to indicate end of buffer. */ + b += datasize; + + if (delim == '=') + { /* equivalence class */ + /* Calculate the offset of char_ranges, + which is next to equivalence_classes. */ + offset = laststart[1] + laststart[2] + + laststart[3] +6; + /* Insert space. */ + insert_space(datasize, laststart + offset, b - 1); + + /* Write the equivalence_class and \0. */ + for (i = 0 ; i < datasize ; i++) + laststart[offset + i] = str[i]; + + /* Update the length of equivalence_classes. */ + laststart[3] += datasize; + had_char_class = true; + } + else /* delim == '.' */ + { /* collating symbol */ + /* Calculate the offset of the equivalence_classes, + which is next to collating_symbols. */ + offset = laststart[1] + laststart[2] + 6; + /* Insert space and write the collationg_symbol + and \0. */ + insert_space(datasize, laststart + offset, b-1); + for (i = 0 ; i < datasize ; i++) + laststart[offset + i] = str[i]; + + /* In re_match_2_internal if range_start < -1, we + assume -range_start is the offset of the + collating symbol which is specified as + the character of the range start. So we assign + -(laststart[1] + laststart[2] + 6) to + range_start. */ + range_start = -(laststart[1] + laststart[2] + 6); + /* Update the length of collating_symbol. */ + laststart[2] += datasize; + had_char_class = false; + } + } + else + { + c1++; + while (c1--) + PATUNFETCH; + BUF_PUSH ('['); + BUF_PUSH (delim); + laststart[5] += 2; /* Update the length of characters */ + range_start = delim; + had_char_class = false; + } + } + else + { + had_char_class = false; + BUF_PUSH(c); + laststart[5]++; /* Update the length of characters */ + range_start = c; + } + } + +#else /* BYTE */ + /* Ensure that we have enough space to push a charset: the + opcode, the length count, and the bitset; 34 bytes in all. */ + GET_BUFFER_SPACE (34); + + laststart = b; + + /* We test `*p == '^' twice, instead of using an if + statement, so we only need one BUF_PUSH. */ + BUF_PUSH (*p == '^' ? charset_not : charset); + if (*p == '^') + p++; + + /* Remember the first position in the bracket expression. */ + p1 = p; + + /* Push the number of bytes in the bitmap. */ + BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); + + /* Clear the whole map. */ + bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); + + /* charset_not matches newline according to a syntax bit. */ + if ((re_opcode_t) b[-2] == charset_not + && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) + SET_LIST_BIT ('\n'); + + /* Read in characters and ranges, setting map bits. */ + for (;;) + { + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + PATFETCH (c); + + /* \ might escape characters inside [...] and [^...]. */ + if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') + { + if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); + + PATFETCH (c1); + SET_LIST_BIT (c1); + range_start = c1; + continue; + } + + /* Could be the end of the bracket expression. If it's + not (i.e., when the bracket expression is `[]' so + far), the ']' character bit gets set way below. */ + if (c == ']' && p != p1 + 1) + break; + + /* Look ahead to see if it's a range when the last thing + was a character class. */ + if (had_char_class && c == '-' && *p != ']') + FREE_STACK_RETURN (REG_ERANGE); + + /* Look ahead to see if it's a range when the last thing + was a character: if this is a hyphen not at the + beginning or the end of a list, then it's the range + operator. */ + if (c == '-' + && !(p - 2 >= pattern && p[-2] == '[') + && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') + && *p != ']') + { + reg_errcode_t ret + = byte_compile_range (range_start, &p, pend, translate, + syntax, b); + if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); + range_start = 0xffffffff; + } + + else if (p[0] == '-' && p[1] != ']') + { /* This handles ranges made up of characters only. */ + reg_errcode_t ret; + + /* Move past the `-'. */ + PATFETCH (c1); + + ret = byte_compile_range (c, &p, pend, translate, syntax, b); + if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); + range_start = 0xffffffff; + } + + /* See if we're at the beginning of a possible character + class. */ + + else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') + { /* Leave room for the null. */ + char str[CHAR_CLASS_MAX_LENGTH + 1]; + + PATFETCH (c); + c1 = 0; + + /* If pattern is `[[:'. */ + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (;;) + { + PATFETCH (c); + if ((c == ':' && *p == ']') || p == pend) + break; + if (c1 < CHAR_CLASS_MAX_LENGTH) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[c1] = '\0'; + + /* If isn't a word bracketed by `[:' and `:]': + undo the ending character, the letters, and leave + the leading `:' and `[' (but set bits for them). */ + if (c == ':' && *p == ']') + { +# if defined _LIBC || WIDE_CHAR_SUPPORT + boolean is_lower = STREQ (str, "lower"); + boolean is_upper = STREQ (str, "upper"); + wctype_t wt; + int ch; + + wt = IS_CHAR_CLASS (str); + if (wt == 0) + FREE_STACK_RETURN (REG_ECTYPE); + + /* Throw away the ] at the end of the character + class. */ + PATFETCH (c); + + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (ch = 0; ch < 1 << BYTEWIDTH; ++ch) + { +# ifdef _LIBC + if (__iswctype (__btowc (ch), wt)) + SET_LIST_BIT (ch); +# else + if (iswctype (btowc (ch), wt)) + SET_LIST_BIT (ch); +# endif + + if (translate && (is_upper || is_lower) + && (ISUPPER (ch) || ISLOWER (ch))) + SET_LIST_BIT (ch); + } + + had_char_class = true; +# else + int ch; + boolean is_alnum = STREQ (str, "alnum"); + boolean is_alpha = STREQ (str, "alpha"); + boolean is_blank = STREQ (str, "blank"); + boolean is_cntrl = STREQ (str, "cntrl"); + boolean is_digit = STREQ (str, "digit"); + boolean is_graph = STREQ (str, "graph"); + boolean is_lower = STREQ (str, "lower"); + boolean is_print = STREQ (str, "print"); + boolean is_punct = STREQ (str, "punct"); + boolean is_space = STREQ (str, "space"); + boolean is_upper = STREQ (str, "upper"); + boolean is_xdigit = STREQ (str, "xdigit"); + + if (!IS_CHAR_CLASS (str)) + FREE_STACK_RETURN (REG_ECTYPE); + + /* Throw away the ] at the end of the character + class. */ + PATFETCH (c); + + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (ch = 0; ch < 1 << BYTEWIDTH; ch++) + { + /* This was split into 3 if's to + avoid an arbitrary limit in some compiler. */ + if ( (is_alnum && ISALNUM (ch)) + || (is_alpha && ISALPHA (ch)) + || (is_blank && ISBLANK (ch)) + || (is_cntrl && ISCNTRL (ch))) + SET_LIST_BIT (ch); + if ( (is_digit && ISDIGIT (ch)) + || (is_graph && ISGRAPH (ch)) + || (is_lower && ISLOWER (ch)) + || (is_print && ISPRINT (ch))) + SET_LIST_BIT (ch); + if ( (is_punct && ISPUNCT (ch)) + || (is_space && ISSPACE (ch)) + || (is_upper && ISUPPER (ch)) + || (is_xdigit && ISXDIGIT (ch))) + SET_LIST_BIT (ch); + if ( translate && (is_upper || is_lower) + && (ISUPPER (ch) || ISLOWER (ch))) + SET_LIST_BIT (ch); + } + had_char_class = true; +# endif /* libc || wctype.h */ + } + else + { + c1++; + while (c1--) + PATUNFETCH; + SET_LIST_BIT ('['); + SET_LIST_BIT (':'); + range_start = ':'; + had_char_class = false; + } + } + else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == '=') + { + unsigned char str[MB_LEN_MAX + 1]; +# ifdef _LIBC + uint32_t nrules = + _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); +# endif + + PATFETCH (c); + c1 = 0; + + /* If pattern is `[[='. */ + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (;;) + { + PATFETCH (c); + if ((c == '=' && *p == ']') || p == pend) + break; + if (c1 < MB_LEN_MAX) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[c1] = '\0'; + + if (c == '=' && *p == ']' && str[0] != '\0') + { + /* If we have no collation data we use the default + collation in which each character is in a class + by itself. It also means that ASCII is the + character set and therefore we cannot have character + with more than one byte in the multibyte + representation. */ +# ifdef _LIBC + if (nrules == 0) +# endif + { + if (c1 != 1) + FREE_STACK_RETURN (REG_ECOLLATE); + + /* Throw away the ] at the end of the equivalence + class. */ + PATFETCH (c); + + /* Set the bit for the character. */ + SET_LIST_BIT (str[0]); + } +# ifdef _LIBC + else + { + /* Try to match the byte sequence in `str' against + those known to the collate implementation. + First find out whether the bytes in `str' are + actually from exactly one character. */ + const int32_t *table; + const unsigned char *weights; + const unsigned char *extra; + const int32_t *indirect; + int32_t idx; + const unsigned char *cp = str; + int ch; + + /* This #include defines a local function! */ +# include + + table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); + weights = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB); + indirect = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB); + + idx = findidx (&cp); + if (idx == 0 || cp < str + c1) + /* This is no valid character. */ + FREE_STACK_RETURN (REG_ECOLLATE); + + /* Throw away the ] at the end of the equivalence + class. */ + PATFETCH (c); + + /* Now we have to go throught the whole table + and find all characters which have the same + first level weight. + + XXX Note that this is not entirely correct. + we would have to match multibyte sequences + but this is not possible with the current + implementation. */ + for (ch = 1; ch < 256; ++ch) + /* XXX This test would have to be changed if we + would allow matching multibyte sequences. */ + if (table[ch] > 0) + { + int32_t idx2 = table[ch]; + size_t len = weights[idx2]; + + /* Test whether the lenghts match. */ + if (weights[idx] == len) + { + /* They do. New compare the bytes of + the weight. */ + size_t cnt = 0; + + while (cnt < len + && (weights[idx + 1 + cnt] + == weights[idx2 + 1 + cnt])) + ++cnt; + + if (cnt == len) + /* They match. Mark the character as + acceptable. */ + SET_LIST_BIT (ch); + } + } + } +# endif + had_char_class = true; + } + else + { + c1++; + while (c1--) + PATUNFETCH; + SET_LIST_BIT ('['); + SET_LIST_BIT ('='); + range_start = '='; + had_char_class = false; + } + } + else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == '.') + { + unsigned char str[128]; /* Should be large enough. */ +# ifdef _LIBC + uint32_t nrules = + _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); +# endif + + PATFETCH (c); + c1 = 0; + + /* If pattern is `[[.'. */ + if (p == pend) FREE_STACK_RETURN (REG_EBRACK); + + for (;;) + { + PATFETCH (c); + if ((c == '.' && *p == ']') || p == pend) + break; + if (c1 < sizeof (str)) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[c1] = '\0'; + + if (c == '.' && *p == ']' && str[0] != '\0') + { + /* If we have no collation data we use the default + collation in which each character is the name + for its own class which contains only the one + character. It also means that ASCII is the + character set and therefore we cannot have character + with more than one byte in the multibyte + representation. */ +# ifdef _LIBC + if (nrules == 0) +# endif + { + if (c1 != 1) + FREE_STACK_RETURN (REG_ECOLLATE); + + /* Throw away the ] at the end of the equivalence + class. */ + PATFETCH (c); + + /* Set the bit for the character. */ + SET_LIST_BIT (str[0]); + range_start = ((const unsigned char *) str)[0]; + } +# ifdef _LIBC + else + { + /* Try to match the byte sequence in `str' against + those known to the collate implementation. + First find out whether the bytes in `str' are + actually from exactly one character. */ + int32_t table_size; + const int32_t *symb_table; + const unsigned char *extra; + int32_t idx; + int32_t elem; + int32_t second; + int32_t hash; + + table_size = + _NL_CURRENT_WORD (LC_COLLATE, + _NL_COLLATE_SYMB_HASH_SIZEMB); + symb_table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_TABLEMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_EXTRAMB); + + /* Locate the character in the hashing table. */ + hash = elem_hash (str, c1); + + idx = 0; + elem = hash % table_size; + second = hash % (table_size - 2); + while (symb_table[2 * elem] != 0) + { + /* First compare the hashing value. */ + if (symb_table[2 * elem] == hash + && c1 == extra[symb_table[2 * elem + 1]] + && memcmp (str, + &extra[symb_table[2 * elem + 1] + + 1], + c1) == 0) + { + /* Yep, this is the entry. */ + idx = symb_table[2 * elem + 1]; + idx += 1 + extra[idx]; + break; + } + + /* Next entry. */ + elem += second; + } + + if (symb_table[2 * elem] == 0) + /* This is no valid character. */ + FREE_STACK_RETURN (REG_ECOLLATE); + + /* Throw away the ] at the end of the equivalence + class. */ + PATFETCH (c); + + /* Now add the multibyte character(s) we found + to the accept list. + + XXX Note that this is not entirely correct. + we would have to match multibyte sequences + but this is not possible with the current + implementation. Also, we have to match + collating symbols, which expand to more than + one file, as a whole and not allow the + individual bytes. */ + c1 = extra[idx++]; + if (c1 == 1) + range_start = extra[idx]; + while (c1-- > 0) + { + SET_LIST_BIT (extra[idx]); + ++idx; + } + } +# endif + had_char_class = false; + } + else + { + c1++; + while (c1--) + PATUNFETCH; + SET_LIST_BIT ('['); + SET_LIST_BIT ('.'); + range_start = '.'; + had_char_class = false; + } + } + else + { + had_char_class = false; + SET_LIST_BIT (c); + range_start = c; + } + } + + /* Discard any (non)matching list bytes that are all 0 at the + end of the map. Decrease the map-length byte too. */ + while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) + b[-1]--; + b += b[-1]; +#endif /* WCHAR */ + } + break; + + + case '(': + if (syntax & RE_NO_BK_PARENS) + goto handle_open; + else + goto normal_char; + + + case ')': + if (syntax & RE_NO_BK_PARENS) + goto handle_close; + else + goto normal_char; + + + case '\n': + if (syntax & RE_NEWLINE_ALT) + goto handle_alt; + else + goto normal_char; + + + case '|': + if (syntax & RE_NO_BK_VBAR) + goto handle_alt; + else + goto normal_char; + + + case '{': + if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) + goto handle_interval; + else + goto normal_char; + + + case '\\': + if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); + + /* Do not translate the character after the \, so that we can + distinguish, e.g., \B from \b, even if we normally would + translate, e.g., B to b. */ + PATFETCH_RAW (c); + + switch (c) + { + case '(': + if (syntax & RE_NO_BK_PARENS) + goto normal_backslash; + + handle_open: + bufp->re_nsub++; + regnum++; + + if (COMPILE_STACK_FULL) + { + RETALLOC (compile_stack.stack, compile_stack.size << 1, + compile_stack_elt_t); + if (compile_stack.stack == NULL) return REG_ESPACE; + + compile_stack.size <<= 1; + } + + /* These are the values to restore when we hit end of this + group. They are all relative offsets, so that if the + whole pattern moves because of realloc, they will still + be valid. */ + COMPILE_STACK_TOP.begalt_offset = begalt - COMPILED_BUFFER_VAR; + COMPILE_STACK_TOP.fixup_alt_jump + = fixup_alt_jump ? fixup_alt_jump - COMPILED_BUFFER_VAR + 1 : 0; + COMPILE_STACK_TOP.laststart_offset = b - COMPILED_BUFFER_VAR; + COMPILE_STACK_TOP.regnum = regnum; + + /* We will eventually replace the 0 with the number of + groups inner to this one. But do not push a + start_memory for groups beyond the last one we can + represent in the compiled pattern. */ + if (regnum <= MAX_REGNUM) + { + COMPILE_STACK_TOP.inner_group_offset = b + - COMPILED_BUFFER_VAR + 2; + BUF_PUSH_3 (start_memory, regnum, 0); + } + + compile_stack.avail++; + + fixup_alt_jump = 0; + laststart = 0; + begalt = b; + /* If we've reached MAX_REGNUM groups, then this open + won't actually generate any code, so we'll have to + clear pending_exact explicitly. */ + pending_exact = 0; + break; + + + case ')': + if (syntax & RE_NO_BK_PARENS) goto normal_backslash; + + if (COMPILE_STACK_EMPTY) + { + if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) + goto normal_backslash; + else + FREE_STACK_RETURN (REG_ERPAREN); + } + + handle_close: + if (fixup_alt_jump) + { /* Push a dummy failure point at the end of the + alternative for a possible future + `pop_failure_jump' to pop. See comments at + `push_dummy_failure' in `re_match_2'. */ + BUF_PUSH (push_dummy_failure); + + /* We allocated space for this jump when we assigned + to `fixup_alt_jump', in the `handle_alt' case below. */ + STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1); + } + + /* See similar code for backslashed left paren above. */ + if (COMPILE_STACK_EMPTY) + { + if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) + goto normal_char; + else + FREE_STACK_RETURN (REG_ERPAREN); + } + + /* Since we just checked for an empty stack above, this + ``can't happen''. */ + assert (compile_stack.avail != 0); + { + /* We don't just want to restore into `regnum', because + later groups should continue to be numbered higher, + as in `(ab)c(de)' -- the second group is #2. */ + regnum_t this_group_regnum; + + compile_stack.avail--; + begalt = COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.begalt_offset; + fixup_alt_jump + = COMPILE_STACK_TOP.fixup_alt_jump + ? COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.fixup_alt_jump - 1 + : 0; + laststart = COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.laststart_offset; + this_group_regnum = COMPILE_STACK_TOP.regnum; + /* If we've reached MAX_REGNUM groups, then this open + won't actually generate any code, so we'll have to + clear pending_exact explicitly. */ + pending_exact = 0; + + /* We're at the end of the group, so now we know how many + groups were inside this one. */ + if (this_group_regnum <= MAX_REGNUM) + { + UCHAR_T *inner_group_loc + = COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.inner_group_offset; + + *inner_group_loc = regnum - this_group_regnum; + BUF_PUSH_3 (stop_memory, this_group_regnum, + regnum - this_group_regnum); + } + } + break; + + + case '|': /* `\|'. */ + if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) + goto normal_backslash; + handle_alt: + if (syntax & RE_LIMITED_OPS) + goto normal_char; + + /* Insert before the previous alternative a jump which + jumps to this alternative if the former fails. */ + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + INSERT_JUMP (on_failure_jump, begalt, + b + 2 + 2 * OFFSET_ADDRESS_SIZE); + pending_exact = 0; + b += 1 + OFFSET_ADDRESS_SIZE; + + /* The alternative before this one has a jump after it + which gets executed if it gets matched. Adjust that + jump so it will jump to this alternative's analogous + jump (put in below, which in turn will jump to the next + (if any) alternative's such jump, etc.). The last such + jump jumps to the correct final destination. A picture: + _____ _____ + | | | | + | v | v + a | b | c + + If we are at `b', then fixup_alt_jump right now points to a + three-byte space after `a'. We'll put in the jump, set + fixup_alt_jump to right after `b', and leave behind three + bytes which we'll fill in when we get to after `c'. */ + + if (fixup_alt_jump) + STORE_JUMP (jump_past_alt, fixup_alt_jump, b); + + /* Mark and leave space for a jump after this alternative, + to be filled in later either by next alternative or + when know we're at the end of a series of alternatives. */ + fixup_alt_jump = b; + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + b += 1 + OFFSET_ADDRESS_SIZE; + + laststart = 0; + begalt = b; + break; + + + case '{': + /* If \{ is a literal. */ + if (!(syntax & RE_INTERVALS) + /* If we're at `\{' and it's not the open-interval + operator. */ + || (syntax & RE_NO_BK_BRACES)) + goto normal_backslash; + + handle_interval: + { + /* If got here, then the syntax allows intervals. */ + + /* At least (most) this many matches must be made. */ + int lower_bound = -1, upper_bound = -1; + + /* Place in the uncompiled pattern (i.e., just after + the '{') to go back to if the interval is invalid. */ + const CHAR_T *beg_interval = p; + + if (p == pend) + goto invalid_interval; + + GET_UNSIGNED_NUMBER (lower_bound); + + if (c == ',') + { + GET_UNSIGNED_NUMBER (upper_bound); + if (upper_bound < 0) + upper_bound = RE_DUP_MAX; + } + else + /* Interval such as `{1}' => match exactly once. */ + upper_bound = lower_bound; + + if (! (0 <= lower_bound && lower_bound <= upper_bound)) + goto invalid_interval; + + if (!(syntax & RE_NO_BK_BRACES)) + { + if (c != '\\' || p == pend) + goto invalid_interval; + PATFETCH (c); + } + + if (c != '}') + goto invalid_interval; + + /* If it's invalid to have no preceding re. */ + if (!laststart) + { + if (syntax & RE_CONTEXT_INVALID_OPS + && !(syntax & RE_INVALID_INTERVAL_ORD)) + FREE_STACK_RETURN (REG_BADRPT); + else if (syntax & RE_CONTEXT_INDEP_OPS) + laststart = b; + else + goto unfetch_interval; + } + + /* We just parsed a valid interval. */ + + if (RE_DUP_MAX < upper_bound) + FREE_STACK_RETURN (REG_BADBR); + + /* If the upper bound is zero, don't want to succeed at + all; jump from `laststart' to `b + 3', which will be + the end of the buffer after we insert the jump. */ + /* ifdef WCHAR, 'b + 1 + OFFSET_ADDRESS_SIZE' + instead of 'b + 3'. */ + if (upper_bound == 0) + { + GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); + INSERT_JUMP (jump, laststart, b + 1 + + OFFSET_ADDRESS_SIZE); + b += 1 + OFFSET_ADDRESS_SIZE; + } + + /* Otherwise, we have a nontrivial interval. When + we're all done, the pattern will look like: + set_number_at + set_number_at + succeed_n + + jump_n + (The upper bound and `jump_n' are omitted if + `upper_bound' is 1, though.) */ + else + { /* If the upper bound is > 1, we need to insert + more at the end of the loop. */ + unsigned nbytes = 2 + 4 * OFFSET_ADDRESS_SIZE + + (upper_bound > 1) * (2 + 4 * OFFSET_ADDRESS_SIZE); + + GET_BUFFER_SPACE (nbytes); + + /* Initialize lower bound of the `succeed_n', even + though it will be set during matching by its + attendant `set_number_at' (inserted next), + because `re_compile_fastmap' needs to know. + Jump to the `jump_n' we might insert below. */ + INSERT_JUMP2 (succeed_n, laststart, + b + 1 + 2 * OFFSET_ADDRESS_SIZE + + (upper_bound > 1) * (1 + 2 * OFFSET_ADDRESS_SIZE) + , lower_bound); + b += 1 + 2 * OFFSET_ADDRESS_SIZE; + + /* Code to initialize the lower bound. Insert + before the `succeed_n'. The `5' is the last two + bytes of this `set_number_at', plus 3 bytes of + the following `succeed_n'. */ + /* ifdef WCHAR, The '1+2*OFFSET_ADDRESS_SIZE' + is the 'set_number_at', plus '1+OFFSET_ADDRESS_SIZE' + of the following `succeed_n'. */ + PREFIX(insert_op2) (set_number_at, laststart, 1 + + 2 * OFFSET_ADDRESS_SIZE, lower_bound, b); + b += 1 + 2 * OFFSET_ADDRESS_SIZE; + + if (upper_bound > 1) + { /* More than one repetition is allowed, so + append a backward jump to the `succeed_n' + that starts this interval. + + When we've reached this during matching, + we'll have matched the interval once, so + jump back only `upper_bound - 1' times. */ + STORE_JUMP2 (jump_n, b, laststart + + 2 * OFFSET_ADDRESS_SIZE + 1, + upper_bound - 1); + b += 1 + 2 * OFFSET_ADDRESS_SIZE; + + /* The location we want to set is the second + parameter of the `jump_n'; that is `b-2' as + an absolute address. `laststart' will be + the `set_number_at' we're about to insert; + `laststart+3' the number to set, the source + for the relative address. But we are + inserting into the middle of the pattern -- + so everything is getting moved up by 5. + Conclusion: (b - 2) - (laststart + 3) + 5, + i.e., b - laststart. + + We insert this at the beginning of the loop + so that if we fail during matching, we'll + reinitialize the bounds. */ + PREFIX(insert_op2) (set_number_at, laststart, + b - laststart, + upper_bound - 1, b); + b += 1 + 2 * OFFSET_ADDRESS_SIZE; + } + } + pending_exact = 0; + break; + + invalid_interval: + if (!(syntax & RE_INVALID_INTERVAL_ORD)) + FREE_STACK_RETURN (p == pend ? REG_EBRACE : REG_BADBR); + unfetch_interval: + /* Match the characters as literals. */ + p = beg_interval; + c = '{'; + if (syntax & RE_NO_BK_BRACES) + goto normal_char; + else + goto normal_backslash; + } + +#ifdef emacs + /* There is no way to specify the before_dot and after_dot + operators. rms says this is ok. --karl */ + case '=': + BUF_PUSH (at_dot); + break; + + case 's': + laststart = b; + PATFETCH (c); + BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); + break; + + case 'S': + laststart = b; + PATFETCH (c); + BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); + break; +#endif /* emacs */ + + + case 'w': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + laststart = b; + BUF_PUSH (wordchar); + break; + + + case 'W': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + laststart = b; + BUF_PUSH (notwordchar); + break; + + + case '<': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (wordbeg); + break; + + case '>': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (wordend); + break; + + case 'b': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (wordbound); + break; + + case 'B': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (notwordbound); + break; + + case '`': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (begbuf); + break; + + case '\'': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH (endbuf); + break; + + case '1': case '2': case '3': case '4': case '5': + case '6': case '7': case '8': case '9': + if (syntax & RE_NO_BK_REFS) + goto normal_char; + + c1 = c - '0'; + + if (c1 > regnum) + FREE_STACK_RETURN (REG_ESUBREG); + + /* Can't back reference to a subexpression if inside of it. */ + if (group_in_compile_stack (compile_stack, (regnum_t) c1)) + goto normal_char; + + laststart = b; + BUF_PUSH_2 (duplicate, c1); + break; + + + case '+': + case '?': + if (syntax & RE_BK_PLUS_QM) + goto handle_plus; + else + goto normal_backslash; + + default: + normal_backslash: + /* You might think it would be useful for \ to mean + not to translate; but if we don't translate it + it will never match anything. */ + c = TRANSLATE (c); + goto normal_char; + } + break; + + + default: + /* Expects the character in `c'. */ + normal_char: + /* If no exactn currently being built. */ + if (!pending_exact +#ifdef WCHAR + /* If last exactn handle binary(or character) and + new exactn handle character(or binary). */ + || is_exactn_bin != is_binary[p - 1 - pattern] +#endif /* WCHAR */ + + /* If last exactn not at current position. */ + || pending_exact + *pending_exact + 1 != b + + /* We have only one byte following the exactn for the count. */ + || *pending_exact == (1 << BYTEWIDTH) - 1 + + /* If followed by a repetition operator. */ + || *p == '*' || *p == '^' + || ((syntax & RE_BK_PLUS_QM) + ? *p == '\\' && (p[1] == '+' || p[1] == '?') + : (*p == '+' || *p == '?')) + || ((syntax & RE_INTERVALS) + && ((syntax & RE_NO_BK_BRACES) + ? *p == '{' + : (p[0] == '\\' && p[1] == '{')))) + { + /* Start building a new exactn. */ + + laststart = b; + +#ifdef WCHAR + /* Is this exactn binary data or character? */ + is_exactn_bin = is_binary[p - 1 - pattern]; + if (is_exactn_bin) + BUF_PUSH_2 (exactn_bin, 0); + else + BUF_PUSH_2 (exactn, 0); +#else + BUF_PUSH_2 (exactn, 0); +#endif /* WCHAR */ + pending_exact = b - 1; + } + + BUF_PUSH (c); + (*pending_exact)++; + break; + } /* switch (c) */ + } /* while p != pend */ + + + /* Through the pattern now. */ + + if (fixup_alt_jump) + STORE_JUMP (jump_past_alt, fixup_alt_jump, b); + + if (!COMPILE_STACK_EMPTY) + FREE_STACK_RETURN (REG_EPAREN); + + /* If we don't want backtracking, force success + the first time we reach the end of the compiled pattern. */ + if (syntax & RE_NO_POSIX_BACKTRACKING) + BUF_PUSH (succeed); + +#ifdef WCHAR + free (pattern); + free (mbs_offset); + free (is_binary); +#endif + free (compile_stack.stack); + + /* We have succeeded; set the length of the buffer. */ +#ifdef WCHAR + bufp->used = (uintptr_t) b - (uintptr_t) COMPILED_BUFFER_VAR; +#else + bufp->used = b - bufp->buffer; +#endif + +#ifdef DEBUG + if (debug) + { + DEBUG_PRINT1 ("\nCompiled pattern: \n"); + PREFIX(print_compiled_pattern) (bufp); + } +#endif /* DEBUG */ + +#ifndef MATCH_MAY_ALLOCATE + /* Initialize the failure stack to the largest possible stack. This + isn't necessary unless we're trying to avoid calling alloca in + the search and match routines. */ + { + int num_regs = bufp->re_nsub + 1; + + /* Since DOUBLE_FAIL_STACK refuses to double only if the current size + is strictly greater than re_max_failures, the largest possible stack + is 2 * re_max_failures failure points. */ + if (fail_stack.size < (2 * re_max_failures * MAX_FAILURE_ITEMS)) + { + fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS); + +# ifdef emacs + if (! fail_stack.stack) + fail_stack.stack + = (PREFIX(fail_stack_elt_t) *) xmalloc (fail_stack.size + * sizeof (PREFIX(fail_stack_elt_t))); + else + fail_stack.stack + = (PREFIX(fail_stack_elt_t) *) xrealloc (fail_stack.stack, + (fail_stack.size + * sizeof (PREFIX(fail_stack_elt_t)))); +# else /* not emacs */ + if (! fail_stack.stack) + fail_stack.stack + = (PREFIX(fail_stack_elt_t) *) malloc (fail_stack.size + * sizeof (PREFIX(fail_stack_elt_t))); + else + fail_stack.stack + = (PREFIX(fail_stack_elt_t) *) realloc (fail_stack.stack, + (fail_stack.size + * sizeof (PREFIX(fail_stack_elt_t)))); +# endif /* not emacs */ + } + + PREFIX(regex_grow_registers) (num_regs); + } +#endif /* not MATCH_MAY_ALLOCATE */ + + return REG_NOERROR; +} /* regex_compile */ + +/* Subroutines for `regex_compile'. */ + +/* Store OP at LOC followed by two-byte integer parameter ARG. */ +/* ifdef WCHAR, integer parameter is 1 wchar_t. */ + +static void +PREFIX(store_op1) (op, loc, arg) + re_opcode_t op; + UCHAR_T *loc; + int arg; +{ + *loc = (UCHAR_T) op; + STORE_NUMBER (loc + 1, arg); +} + + +/* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ +/* ifdef WCHAR, integer parameter is 1 wchar_t. */ + +static void +PREFIX(store_op2) (op, loc, arg1, arg2) + re_opcode_t op; + UCHAR_T *loc; + int arg1, arg2; +{ + *loc = (UCHAR_T) op; + STORE_NUMBER (loc + 1, arg1); + STORE_NUMBER (loc + 1 + OFFSET_ADDRESS_SIZE, arg2); +} + + +/* Copy the bytes from LOC to END to open up three bytes of space at LOC + for OP followed by two-byte integer parameter ARG. */ +/* ifdef WCHAR, integer parameter is 1 wchar_t. */ + +static void +PREFIX(insert_op1) (op, loc, arg, end) + re_opcode_t op; + UCHAR_T *loc; + int arg; + UCHAR_T *end; +{ + register UCHAR_T *pfrom = end; + register UCHAR_T *pto = end + 1 + OFFSET_ADDRESS_SIZE; + + while (pfrom != loc) + *--pto = *--pfrom; + + PREFIX(store_op1) (op, loc, arg); +} + + +/* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ +/* ifdef WCHAR, integer parameter is 1 wchar_t. */ + +static void +PREFIX(insert_op2) (op, loc, arg1, arg2, end) + re_opcode_t op; + UCHAR_T *loc; + int arg1, arg2; + UCHAR_T *end; +{ + register UCHAR_T *pfrom = end; + register UCHAR_T *pto = end + 1 + 2 * OFFSET_ADDRESS_SIZE; + + while (pfrom != loc) + *--pto = *--pfrom; + + PREFIX(store_op2) (op, loc, arg1, arg2); +} + + +/* P points to just after a ^ in PATTERN. Return true if that ^ comes + after an alternative or a begin-subexpression. We assume there is at + least one character before the ^. */ + +static boolean +PREFIX(at_begline_loc_p) (pattern, p, syntax) + const CHAR_T *pattern, *p; + reg_syntax_t syntax; +{ + const CHAR_T *prev = p - 2; + boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\'; + + return + /* After a subexpression? */ + (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash)) + /* After an alternative? */ + || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash)); +} + + +/* The dual of at_begline_loc_p. This one is for $. We assume there is + at least one character after the $, i.e., `P < PEND'. */ + +static boolean +PREFIX(at_endline_loc_p) (p, pend, syntax) + const CHAR_T *p, *pend; + reg_syntax_t syntax; +{ + const CHAR_T *next = p; + boolean next_backslash = *next == '\\'; + const CHAR_T *next_next = p + 1 < pend ? p + 1 : 0; + + return + /* Before a subexpression? */ + (syntax & RE_NO_BK_PARENS ? *next == ')' + : next_backslash && next_next && *next_next == ')') + /* Before an alternative? */ + || (syntax & RE_NO_BK_VBAR ? *next == '|' + : next_backslash && next_next && *next_next == '|'); +} + +#else /* not INSIDE_RECURSION */ + +/* Returns true if REGNUM is in one of COMPILE_STACK's elements and + false if it's not. */ + +static boolean +group_in_compile_stack (compile_stack, regnum) + compile_stack_type compile_stack; + regnum_t regnum; +{ + int this_element; + + for (this_element = compile_stack.avail - 1; + this_element >= 0; + this_element--) + if (compile_stack.stack[this_element].regnum == regnum) + return true; + + return false; +} +#endif /* not INSIDE_RECURSION */ + +#ifdef INSIDE_RECURSION + +#ifdef WCHAR +/* This insert space, which size is "num", into the pattern at "loc". + "end" must point the end of the allocated buffer. */ +static void +insert_space (num, loc, end) + int num; + CHAR_T *loc; + CHAR_T *end; +{ + register CHAR_T *pto = end; + register CHAR_T *pfrom = end - num; + + while (pfrom >= loc) + *pto-- = *pfrom--; +} +#endif /* WCHAR */ + +#ifdef WCHAR +static reg_errcode_t +wcs_compile_range (range_start_char, p_ptr, pend, translate, syntax, b, + char_set) + CHAR_T range_start_char; + const CHAR_T **p_ptr, *pend; + CHAR_T *char_set, *b; + RE_TRANSLATE_TYPE translate; + reg_syntax_t syntax; +{ + const CHAR_T *p = *p_ptr; + CHAR_T range_start, range_end; + reg_errcode_t ret; +# ifdef _LIBC + uint32_t nrules; + uint32_t start_val, end_val; +# endif + if (p == pend) + return REG_ERANGE; + +# ifdef _LIBC + nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + if (nrules != 0) + { + const char *collseq = (const char *) _NL_CURRENT(LC_COLLATE, + _NL_COLLATE_COLLSEQWC); + const unsigned char *extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); + + if (range_start_char < -1) + { + /* range_start is a collating symbol. */ + int32_t *wextra; + /* Retreive the index and get collation sequence value. */ + wextra = (int32_t*)(extra + char_set[-range_start_char]); + start_val = wextra[1 + *wextra]; + } + else + start_val = collseq_table_lookup(collseq, TRANSLATE(range_start_char)); + + end_val = collseq_table_lookup (collseq, TRANSLATE (p[0])); + + /* Report an error if the range is empty and the syntax prohibits + this. */ + ret = ((syntax & RE_NO_EMPTY_RANGES) + && (start_val > end_val))? REG_ERANGE : REG_NOERROR; + + /* Insert space to the end of the char_ranges. */ + insert_space(2, b - char_set[5] - 2, b - 1); + *(b - char_set[5] - 2) = (wchar_t)start_val; + *(b - char_set[5] - 1) = (wchar_t)end_val; + char_set[4]++; /* ranges_index */ + } + else +# endif + { + range_start = (range_start_char >= 0)? TRANSLATE (range_start_char): + range_start_char; + range_end = TRANSLATE (p[0]); + /* Report an error if the range is empty and the syntax prohibits + this. */ + ret = ((syntax & RE_NO_EMPTY_RANGES) + && (range_start > range_end))? REG_ERANGE : REG_NOERROR; + + /* Insert space to the end of the char_ranges. */ + insert_space(2, b - char_set[5] - 2, b - 1); + *(b - char_set[5] - 2) = range_start; + *(b - char_set[5] - 1) = range_end; + char_set[4]++; /* ranges_index */ + } + /* Have to increment the pointer into the pattern string, so the + caller isn't still at the ending character. */ + (*p_ptr)++; + + return ret; +} +#else /* BYTE */ +/* Read the ending character of a range (in a bracket expression) from the + uncompiled pattern *P_PTR (which ends at PEND). We assume the + starting character is in `P[-2]'. (`P[-1]' is the character `-'.) + Then we set the translation of all bits between the starting and + ending characters (inclusive) in the compiled pattern B. + + Return an error code. + + We use these short variable names so we can use the same macros as + `regex_compile' itself. */ + +static reg_errcode_t +byte_compile_range (range_start_char, p_ptr, pend, translate, syntax, b) + unsigned int range_start_char; + const char **p_ptr, *pend; + RE_TRANSLATE_TYPE translate; + reg_syntax_t syntax; + unsigned char *b; +{ + unsigned this_char; + const char *p = *p_ptr; + reg_errcode_t ret; +# if _LIBC + const unsigned char *collseq; + unsigned int start_colseq; + unsigned int end_colseq; +# else + unsigned end_char; +# endif + + if (p == pend) + return REG_ERANGE; + + /* Have to increment the pointer into the pattern string, so the + caller isn't still at the ending character. */ + (*p_ptr)++; + + /* Report an error if the range is empty and the syntax prohibits this. */ + ret = syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; + +# if _LIBC + collseq = (const unsigned char *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_COLLSEQMB); + + start_colseq = collseq[(unsigned char) TRANSLATE (range_start_char)]; + end_colseq = collseq[(unsigned char) TRANSLATE (p[0])]; + for (this_char = 0; this_char <= (unsigned char) -1; ++this_char) + { + unsigned int this_colseq = collseq[(unsigned char) TRANSLATE (this_char)]; + + if (start_colseq <= this_colseq && this_colseq <= end_colseq) + { + SET_LIST_BIT (TRANSLATE (this_char)); + ret = REG_NOERROR; + } + } +# else + /* Here we see why `this_char' has to be larger than an `unsigned + char' -- we would otherwise go into an infinite loop, since all + characters <= 0xff. */ + range_start_char = TRANSLATE (range_start_char); + /* TRANSLATE(p[0]) is casted to char (not unsigned char) in TRANSLATE, + and some compilers cast it to int implicitly, so following for_loop + may fall to (almost) infinite loop. + e.g. If translate[p[0]] = 0xff, end_char may equals to 0xffffffff. + To avoid this, we cast p[0] to unsigned int and truncate it. */ + end_char = ((unsigned)TRANSLATE(p[0]) & ((1 << BYTEWIDTH) - 1)); + + for (this_char = range_start_char; this_char <= end_char; ++this_char) + { + SET_LIST_BIT (TRANSLATE (this_char)); + ret = REG_NOERROR; + } +# endif + + return ret; +} +#endif /* WCHAR */ + +/* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in + BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible + characters can start a string that matches the pattern. This fastmap + is used by re_search to skip quickly over impossible starting points. + + The caller must supply the address of a (1 << BYTEWIDTH)-byte data + area as BUFP->fastmap. + + We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in + the pattern buffer. + + Returns 0 if we succeed, -2 if an internal error. */ + +#ifdef WCHAR +/* local function for re_compile_fastmap. + truncate wchar_t character to char. */ +static unsigned char truncate_wchar (CHAR_T c); + +static unsigned char +truncate_wchar (c) + CHAR_T c; +{ + unsigned char buf[MB_LEN_MAX]; + int retval = wctomb(buf, c); + return retval > 0 ? buf[0] : (unsigned char)c; +} +#endif /* WCHAR */ + +static int +PREFIX(re_compile_fastmap) (bufp) + struct re_pattern_buffer *bufp; +{ + int j, k; +#ifdef MATCH_MAY_ALLOCATE + PREFIX(fail_stack_type) fail_stack; +#endif +#ifndef REGEX_MALLOC + char *destination; +#endif + + register char *fastmap = bufp->fastmap; + +#ifdef WCHAR + /* We need to cast pattern to (wchar_t*), because we casted this compiled + pattern to (char*) in regex_compile. */ + UCHAR_T *pattern = (UCHAR_T*)bufp->buffer; + register UCHAR_T *pend = (UCHAR_T*) (bufp->buffer + bufp->used); +#else /* BYTE */ + UCHAR_T *pattern = bufp->buffer; + register UCHAR_T *pend = pattern + bufp->used; +#endif /* WCHAR */ + UCHAR_T *p = pattern; + +#ifdef REL_ALLOC + /* This holds the pointer to the failure stack, when + it is allocated relocatably. */ + fail_stack_elt_t *failure_stack_ptr; +#endif + + /* Assume that each path through the pattern can be null until + proven otherwise. We set this false at the bottom of switch + statement, to which we get only if a particular path doesn't + match the empty string. */ + boolean path_can_be_null = true; + + /* We aren't doing a `succeed_n' to begin with. */ + boolean succeed_n_p = false; + + assert (fastmap != NULL && p != NULL); + + INIT_FAIL_STACK (); + bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */ + bufp->fastmap_accurate = 1; /* It will be when we're done. */ + bufp->can_be_null = 0; + + while (1) + { + if (p == pend || *p == succeed) + { + /* We have reached the (effective) end of pattern. */ + if (!FAIL_STACK_EMPTY ()) + { + bufp->can_be_null |= path_can_be_null; + + /* Reset for next path. */ + path_can_be_null = true; + + p = fail_stack.stack[--fail_stack.avail].pointer; + + continue; + } + else + break; + } + + /* We should never be about to go beyond the end of the pattern. */ + assert (p < pend); + + switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) + { + + /* I guess the idea here is to simply not bother with a fastmap + if a backreference is used, since it's too hard to figure out + the fastmap for the corresponding group. Setting + `can_be_null' stops `re_search_2' from using the fastmap, so + that is all we do. */ + case duplicate: + bufp->can_be_null = 1; + goto done; + + + /* Following are the cases which match a character. These end + with `break'. */ + +#ifdef WCHAR + case exactn: + fastmap[truncate_wchar(p[1])] = 1; + break; +#else /* BYTE */ + case exactn: + fastmap[p[1]] = 1; + break; +#endif /* WCHAR */ +#ifdef MBS_SUPPORT + case exactn_bin: + fastmap[p[1]] = 1; + break; +#endif + +#ifdef WCHAR + /* It is hard to distinguish fastmap from (multi byte) characters + which depends on current locale. */ + case charset: + case charset_not: + case wordchar: + case notwordchar: + bufp->can_be_null = 1; + goto done; +#else /* BYTE */ + case charset: + for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) + if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) + fastmap[j] = 1; + break; + + + case charset_not: + /* Chars beyond end of map must be allowed. */ + for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++) + fastmap[j] = 1; + + for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) + if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))) + fastmap[j] = 1; + break; + + + case wordchar: + for (j = 0; j < (1 << BYTEWIDTH); j++) + if (SYNTAX (j) == Sword) + fastmap[j] = 1; + break; + + + case notwordchar: + for (j = 0; j < (1 << BYTEWIDTH); j++) + if (SYNTAX (j) != Sword) + fastmap[j] = 1; + break; +#endif /* WCHAR */ + + case anychar: + { + int fastmap_newline = fastmap['\n']; + + /* `.' matches anything ... */ + for (j = 0; j < (1 << BYTEWIDTH); j++) + fastmap[j] = 1; + + /* ... except perhaps newline. */ + if (!(bufp->syntax & RE_DOT_NEWLINE)) + fastmap['\n'] = fastmap_newline; + + /* Return if we have already set `can_be_null'; if we have, + then the fastmap is irrelevant. Something's wrong here. */ + else if (bufp->can_be_null) + goto done; + + /* Otherwise, have to check alternative paths. */ + break; + } + +#ifdef emacs + case syntaxspec: + k = *p++; + for (j = 0; j < (1 << BYTEWIDTH); j++) + if (SYNTAX (j) == (enum syntaxcode) k) + fastmap[j] = 1; + break; + + + case notsyntaxspec: + k = *p++; + for (j = 0; j < (1 << BYTEWIDTH); j++) + if (SYNTAX (j) != (enum syntaxcode) k) + fastmap[j] = 1; + break; + + + /* All cases after this match the empty string. These end with + `continue'. */ + + + case before_dot: + case at_dot: + case after_dot: + continue; +#endif /* emacs */ + + + case no_op: + case begline: + case endline: + case begbuf: + case endbuf: + case wordbound: + case notwordbound: + case wordbeg: + case wordend: + case push_dummy_failure: + continue; + + + case jump_n: + case pop_failure_jump: + case maybe_pop_jump: + case jump: + case jump_past_alt: + case dummy_failure_jump: + EXTRACT_NUMBER_AND_INCR (j, p); + p += j; + if (j > 0) + continue; + + /* Jump backward implies we just went through the body of a + loop and matched nothing. Opcode jumped to should be + `on_failure_jump' or `succeed_n'. Just treat it like an + ordinary jump. For a * loop, it has pushed its failure + point already; if so, discard that as redundant. */ + if ((re_opcode_t) *p != on_failure_jump + && (re_opcode_t) *p != succeed_n) + continue; + + p++; + EXTRACT_NUMBER_AND_INCR (j, p); + p += j; + + /* If what's on the stack is where we are now, pop it. */ + if (!FAIL_STACK_EMPTY () + && fail_stack.stack[fail_stack.avail - 1].pointer == p) + fail_stack.avail--; + + continue; + + + case on_failure_jump: + case on_failure_keep_string_jump: + handle_on_failure_jump: + EXTRACT_NUMBER_AND_INCR (j, p); + + /* For some patterns, e.g., `(a?)?', `p+j' here points to the + end of the pattern. We don't want to push such a point, + since when we restore it above, entering the switch will + increment `p' past the end of the pattern. We don't need + to push such a point since we obviously won't find any more + fastmap entries beyond `pend'. Such a pattern can match + the null string, though. */ + if (p + j < pend) + { + if (!PUSH_PATTERN_OP (p + j, fail_stack)) + { + RESET_FAIL_STACK (); + return -2; + } + } + else + bufp->can_be_null = 1; + + if (succeed_n_p) + { + EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */ + succeed_n_p = false; + } + + continue; + + + case succeed_n: + /* Get to the number of times to succeed. */ + p += OFFSET_ADDRESS_SIZE; + + /* Increment p past the n for when k != 0. */ + EXTRACT_NUMBER_AND_INCR (k, p); + if (k == 0) + { + p -= 2 * OFFSET_ADDRESS_SIZE; + succeed_n_p = true; /* Spaghetti code alert. */ + goto handle_on_failure_jump; + } + continue; + + + case set_number_at: + p += 2 * OFFSET_ADDRESS_SIZE; + continue; + + + case start_memory: + case stop_memory: + p += 2; + continue; + + + default: + abort (); /* We have listed all the cases. */ + } /* switch *p++ */ + + /* Getting here means we have found the possible starting + characters for one path of the pattern -- and that the empty + string does not match. We need not follow this path further. + Instead, look at the next alternative (remembered on the + stack), or quit if no more. The test at the top of the loop + does these things. */ + path_can_be_null = false; + p = pend; + } /* while p */ + + /* Set `can_be_null' for the last path (also the first path, if the + pattern is empty). */ + bufp->can_be_null |= path_can_be_null; + + done: + RESET_FAIL_STACK (); + return 0; +} + +#else /* not INSIDE_RECURSION */ + +int +re_compile_fastmap (bufp) + struct re_pattern_buffer *bufp; +{ +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + return wcs_re_compile_fastmap(bufp); + else +# endif + return byte_re_compile_fastmap(bufp); +} /* re_compile_fastmap */ +#ifdef _LIBC +weak_alias (__re_compile_fastmap, re_compile_fastmap) +#endif + + +/* Set REGS to hold NUM_REGS registers, storing them in STARTS and + ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use + this memory for recording register information. STARTS and ENDS + must be allocated using the malloc library routine, and must each + be at least NUM_REGS * sizeof (regoff_t) bytes long. + + If NUM_REGS == 0, then subsequent matches should allocate their own + register data. + + Unless this function is called, the first search or match using + PATTERN_BUFFER will allocate its own register data, without + freeing the old data. */ + +void +re_set_registers (bufp, regs, num_regs, starts, ends) + struct re_pattern_buffer *bufp; + struct re_registers *regs; + unsigned num_regs; + regoff_t *starts, *ends; +{ + if (num_regs) + { + bufp->regs_allocated = REGS_REALLOCATE; + regs->num_regs = num_regs; + regs->start = starts; + regs->end = ends; + } + else + { + bufp->regs_allocated = REGS_UNALLOCATED; + regs->num_regs = 0; + regs->start = regs->end = (regoff_t *) 0; + } +} +#ifdef _LIBC +weak_alias (__re_set_registers, re_set_registers) +#endif + +/* Searching routines. */ + +/* Like re_search_2, below, but only one string is specified, and + doesn't let you say where to stop matching. */ + +int +re_search (bufp, string, size, startpos, range, regs) + struct re_pattern_buffer *bufp; + const char *string; + int size, startpos, range; + struct re_registers *regs; +{ + return re_search_2 (bufp, NULL, 0, string, size, startpos, range, + regs, size); +} +#ifdef _LIBC +weak_alias (__re_search, re_search) +#endif + + +/* Using the compiled pattern in BUFP->buffer, first tries to match the + virtual concatenation of STRING1 and STRING2, starting first at index + STARTPOS, then at STARTPOS + 1, and so on. + + STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. + + RANGE is how far to scan while trying to match. RANGE = 0 means try + only at STARTPOS; in general, the last start tried is STARTPOS + + RANGE. + + In REGS, return the indices of the virtual concatenation of STRING1 + and STRING2 that matched the entire BUFP->buffer and its contained + subexpressions. + + Do not consider matching one past the index STOP in the virtual + concatenation of STRING1 and STRING2. + + We return either the position in the strings at which the match was + found, -1 if no match, or -2 if error (such as failure + stack overflow). */ + +int +re_search_2 (bufp, string1, size1, string2, size2, startpos, range, regs, stop) + struct re_pattern_buffer *bufp; + const char *string1, *string2; + int size1, size2; + int startpos; + int range; + struct re_registers *regs; + int stop; +{ +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + return wcs_re_search_2 (bufp, string1, size1, string2, size2, startpos, + range, regs, stop); + else +# endif + return byte_re_search_2 (bufp, string1, size1, string2, size2, startpos, + range, regs, stop); +} /* re_search_2 */ +#ifdef _LIBC +weak_alias (__re_search_2, re_search_2) +#endif + +#endif /* not INSIDE_RECURSION */ + +#ifdef INSIDE_RECURSION + +#ifdef MATCH_MAY_ALLOCATE +# define FREE_VAR(var) if (var) REGEX_FREE (var); var = NULL +#else +# define FREE_VAR(var) if (var) free (var); var = NULL +#endif + +#ifdef WCHAR +# define FREE_WCS_BUFFERS() \ + do { \ + FREE_VAR (string1); \ + FREE_VAR (string2); \ + FREE_VAR (mbs_offset1); \ + FREE_VAR (mbs_offset2); \ + } while (0) + +#endif + +static int +PREFIX(re_search_2) (bufp, string1, size1, string2, size2, startpos, range, + regs, stop) + struct re_pattern_buffer *bufp; + const char *string1, *string2; + int size1, size2; + int startpos; + int range; + struct re_registers *regs; + int stop; +{ + int val; + register char *fastmap = bufp->fastmap; + register RE_TRANSLATE_TYPE translate = bufp->translate; + int total_size = size1 + size2; + int endpos = startpos + range; +#ifdef WCHAR + /* We need wchar_t* buffers correspond to cstring1, cstring2. */ + wchar_t *wcs_string1 = NULL, *wcs_string2 = NULL; + /* We need the size of wchar_t buffers correspond to csize1, csize2. */ + int wcs_size1 = 0, wcs_size2 = 0; + /* offset buffer for optimizatoin. See convert_mbs_to_wc. */ + int *mbs_offset1 = NULL, *mbs_offset2 = NULL; + /* They hold whether each wchar_t is binary data or not. */ + char *is_binary = NULL; +#endif /* WCHAR */ + + /* Check for out-of-range STARTPOS. */ + if (startpos < 0 || startpos > total_size) + return -1; + + /* Fix up RANGE if it might eventually take us outside + the virtual concatenation of STRING1 and STRING2. + Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ + if (endpos < 0) + range = 0 - startpos; + else if (endpos > total_size) + range = total_size - startpos; + + /* If the search isn't to be a backwards one, don't waste time in a + search for a pattern that must be anchored. */ + if (bufp->used > 0 && range > 0 + && ((re_opcode_t) bufp->buffer[0] == begbuf + /* `begline' is like `begbuf' if it cannot match at newlines. */ + || ((re_opcode_t) bufp->buffer[0] == begline + && !bufp->newline_anchor))) + { + if (startpos > 0) + return -1; + else + range = 1; + } + +#ifdef emacs + /* In a forward search for something that starts with \=. + don't keep searching past point. */ + if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) + { + range = PT - startpos; + if (range <= 0) + return -1; + } +#endif /* emacs */ + + /* Update the fastmap now if not correct already. */ + if (fastmap && !bufp->fastmap_accurate) + if (re_compile_fastmap (bufp) == -2) + return -2; + +#ifdef WCHAR + /* Allocate wchar_t array for wcs_string1 and wcs_string2 and + fill them with converted string. */ + if (size1 != 0) + { + wcs_string1 = REGEX_TALLOC (size1 + 1, CHAR_T); + mbs_offset1 = REGEX_TALLOC (size1 + 1, int); + is_binary = REGEX_TALLOC (size1 + 1, char); + if (!wcs_string1 || !mbs_offset1 || !is_binary) + { + FREE_VAR (wcs_string1); + FREE_VAR (mbs_offset1); + FREE_VAR (is_binary); + return -2; + } + wcs_size1 = convert_mbs_to_wcs(wcs_string1, string1, size1, + mbs_offset1, is_binary); + wcs_string1[wcs_size1] = L'\0'; /* for a sentinel */ + FREE_VAR (is_binary); + } + if (size2 != 0) + { + wcs_string2 = REGEX_TALLOC (size2 + 1, CHAR_T); + mbs_offset2 = REGEX_TALLOC (size2 + 1, int); + is_binary = REGEX_TALLOC (size2 + 1, char); + if (!wcs_string2 || !mbs_offset2 || !is_binary) + { + FREE_WCS_BUFFERS (); + FREE_VAR (is_binary); + return -2; + } + wcs_size2 = convert_mbs_to_wcs(wcs_string2, string2, size2, + mbs_offset2, is_binary); + wcs_string2[wcs_size2] = L'\0'; /* for a sentinel */ + FREE_VAR (is_binary); + } +#endif /* WCHAR */ + + + /* Loop through the string, looking for a place to start matching. */ + for (;;) + { + /* If a fastmap is supplied, skip quickly over characters that + cannot be the start of a match. If the pattern can match the + null string, however, we don't need to skip characters; we want + the first null string. */ + if (fastmap && startpos < total_size && !bufp->can_be_null) + { + if (range > 0) /* Searching forwards. */ + { + register const char *d; + register int lim = 0; + int irange = range; + + if (startpos < size1 && startpos + range >= size1) + lim = range - (size1 - startpos); + + d = (startpos >= size1 ? string2 - size1 : string1) + startpos; + + /* Written out as an if-else to avoid testing `translate' + inside the loop. */ + if (translate) + while (range > lim + && !fastmap[(unsigned char) + translate[(unsigned char) *d++]]) + range--; + else + while (range > lim && !fastmap[(unsigned char) *d++]) + range--; + + startpos += irange - range; + } + else /* Searching backwards. */ + { + register CHAR_T c = (size1 == 0 || startpos >= size1 + ? string2[startpos - size1] + : string1[startpos]); + + if (!fastmap[(unsigned char) TRANSLATE (c)]) + goto advance; + } + } + + /* If can't match the null string, and that's all we have left, fail. */ + if (range >= 0 && startpos == total_size && fastmap + && !bufp->can_be_null) + { +#ifdef WCHAR + FREE_WCS_BUFFERS (); +#endif + return -1; + } + +#ifdef WCHAR + val = wcs_re_match_2_internal (bufp, string1, size1, string2, + size2, startpos, regs, stop, + wcs_string1, wcs_size1, + wcs_string2, wcs_size2, + mbs_offset1, mbs_offset2); +#else /* BYTE */ + val = byte_re_match_2_internal (bufp, string1, size1, string2, + size2, startpos, regs, stop); +#endif /* BYTE */ + +#ifndef REGEX_MALLOC +# ifdef C_ALLOCA + alloca (0); +# endif +#endif + + if (val >= 0) + { +#ifdef WCHAR + FREE_WCS_BUFFERS (); +#endif + return startpos; + } + + if (val == -2) + { +#ifdef WCHAR + FREE_WCS_BUFFERS (); +#endif + return -2; + } + + advance: + if (!range) + break; + else if (range > 0) + { + range--; + startpos++; + } + else + { + range++; + startpos--; + } + } +#ifdef WCHAR + FREE_WCS_BUFFERS (); +#endif + return -1; +} + +#ifdef WCHAR +/* This converts PTR, a pointer into one of the search wchar_t strings + `string1' and `string2' into an multibyte string offset from the + beginning of that string. We use mbs_offset to optimize. + See convert_mbs_to_wcs. */ +# define POINTER_TO_OFFSET(ptr) \ + (FIRST_STRING_P (ptr) \ + ? ((regoff_t)(mbs_offset1 != NULL? mbs_offset1[(ptr)-string1] : 0)) \ + : ((regoff_t)((mbs_offset2 != NULL? mbs_offset2[(ptr)-string2] : 0) \ + + csize1))) +#else /* BYTE */ +/* This converts PTR, a pointer into one of the search strings `string1' + and `string2' into an offset from the beginning of that string. */ +# define POINTER_TO_OFFSET(ptr) \ + (FIRST_STRING_P (ptr) \ + ? ((regoff_t) ((ptr) - string1)) \ + : ((regoff_t) ((ptr) - string2 + size1))) +#endif /* WCHAR */ + +/* Macros for dealing with the split strings in re_match_2. */ + +#define MATCHING_IN_FIRST_STRING (dend == end_match_1) + +/* Call before fetching a character with *d. This switches over to + string2 if necessary. */ +#define PREFETCH() \ + while (d == dend) \ + { \ + /* End of string2 => fail. */ \ + if (dend == end_match_2) \ + goto fail; \ + /* End of string1 => advance to string2. */ \ + d = string2; \ + dend = end_match_2; \ + } + +/* Test if at very beginning or at very end of the virtual concatenation + of `string1' and `string2'. If only one string, it's `string2'. */ +#define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) +#define AT_STRINGS_END(d) ((d) == end2) + + +/* Test if D points to a character which is word-constituent. We have + two special cases to check for: if past the end of string1, look at + the first character in string2; and if before the beginning of + string2, look at the last character in string1. */ +#ifdef WCHAR +/* Use internationalized API instead of SYNTAX. */ +# define WORDCHAR_P(d) \ + (iswalnum ((wint_t)((d) == end1 ? *string2 \ + : (d) == string2 - 1 ? *(end1 - 1) : *(d))) != 0) +#else /* BYTE */ +# define WORDCHAR_P(d) \ + (SYNTAX ((d) == end1 ? *string2 \ + : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ + == Sword) +#endif /* WCHAR */ + +/* Disabled due to a compiler bug -- see comment at case wordbound */ +#if 0 +/* Test if the character before D and the one at D differ with respect + to being word-constituent. */ +#define AT_WORD_BOUNDARY(d) \ + (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ + || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) +#endif + +/* Free everything we malloc. */ +#ifdef MATCH_MAY_ALLOCATE +# ifdef WCHAR +# define FREE_VARIABLES() \ + do { \ + REGEX_FREE_STACK (fail_stack.stack); \ + FREE_VAR (regstart); \ + FREE_VAR (regend); \ + FREE_VAR (old_regstart); \ + FREE_VAR (old_regend); \ + FREE_VAR (best_regstart); \ + FREE_VAR (best_regend); \ + FREE_VAR (reg_info); \ + FREE_VAR (reg_dummy); \ + FREE_VAR (reg_info_dummy); \ + if (!cant_free_wcs_buf) \ + { \ + FREE_VAR (string1); \ + FREE_VAR (string2); \ + FREE_VAR (mbs_offset1); \ + FREE_VAR (mbs_offset2); \ + } \ + } while (0) +# else /* BYTE */ +# define FREE_VARIABLES() \ + do { \ + REGEX_FREE_STACK (fail_stack.stack); \ + FREE_VAR (regstart); \ + FREE_VAR (regend); \ + FREE_VAR (old_regstart); \ + FREE_VAR (old_regend); \ + FREE_VAR (best_regstart); \ + FREE_VAR (best_regend); \ + FREE_VAR (reg_info); \ + FREE_VAR (reg_dummy); \ + FREE_VAR (reg_info_dummy); \ + } while (0) +# endif /* WCHAR */ +#else +# ifdef WCHAR +# define FREE_VARIABLES() \ + do { \ + if (!cant_free_wcs_buf) \ + { \ + FREE_VAR (string1); \ + FREE_VAR (string2); \ + FREE_VAR (mbs_offset1); \ + FREE_VAR (mbs_offset2); \ + } \ + } while (0) +# else /* BYTE */ +# define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ +# endif /* WCHAR */ +#endif /* not MATCH_MAY_ALLOCATE */ + +/* These values must meet several constraints. They must not be valid + register values; since we have a limit of 255 registers (because + we use only one byte in the pattern for the register number), we can + use numbers larger than 255. They must differ by 1, because of + NUM_FAILURE_ITEMS above. And the value for the lowest register must + be larger than the value for the highest register, so we do not try + to actually save any registers when none are active. */ +#define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH) +#define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1) + +#else /* not INSIDE_RECURSION */ +/* Matching routines. */ + +#ifndef emacs /* Emacs never uses this. */ +/* re_match is like re_match_2 except it takes only a single string. */ + +int +re_match (bufp, string, size, pos, regs) + struct re_pattern_buffer *bufp; + const char *string; + int size, pos; + struct re_registers *regs; +{ + int result; +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + result = wcs_re_match_2_internal (bufp, NULL, 0, string, size, + pos, regs, size, + NULL, 0, NULL, 0, NULL, NULL); + else +# endif + result = byte_re_match_2_internal (bufp, NULL, 0, string, size, + pos, regs, size); +# ifndef REGEX_MALLOC +# ifdef C_ALLOCA + alloca (0); +# endif +# endif + return result; +} +# ifdef _LIBC +weak_alias (__re_match, re_match) +# endif +#endif /* not emacs */ + +#endif /* not INSIDE_RECURSION */ + +#ifdef INSIDE_RECURSION +static boolean PREFIX(group_match_null_string_p) _RE_ARGS ((UCHAR_T **p, + UCHAR_T *end, + PREFIX(register_info_type) *reg_info)); +static boolean PREFIX(alt_match_null_string_p) _RE_ARGS ((UCHAR_T *p, + UCHAR_T *end, + PREFIX(register_info_type) *reg_info)); +static boolean PREFIX(common_op_match_null_string_p) _RE_ARGS ((UCHAR_T **p, + UCHAR_T *end, + PREFIX(register_info_type) *reg_info)); +static int PREFIX(bcmp_translate) _RE_ARGS ((const CHAR_T *s1, const CHAR_T *s2, + int len, char *translate)); +#else /* not INSIDE_RECURSION */ + +/* re_match_2 matches the compiled pattern in BUFP against the + the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 + and SIZE2, respectively). We start matching at POS, and stop + matching at STOP. + + If REGS is non-null and the `no_sub' field of BUFP is nonzero, we + store offsets for the substring each group matched in REGS. See the + documentation for exactly how many groups we fill. + + We return -1 if no match, -2 if an internal error (such as the + failure stack overflowing). Otherwise, we return the length of the + matched substring. */ + +int +re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) + struct re_pattern_buffer *bufp; + const char *string1, *string2; + int size1, size2; + int pos; + struct re_registers *regs; + int stop; +{ + int result; +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + result = wcs_re_match_2_internal (bufp, string1, size1, string2, size2, + pos, regs, stop, + NULL, 0, NULL, 0, NULL, NULL); + else +# endif + result = byte_re_match_2_internal (bufp, string1, size1, string2, size2, + pos, regs, stop); + +#ifndef REGEX_MALLOC +# ifdef C_ALLOCA + alloca (0); +# endif +#endif + return result; +} +#ifdef _LIBC +weak_alias (__re_match_2, re_match_2) +#endif + +#endif /* not INSIDE_RECURSION */ + +#ifdef INSIDE_RECURSION + +#ifdef WCHAR +static int count_mbs_length PARAMS ((int *, int)); + +/* This check the substring (from 0, to length) of the multibyte string, + to which offset_buffer correspond. And count how many wchar_t_characters + the substring occupy. We use offset_buffer to optimization. + See convert_mbs_to_wcs. */ + +static int +count_mbs_length(offset_buffer, length) + int *offset_buffer; + int length; +{ + int upper, lower; + + /* Check whether the size is valid. */ + if (length < 0) + return -1; + + if (offset_buffer == NULL) + return 0; + + /* If there are no multibyte character, offset_buffer[i] == i. + Optmize for this case. */ + if (offset_buffer[length] == length) + return length; + + /* Set up upper with length. (because for all i, offset_buffer[i] >= i) */ + upper = length; + lower = 0; + + while (true) + { + int middle = (lower + upper) / 2; + if (middle == lower || middle == upper) + break; + if (offset_buffer[middle] > length) + upper = middle; + else if (offset_buffer[middle] < length) + lower = middle; + else + return middle; + } + + return -1; +} +#endif /* WCHAR */ + +/* This is a separate function so that we can force an alloca cleanup + afterwards. */ +#ifdef WCHAR +static int +wcs_re_match_2_internal (bufp, cstring1, csize1, cstring2, csize2, pos, + regs, stop, string1, size1, string2, size2, + mbs_offset1, mbs_offset2) + struct re_pattern_buffer *bufp; + const char *cstring1, *cstring2; + int csize1, csize2; + int pos; + struct re_registers *regs; + int stop; + /* string1 == string2 == NULL means string1/2, size1/2 and + mbs_offset1/2 need seting up in this function. */ + /* We need wchar_t* buffers correspond to cstring1, cstring2. */ + wchar_t *string1, *string2; + /* We need the size of wchar_t buffers correspond to csize1, csize2. */ + int size1, size2; + /* offset buffer for optimizatoin. See convert_mbs_to_wc. */ + int *mbs_offset1, *mbs_offset2; +#else /* BYTE */ +static int +byte_re_match_2_internal (bufp, string1, size1,string2, size2, pos, + regs, stop) + struct re_pattern_buffer *bufp; + const char *string1, *string2; + int size1, size2; + int pos; + struct re_registers *regs; + int stop; +#endif /* BYTE */ +{ + /* General temporaries. */ + int mcnt; + UCHAR_T *p1; +#ifdef WCHAR + /* They hold whether each wchar_t is binary data or not. */ + char *is_binary = NULL; + /* If true, we can't free string1/2, mbs_offset1/2. */ + int cant_free_wcs_buf = 1; +#endif /* WCHAR */ + + /* Just past the end of the corresponding string. */ + const CHAR_T *end1, *end2; + + /* Pointers into string1 and string2, just past the last characters in + each to consider matching. */ + const CHAR_T *end_match_1, *end_match_2; + + /* Where we are in the data, and the end of the current string. */ + const CHAR_T *d, *dend; + + /* Where we are in the pattern, and the end of the pattern. */ +#ifdef WCHAR + UCHAR_T *pattern, *p; + register UCHAR_T *pend; +#else /* BYTE */ + UCHAR_T *p = bufp->buffer; + register UCHAR_T *pend = p + bufp->used; +#endif /* WCHAR */ + + /* Mark the opcode just after a start_memory, so we can test for an + empty subpattern when we get to the stop_memory. */ + UCHAR_T *just_past_start_mem = 0; + + /* We use this to map every character in the string. */ + RE_TRANSLATE_TYPE translate = bufp->translate; + + /* Failure point stack. Each place that can handle a failure further + down the line pushes a failure point on this stack. It consists of + restart, regend, and reg_info for all registers corresponding to + the subexpressions we're currently inside, plus the number of such + registers, and, finally, two char *'s. The first char * is where + to resume scanning the pattern; the second one is where to resume + scanning the strings. If the latter is zero, the failure point is + a ``dummy''; if a failure happens and the failure point is a dummy, + it gets discarded and the next next one is tried. */ +#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ + PREFIX(fail_stack_type) fail_stack; +#endif +#ifdef DEBUG + static unsigned failure_id; + unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; +#endif + +#ifdef REL_ALLOC + /* This holds the pointer to the failure stack, when + it is allocated relocatably. */ + fail_stack_elt_t *failure_stack_ptr; +#endif + + /* We fill all the registers internally, independent of what we + return, for use in backreferences. The number here includes + an element for register zero. */ + size_t num_regs = bufp->re_nsub + 1; + + /* The currently active registers. */ + active_reg_t lowest_active_reg = NO_LOWEST_ACTIVE_REG; + active_reg_t highest_active_reg = NO_HIGHEST_ACTIVE_REG; + + /* Information on the contents of registers. These are pointers into + the input strings; they record just what was matched (on this + attempt) by a subexpression part of the pattern, that is, the + regnum-th regstart pointer points to where in the pattern we began + matching and the regnum-th regend points to right after where we + stopped matching the regnum-th subexpression. (The zeroth register + keeps track of what the whole pattern matches.) */ +#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ + const CHAR_T **regstart, **regend; +#endif + + /* If a group that's operated upon by a repetition operator fails to + match anything, then the register for its start will need to be + restored because it will have been set to wherever in the string we + are when we last see its open-group operator. Similarly for a + register's end. */ +#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ + const CHAR_T **old_regstart, **old_regend; +#endif + + /* The is_active field of reg_info helps us keep track of which (possibly + nested) subexpressions we are currently in. The matched_something + field of reg_info[reg_num] helps us tell whether or not we have + matched any of the pattern so far this time through the reg_num-th + subexpression. These two fields get reset each time through any + loop their register is in. */ +#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ + PREFIX(register_info_type) *reg_info; +#endif + + /* The following record the register info as found in the above + variables when we find a match better than any we've seen before. + This happens as we backtrack through the failure points, which in + turn happens only if we have not yet matched the entire string. */ + unsigned best_regs_set = false; +#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ + const CHAR_T **best_regstart, **best_regend; +#endif + + /* Logically, this is `best_regend[0]'. But we don't want to have to + allocate space for that if we're not allocating space for anything + else (see below). Also, we never need info about register 0 for + any of the other register vectors, and it seems rather a kludge to + treat `best_regend' differently than the rest. So we keep track of + the end of the best match so far in a separate variable. We + initialize this to NULL so that when we backtrack the first time + and need to test it, it's not garbage. */ + const CHAR_T *match_end = NULL; + + /* This helps SET_REGS_MATCHED avoid doing redundant work. */ + int set_regs_matched_done = 0; + + /* Used when we pop values we don't care about. */ +#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ + const CHAR_T **reg_dummy; + PREFIX(register_info_type) *reg_info_dummy; +#endif + +#ifdef DEBUG + /* Counts the total number of registers pushed. */ + unsigned num_regs_pushed = 0; +#endif + + DEBUG_PRINT1 ("\n\nEntering re_match_2.\n"); + + INIT_FAIL_STACK (); + +#ifdef MATCH_MAY_ALLOCATE + /* Do not bother to initialize all the register variables if there are + no groups in the pattern, as it takes a fair amount of time. If + there are groups, we include space for register 0 (the whole + pattern), even though we never use it, since it simplifies the + array indexing. We should fix this. */ + if (bufp->re_nsub) + { + regstart = REGEX_TALLOC (num_regs, const CHAR_T *); + regend = REGEX_TALLOC (num_regs, const CHAR_T *); + old_regstart = REGEX_TALLOC (num_regs, const CHAR_T *); + old_regend = REGEX_TALLOC (num_regs, const CHAR_T *); + best_regstart = REGEX_TALLOC (num_regs, const CHAR_T *); + best_regend = REGEX_TALLOC (num_regs, const CHAR_T *); + reg_info = REGEX_TALLOC (num_regs, PREFIX(register_info_type)); + reg_dummy = REGEX_TALLOC (num_regs, const CHAR_T *); + reg_info_dummy = REGEX_TALLOC (num_regs, PREFIX(register_info_type)); + + if (!(regstart && regend && old_regstart && old_regend && reg_info + && best_regstart && best_regend && reg_dummy && reg_info_dummy)) + { + FREE_VARIABLES (); + return -2; + } + } + else + { + /* We must initialize all our variables to NULL, so that + `FREE_VARIABLES' doesn't try to free them. */ + regstart = regend = old_regstart = old_regend = best_regstart + = best_regend = reg_dummy = NULL; + reg_info = reg_info_dummy = (PREFIX(register_info_type) *) NULL; + } +#endif /* MATCH_MAY_ALLOCATE */ + + /* The starting position is bogus. */ +#ifdef WCHAR + if (pos < 0 || pos > csize1 + csize2) +#else /* BYTE */ + if (pos < 0 || pos > size1 + size2) +#endif + { + FREE_VARIABLES (); + return -1; + } + +#ifdef WCHAR + /* Allocate wchar_t array for string1 and string2 and + fill them with converted string. */ + if (string1 == NULL && string2 == NULL) + { + /* We need seting up buffers here. */ + + /* We must free wcs buffers in this function. */ + cant_free_wcs_buf = 0; + + if (csize1 != 0) + { + string1 = REGEX_TALLOC (csize1 + 1, CHAR_T); + mbs_offset1 = REGEX_TALLOC (csize1 + 1, int); + is_binary = REGEX_TALLOC (csize1 + 1, char); + if (!string1 || !mbs_offset1 || !is_binary) + { + FREE_VAR (string1); + FREE_VAR (mbs_offset1); + FREE_VAR (is_binary); + return -2; + } + } + if (csize2 != 0) + { + string2 = REGEX_TALLOC (csize2 + 1, CHAR_T); + mbs_offset2 = REGEX_TALLOC (csize2 + 1, int); + is_binary = REGEX_TALLOC (csize2 + 1, char); + if (!string2 || !mbs_offset2 || !is_binary) + { + FREE_VAR (string1); + FREE_VAR (mbs_offset1); + FREE_VAR (string2); + FREE_VAR (mbs_offset2); + FREE_VAR (is_binary); + return -2; + } + size2 = convert_mbs_to_wcs(string2, cstring2, csize2, + mbs_offset2, is_binary); + string2[size2] = L'\0'; /* for a sentinel */ + FREE_VAR (is_binary); + } + } + + /* We need to cast pattern to (wchar_t*), because we casted this compiled + pattern to (char*) in regex_compile. */ + p = pattern = (CHAR_T*)bufp->buffer; + pend = (CHAR_T*)(bufp->buffer + bufp->used); + +#endif /* WCHAR */ + + /* Initialize subexpression text positions to -1 to mark ones that no + start_memory/stop_memory has been seen for. Also initialize the + register information struct. */ + for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++) + { + regstart[mcnt] = regend[mcnt] + = old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE; + + REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE; + IS_ACTIVE (reg_info[mcnt]) = 0; + MATCHED_SOMETHING (reg_info[mcnt]) = 0; + EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0; + } + + /* We move `string1' into `string2' if the latter's empty -- but not if + `string1' is null. */ + if (size2 == 0 && string1 != NULL) + { + string2 = string1; + size2 = size1; + string1 = 0; + size1 = 0; +#ifdef WCHAR + mbs_offset2 = mbs_offset1; + csize2 = csize1; + mbs_offset1 = NULL; + csize1 = 0; +#endif + } + end1 = string1 + size1; + end2 = string2 + size2; + + /* Compute where to stop matching, within the two strings. */ +#ifdef WCHAR + if (stop <= csize1) + { + mcnt = count_mbs_length(mbs_offset1, stop); + end_match_1 = string1 + mcnt; + end_match_2 = string2; + } + else + { + if (stop > csize1 + csize2) + stop = csize1 + csize2; + end_match_1 = end1; + mcnt = count_mbs_length(mbs_offset2, stop-csize1); + end_match_2 = string2 + mcnt; + } + if (mcnt < 0) + { /* count_mbs_length return error. */ + FREE_VARIABLES (); + return -1; + } +#else + if (stop <= size1) + { + end_match_1 = string1 + stop; + end_match_2 = string2; + } + else + { + end_match_1 = end1; + end_match_2 = string2 + stop - size1; + } +#endif /* WCHAR */ + + /* `p' scans through the pattern as `d' scans through the data. + `dend' is the end of the input string that `d' points within. `d' + is advanced into the following input string whenever necessary, but + this happens before fetching; therefore, at the beginning of the + loop, `d' can be pointing at the end of a string, but it cannot + equal `string2'. */ +#ifdef WCHAR + if (size1 > 0 && pos <= csize1) + { + mcnt = count_mbs_length(mbs_offset1, pos); + d = string1 + mcnt; + dend = end_match_1; + } + else + { + mcnt = count_mbs_length(mbs_offset2, pos-csize1); + d = string2 + mcnt; + dend = end_match_2; + } + + if (mcnt < 0) + { /* count_mbs_length return error. */ + FREE_VARIABLES (); + return -1; + } +#else + if (size1 > 0 && pos <= size1) + { + d = string1 + pos; + dend = end_match_1; + } + else + { + d = string2 + pos - size1; + dend = end_match_2; + } +#endif /* WCHAR */ + + DEBUG_PRINT1 ("The compiled pattern is:\n"); + DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); + DEBUG_PRINT1 ("The string to match is: `"); + DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); + DEBUG_PRINT1 ("'\n"); + + /* This loops over pattern commands. It exits by returning from the + function if the match is complete, or it drops through if the match + fails at this starting point in the input data. */ + for (;;) + { +#ifdef _LIBC + DEBUG_PRINT2 ("\n%p: ", p); +#else + DEBUG_PRINT2 ("\n0x%x: ", p); +#endif + + if (p == pend) + { /* End of pattern means we might have succeeded. */ + DEBUG_PRINT1 ("end of pattern ... "); + + /* If we haven't matched the entire string, and we want the + longest match, try backtracking. */ + if (d != end_match_2) + { + /* 1 if this match ends in the same string (string1 or string2) + as the best previous match. */ + boolean same_str_p = (FIRST_STRING_P (match_end) + == MATCHING_IN_FIRST_STRING); + /* 1 if this match is the best seen so far. */ + boolean best_match_p; + + /* AIX compiler got confused when this was combined + with the previous declaration. */ + if (same_str_p) + best_match_p = d > match_end; + else + best_match_p = !MATCHING_IN_FIRST_STRING; + + DEBUG_PRINT1 ("backtracking.\n"); + + if (!FAIL_STACK_EMPTY ()) + { /* More failure points to try. */ + + /* If exceeds best match so far, save it. */ + if (!best_regs_set || best_match_p) + { + best_regs_set = true; + match_end = d; + + DEBUG_PRINT1 ("\nSAVING match as best so far.\n"); + + for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++) + { + best_regstart[mcnt] = regstart[mcnt]; + best_regend[mcnt] = regend[mcnt]; + } + } + goto fail; + } + + /* If no failure points, don't restore garbage. And if + last match is real best match, don't restore second + best one. */ + else if (best_regs_set && !best_match_p) + { + restore_best_regs: + /* Restore best match. It may happen that `dend == + end_match_1' while the restored d is in string2. + For example, the pattern `x.*y.*z' against the + strings `x-' and `y-z-', if the two strings are + not consecutive in memory. */ + DEBUG_PRINT1 ("Restoring best registers.\n"); + + d = match_end; + dend = ((d >= string1 && d <= end1) + ? end_match_1 : end_match_2); + + for (mcnt = 1; (unsigned) mcnt < num_regs; mcnt++) + { + regstart[mcnt] = best_regstart[mcnt]; + regend[mcnt] = best_regend[mcnt]; + } + } + } /* d != end_match_2 */ + + succeed_label: + DEBUG_PRINT1 ("Accepting match.\n"); + /* If caller wants register contents data back, do it. */ + if (regs && !bufp->no_sub) + { + /* Have the register data arrays been allocated? */ + if (bufp->regs_allocated == REGS_UNALLOCATED) + { /* No. So allocate them with malloc. We need one + extra element beyond `num_regs' for the `-1' marker + GNU code uses. */ + regs->num_regs = MAX (RE_NREGS, num_regs + 1); + regs->start = TALLOC (regs->num_regs, regoff_t); + regs->end = TALLOC (regs->num_regs, regoff_t); + if (regs->start == NULL || regs->end == NULL) + { + FREE_VARIABLES (); + return -2; + } + bufp->regs_allocated = REGS_REALLOCATE; + } + else if (bufp->regs_allocated == REGS_REALLOCATE) + { /* Yes. If we need more elements than were already + allocated, reallocate them. If we need fewer, just + leave it alone. */ + if (regs->num_regs < num_regs + 1) + { + regs->num_regs = num_regs + 1; + RETALLOC (regs->start, regs->num_regs, regoff_t); + RETALLOC (regs->end, regs->num_regs, regoff_t); + if (regs->start == NULL || regs->end == NULL) + { + FREE_VARIABLES (); + return -2; + } + } + } + else + { + /* These braces fend off a "empty body in an else-statement" + warning under GCC when assert expands to nothing. */ + assert (bufp->regs_allocated == REGS_FIXED); + } + + /* Convert the pointer data in `regstart' and `regend' to + indices. Register zero has to be set differently, + since we haven't kept track of any info for it. */ + if (regs->num_regs > 0) + { + regs->start[0] = pos; +#ifdef WCHAR + if (MATCHING_IN_FIRST_STRING) + regs->end[0] = mbs_offset1 != NULL ? + mbs_offset1[d-string1] : 0; + else + regs->end[0] = csize1 + (mbs_offset2 != NULL ? + mbs_offset2[d-string2] : 0); +#else + regs->end[0] = (MATCHING_IN_FIRST_STRING + ? ((regoff_t) (d - string1)) + : ((regoff_t) (d - string2 + size1))); +#endif /* WCHAR */ + } + + /* Go through the first `min (num_regs, regs->num_regs)' + registers, since that is all we initialized. */ + for (mcnt = 1; (unsigned) mcnt < MIN (num_regs, regs->num_regs); + mcnt++) + { + if (REG_UNSET (regstart[mcnt]) || REG_UNSET (regend[mcnt])) + regs->start[mcnt] = regs->end[mcnt] = -1; + else + { + regs->start[mcnt] + = (regoff_t) POINTER_TO_OFFSET (regstart[mcnt]); + regs->end[mcnt] + = (regoff_t) POINTER_TO_OFFSET (regend[mcnt]); + } + } + + /* If the regs structure we return has more elements than + were in the pattern, set the extra elements to -1. If + we (re)allocated the registers, this is the case, + because we always allocate enough to have at least one + -1 at the end. */ + for (mcnt = num_regs; (unsigned) mcnt < regs->num_regs; mcnt++) + regs->start[mcnt] = regs->end[mcnt] = -1; + } /* regs && !bufp->no_sub */ + + DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", + nfailure_points_pushed, nfailure_points_popped, + nfailure_points_pushed - nfailure_points_popped); + DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed); + +#ifdef WCHAR + if (MATCHING_IN_FIRST_STRING) + mcnt = mbs_offset1 != NULL ? mbs_offset1[d-string1] : 0; + else + mcnt = (mbs_offset2 != NULL ? mbs_offset2[d-string2] : 0) + + csize1; + mcnt -= pos; +#else + mcnt = d - pos - (MATCHING_IN_FIRST_STRING + ? string1 + : string2 - size1); +#endif /* WCHAR */ + + DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt); + + FREE_VARIABLES (); + return mcnt; + } + + /* Otherwise match next pattern command. */ + switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) + { + /* Ignore these. Used to ignore the n of succeed_n's which + currently have n == 0. */ + case no_op: + DEBUG_PRINT1 ("EXECUTING no_op.\n"); + break; + + case succeed: + DEBUG_PRINT1 ("EXECUTING succeed.\n"); + goto succeed_label; + + /* Match the next n pattern characters exactly. The following + byte in the pattern defines n, and the n bytes after that + are the characters to match. */ + case exactn: +#ifdef MBS_SUPPORT + case exactn_bin: +#endif + mcnt = *p++; + DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt); + + /* This is written out as an if-else so we don't waste time + testing `translate' inside the loop. */ + if (translate) + { + do + { + PREFETCH (); +#ifdef WCHAR + if (*d <= 0xff) + { + if ((UCHAR_T) translate[(unsigned char) *d++] + != (UCHAR_T) *p++) + goto fail; + } + else + { + if (*d++ != (CHAR_T) *p++) + goto fail; + } +#else + if ((UCHAR_T) translate[(unsigned char) *d++] + != (UCHAR_T) *p++) + goto fail; +#endif /* WCHAR */ + } + while (--mcnt); + } + else + { + do + { + PREFETCH (); + if (*d++ != (CHAR_T) *p++) goto fail; + } + while (--mcnt); + } + SET_REGS_MATCHED (); + break; + + + /* Match any character except possibly a newline or a null. */ + case anychar: + DEBUG_PRINT1 ("EXECUTING anychar.\n"); + + PREFETCH (); + + if ((!(bufp->syntax & RE_DOT_NEWLINE) && TRANSLATE (*d) == '\n') + || (bufp->syntax & RE_DOT_NOT_NULL && TRANSLATE (*d) == '\000')) + goto fail; + + SET_REGS_MATCHED (); + DEBUG_PRINT2 (" Matched `%ld'.\n", (long int) *d); + d++; + break; + + + case charset: + case charset_not: + { + register UCHAR_T c; +#ifdef WCHAR + unsigned int i, char_class_length, coll_symbol_length, + equiv_class_length, ranges_length, chars_length, length; + CHAR_T *workp, *workp2, *charset_top; +#define WORK_BUFFER_SIZE 128 + CHAR_T str_buf[WORK_BUFFER_SIZE]; +# ifdef _LIBC + uint32_t nrules; +# endif /* _LIBC */ +#endif /* WCHAR */ + boolean not = (re_opcode_t) *(p - 1) == charset_not; + + DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : ""); + PREFETCH (); + c = TRANSLATE (*d); /* The character to match. */ +#ifdef WCHAR +# ifdef _LIBC + nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); +# endif /* _LIBC */ + charset_top = p - 1; + char_class_length = *p++; + coll_symbol_length = *p++; + equiv_class_length = *p++; + ranges_length = *p++; + chars_length = *p++; + /* p points charset[6], so the address of the next instruction + (charset[l+m+n+2o+k+p']) equals p[l+m+n+2*o+p'], + where l=length of char_classes, m=length of collating_symbol, + n=equivalence_class, o=length of char_range, + p'=length of character. */ + workp = p; + /* Update p to indicate the next instruction. */ + p += char_class_length + coll_symbol_length+ equiv_class_length + + 2*ranges_length + chars_length; + + /* match with char_class? */ + for (i = 0; i < char_class_length ; i += CHAR_CLASS_SIZE) + { + wctype_t wctype; + uintptr_t alignedp = ((uintptr_t)workp + + __alignof__(wctype_t) - 1) + & ~(uintptr_t)(__alignof__(wctype_t) - 1); + wctype = *((wctype_t*)alignedp); + workp += CHAR_CLASS_SIZE; + if (iswctype((wint_t)c, wctype)) + goto char_set_matched; + } + + /* match with collating_symbol? */ +# ifdef _LIBC + if (nrules != 0) + { + const unsigned char *extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); + + for (workp2 = workp + coll_symbol_length ; workp < workp2 ; + workp++) + { + int32_t *wextra; + wextra = (int32_t*)(extra + *workp++); + for (i = 0; i < *wextra; ++i) + if (TRANSLATE(d[i]) != wextra[1 + i]) + break; + + if (i == *wextra) + { + /* Update d, however d will be incremented at + char_set_matched:, we decrement d here. */ + d += i - 1; + goto char_set_matched; + } + } + } + else /* (nrules == 0) */ +# endif + /* If we can't look up collation data, we use wcscoll + instead. */ + { + for (workp2 = workp + coll_symbol_length ; workp < workp2 ;) + { + const CHAR_T *backup_d = d, *backup_dend = dend; + length = wcslen(workp); + + /* If wcscoll(the collating symbol, whole string) > 0, + any substring of the string never match with the + collating symbol. */ + if (wcscoll(workp, d) > 0) + { + workp += length + 1; + continue; + } + + /* First, we compare the collating symbol with + the first character of the string. + If it don't match, we add the next character to + the compare buffer in turn. */ + for (i = 0 ; i < WORK_BUFFER_SIZE-1 ; i++, d++) + { + int match; + if (d == dend) + { + if (dend == end_match_2) + break; + d = string2; + dend = end_match_2; + } + + /* add next character to the compare buffer. */ + str_buf[i] = TRANSLATE(*d); + str_buf[i+1] = '\0'; + + match = wcscoll(workp, str_buf); + if (match == 0) + goto char_set_matched; + + if (match < 0) + /* (str_buf > workp) indicate (str_buf + X > workp), + because for all X (str_buf + X > str_buf). + So we don't need continue this loop. */ + break; + + /* Otherwise(str_buf < workp), + (str_buf+next_character) may equals (workp). + So we continue this loop. */ + } + /* not matched */ + d = backup_d; + dend = backup_dend; + workp += length + 1; + } + } + /* match with equivalence_class? */ +# ifdef _LIBC + if (nrules != 0) + { + const CHAR_T *backup_d = d, *backup_dend = dend; + /* Try to match the equivalence class against + those known to the collate implementation. */ + const int32_t *table; + const int32_t *weights; + const int32_t *extra; + const int32_t *indirect; + int32_t idx, idx2; + wint_t *cp; + size_t len; + + /* This #include defines a local function! */ +# include + + table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEWC); + weights = (const wint_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTWC); + extra = (const wint_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAWC); + indirect = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTWC); + + /* Write 1 collating element to str_buf, and + get its index. */ + idx2 = 0; + + for (i = 0 ; idx2 == 0 && i < WORK_BUFFER_SIZE - 1; i++) + { + cp = (wint_t*)str_buf; + if (d == dend) + { + if (dend == end_match_2) + break; + d = string2; + dend = end_match_2; + } + str_buf[i] = TRANSLATE(*(d+i)); + str_buf[i+1] = '\0'; /* sentinel */ + idx2 = findidx ((const wint_t**)&cp); + } + + /* Update d, however d will be incremented at + char_set_matched:, we decrement d here. */ + d = backup_d + ((wchar_t*)cp - (wchar_t*)str_buf - 1); + if (d >= dend) + { + if (dend == end_match_2) + d = dend; + else + { + d = string2; + dend = end_match_2; + } + } + + len = weights[idx2]; + + for (workp2 = workp + equiv_class_length ; workp < workp2 ; + workp++) + { + idx = (int32_t)*workp; + /* We already checked idx != 0 in regex_compile. */ + + if (idx2 != 0 && len == weights[idx]) + { + int cnt = 0; + while (cnt < len && (weights[idx + 1 + cnt] + == weights[idx2 + 1 + cnt])) + ++cnt; + + if (cnt == len) + goto char_set_matched; + } + } + /* not matched */ + d = backup_d; + dend = backup_dend; + } + else /* (nrules == 0) */ +# endif + /* If we can't look up collation data, we use wcscoll + instead. */ + { + for (workp2 = workp + equiv_class_length ; workp < workp2 ;) + { + const CHAR_T *backup_d = d, *backup_dend = dend; + length = wcslen(workp); + + /* If wcscoll(the collating symbol, whole string) > 0, + any substring of the string never match with the + collating symbol. */ + if (wcscoll(workp, d) > 0) + { + workp += length + 1; + break; + } + + /* First, we compare the equivalence class with + the first character of the string. + If it don't match, we add the next character to + the compare buffer in turn. */ + for (i = 0 ; i < WORK_BUFFER_SIZE - 1 ; i++, d++) + { + int match; + if (d == dend) + { + if (dend == end_match_2) + break; + d = string2; + dend = end_match_2; + } + + /* add next character to the compare buffer. */ + str_buf[i] = TRANSLATE(*d); + str_buf[i+1] = '\0'; + + match = wcscoll(workp, str_buf); + + if (match == 0) + goto char_set_matched; + + if (match < 0) + /* (str_buf > workp) indicate (str_buf + X > workp), + because for all X (str_buf + X > str_buf). + So we don't need continue this loop. */ + break; + + /* Otherwise(str_buf < workp), + (str_buf+next_character) may equals (workp). + So we continue this loop. */ + } + /* not matched */ + d = backup_d; + dend = backup_dend; + workp += length + 1; + } + } + + /* match with char_range? */ +#ifdef _LIBC + if (nrules != 0) + { + uint32_t collseqval; + const char *collseq = (const char *) + _NL_CURRENT(LC_COLLATE, _NL_COLLATE_COLLSEQWC); + + collseqval = collseq_table_lookup (collseq, c); + + for (; workp < p - chars_length ;) + { + uint32_t start_val, end_val; + + /* We already compute the collation sequence value + of the characters (or collating symbols). */ + start_val = (uint32_t) *workp++; /* range_start */ + end_val = (uint32_t) *workp++; /* range_end */ + + if (start_val <= collseqval && collseqval <= end_val) + goto char_set_matched; + } + } + else +#endif + { + /* We set range_start_char at str_buf[0], range_end_char + at str_buf[4], and compared char at str_buf[2]. */ + str_buf[1] = 0; + str_buf[2] = c; + str_buf[3] = 0; + str_buf[5] = 0; + for (; workp < p - chars_length ;) + { + wchar_t *range_start_char, *range_end_char; + + /* match if (range_start_char <= c <= range_end_char). */ + + /* If range_start(or end) < 0, we assume -range_start(end) + is the offset of the collating symbol which is specified + as the character of the range start(end). */ + + /* range_start */ + if (*workp < 0) + range_start_char = charset_top - (*workp++); + else + { + str_buf[0] = *workp++; + range_start_char = str_buf; + } + + /* range_end */ + if (*workp < 0) + range_end_char = charset_top - (*workp++); + else + { + str_buf[4] = *workp++; + range_end_char = str_buf + 4; + } + + if (wcscoll(range_start_char, str_buf+2) <= 0 && + wcscoll(str_buf+2, range_end_char) <= 0) + + goto char_set_matched; + } + } + + /* match with char? */ + for (; workp < p ; workp++) + if (c == *workp) + goto char_set_matched; + + not = !not; + + char_set_matched: + if (not) goto fail; +#else + /* Cast to `unsigned' instead of `unsigned char' in case the + bit list is a full 32 bytes long. */ + if (c < (unsigned) (*p * BYTEWIDTH) + && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) + not = !not; + + p += 1 + *p; + + if (!not) goto fail; +#undef WORK_BUFFER_SIZE +#endif /* WCHAR */ + SET_REGS_MATCHED (); + d++; + break; + } + + + /* The beginning of a group is represented by start_memory. + The arguments are the register number in the next byte, and the + number of groups inner to this one in the next. The text + matched within the group is recorded (in the internal + registers data structure) under the register number. */ + case start_memory: + DEBUG_PRINT3 ("EXECUTING start_memory %ld (%ld):\n", + (long int) *p, (long int) p[1]); + + /* Find out if this group can match the empty string. */ + p1 = p; /* To send to group_match_null_string_p. */ + + if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE) + REG_MATCH_NULL_STRING_P (reg_info[*p]) + = PREFIX(group_match_null_string_p) (&p1, pend, reg_info); + + /* Save the position in the string where we were the last time + we were at this open-group operator in case the group is + operated upon by a repetition operator, e.g., with `(a*)*b' + against `ab'; then we want to ignore where we are now in + the string in case this attempt to match fails. */ + old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) + ? REG_UNSET (regstart[*p]) ? d : regstart[*p] + : regstart[*p]; + DEBUG_PRINT2 (" old_regstart: %d\n", + POINTER_TO_OFFSET (old_regstart[*p])); + + regstart[*p] = d; + DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); + + IS_ACTIVE (reg_info[*p]) = 1; + MATCHED_SOMETHING (reg_info[*p]) = 0; + + /* Clear this whenever we change the register activity status. */ + set_regs_matched_done = 0; + + /* This is the new highest active register. */ + highest_active_reg = *p; + + /* If nothing was active before, this is the new lowest active + register. */ + if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) + lowest_active_reg = *p; + + /* Move past the register number and inner group count. */ + p += 2; + just_past_start_mem = p; + + break; + + + /* The stop_memory opcode represents the end of a group. Its + arguments are the same as start_memory's: the register + number, and the number of inner groups. */ + case stop_memory: + DEBUG_PRINT3 ("EXECUTING stop_memory %ld (%ld):\n", + (long int) *p, (long int) p[1]); + + /* We need to save the string position the last time we were at + this close-group operator in case the group is operated + upon by a repetition operator, e.g., with `((a*)*(b*)*)*' + against `aba'; then we want to ignore where we are now in + the string in case this attempt to match fails. */ + old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) + ? REG_UNSET (regend[*p]) ? d : regend[*p] + : regend[*p]; + DEBUG_PRINT2 (" old_regend: %d\n", + POINTER_TO_OFFSET (old_regend[*p])); + + regend[*p] = d; + DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); + + /* This register isn't active anymore. */ + IS_ACTIVE (reg_info[*p]) = 0; + + /* Clear this whenever we change the register activity status. */ + set_regs_matched_done = 0; + + /* If this was the only register active, nothing is active + anymore. */ + if (lowest_active_reg == highest_active_reg) + { + lowest_active_reg = NO_LOWEST_ACTIVE_REG; + highest_active_reg = NO_HIGHEST_ACTIVE_REG; + } + else + { /* We must scan for the new highest active register, since + it isn't necessarily one less than now: consider + (a(b)c(d(e)f)g). When group 3 ends, after the f), the + new highest active register is 1. */ + UCHAR_T r = *p - 1; + while (r > 0 && !IS_ACTIVE (reg_info[r])) + r--; + + /* If we end up at register zero, that means that we saved + the registers as the result of an `on_failure_jump', not + a `start_memory', and we jumped to past the innermost + `stop_memory'. For example, in ((.)*) we save + registers 1 and 2 as a result of the *, but when we pop + back to the second ), we are at the stop_memory 1. + Thus, nothing is active. */ + if (r == 0) + { + lowest_active_reg = NO_LOWEST_ACTIVE_REG; + highest_active_reg = NO_HIGHEST_ACTIVE_REG; + } + else + highest_active_reg = r; + } + + /* If just failed to match something this time around with a + group that's operated on by a repetition operator, try to + force exit from the ``loop'', and restore the register + information for this group that we had before trying this + last match. */ + if ((!MATCHED_SOMETHING (reg_info[*p]) + || just_past_start_mem == p - 1) + && (p + 2) < pend) + { + boolean is_a_jump_n = false; + + p1 = p + 2; + mcnt = 0; + switch ((re_opcode_t) *p1++) + { + case jump_n: + is_a_jump_n = true; + case pop_failure_jump: + case maybe_pop_jump: + case jump: + case dummy_failure_jump: + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + if (is_a_jump_n) + p1 += OFFSET_ADDRESS_SIZE; + break; + + default: + /* do nothing */ ; + } + p1 += mcnt; + + /* If the next operation is a jump backwards in the pattern + to an on_failure_jump right before the start_memory + corresponding to this stop_memory, exit from the loop + by forcing a failure after pushing on the stack the + on_failure_jump's jump in the pattern, and d. */ + if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump + && (re_opcode_t) p1[1+OFFSET_ADDRESS_SIZE] == start_memory + && p1[2+OFFSET_ADDRESS_SIZE] == *p) + { + /* If this group ever matched anything, then restore + what its registers were before trying this last + failed match, e.g., with `(a*)*b' against `ab' for + regstart[1], and, e.g., with `((a*)*(b*)*)*' + against `aba' for regend[3]. + + Also restore the registers for inner groups for, + e.g., `((a*)(b*))*' against `aba' (register 3 would + otherwise get trashed). */ + + if (EVER_MATCHED_SOMETHING (reg_info[*p])) + { + unsigned r; + + EVER_MATCHED_SOMETHING (reg_info[*p]) = 0; + + /* Restore this and inner groups' (if any) registers. */ + for (r = *p; r < (unsigned) *p + (unsigned) *(p + 1); + r++) + { + regstart[r] = old_regstart[r]; + + /* xx why this test? */ + if (old_regend[r] >= regstart[r]) + regend[r] = old_regend[r]; + } + } + p1++; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + PUSH_FAILURE_POINT (p1 + mcnt, d, -2); + + goto fail; + } + } + + /* Move past the register number and the inner group count. */ + p += 2; + break; + + + /* \ has been turned into a `duplicate' command which is + followed by the numeric value of as the register number. */ + case duplicate: + { + register const CHAR_T *d2, *dend2; + int regno = *p++; /* Get which register to match against. */ + DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno); + + /* Can't back reference a group which we've never matched. */ + if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) + goto fail; + + /* Where in input to try to start matching. */ + d2 = regstart[regno]; + + /* Where to stop matching; if both the place to start and + the place to stop matching are in the same string, then + set to the place to stop, otherwise, for now have to use + the end of the first string. */ + + dend2 = ((FIRST_STRING_P (regstart[regno]) + == FIRST_STRING_P (regend[regno])) + ? regend[regno] : end_match_1); + for (;;) + { + /* If necessary, advance to next segment in register + contents. */ + while (d2 == dend2) + { + if (dend2 == end_match_2) break; + if (dend2 == regend[regno]) break; + + /* End of string1 => advance to string2. */ + d2 = string2; + dend2 = regend[regno]; + } + /* At end of register contents => success */ + if (d2 == dend2) break; + + /* If necessary, advance to next segment in data. */ + PREFETCH (); + + /* How many characters left in this segment to match. */ + mcnt = dend - d; + + /* Want how many consecutive characters we can match in + one shot, so, if necessary, adjust the count. */ + if (mcnt > dend2 - d2) + mcnt = dend2 - d2; + + /* Compare that many; failure if mismatch, else move + past them. */ + if (translate + ? PREFIX(bcmp_translate) (d, d2, mcnt, translate) + : memcmp (d, d2, mcnt*sizeof(UCHAR_T))) + goto fail; + d += mcnt, d2 += mcnt; + + /* Do this because we've match some characters. */ + SET_REGS_MATCHED (); + } + } + break; + + + /* begline matches the empty string at the beginning of the string + (unless `not_bol' is set in `bufp'), and, if + `newline_anchor' is set, after newlines. */ + case begline: + DEBUG_PRINT1 ("EXECUTING begline.\n"); + + if (AT_STRINGS_BEG (d)) + { + if (!bufp->not_bol) break; + } + else if (d[-1] == '\n' && bufp->newline_anchor) + { + break; + } + /* In all other cases, we fail. */ + goto fail; + + + /* endline is the dual of begline. */ + case endline: + DEBUG_PRINT1 ("EXECUTING endline.\n"); + + if (AT_STRINGS_END (d)) + { + if (!bufp->not_eol) break; + } + + /* We have to ``prefetch'' the next character. */ + else if ((d == end1 ? *string2 : *d) == '\n' + && bufp->newline_anchor) + { + break; + } + goto fail; + + + /* Match at the very beginning of the data. */ + case begbuf: + DEBUG_PRINT1 ("EXECUTING begbuf.\n"); + if (AT_STRINGS_BEG (d)) + break; + goto fail; + + + /* Match at the very end of the data. */ + case endbuf: + DEBUG_PRINT1 ("EXECUTING endbuf.\n"); + if (AT_STRINGS_END (d)) + break; + goto fail; + + + /* on_failure_keep_string_jump is used to optimize `.*\n'. It + pushes NULL as the value for the string on the stack. Then + `pop_failure_point' will keep the current value for the + string, instead of restoring it. To see why, consider + matching `foo\nbar' against `.*\n'. The .* matches the foo; + then the . fails against the \n. But the next thing we want + to do is match the \n against the \n; if we restored the + string value, we would be back at the foo. + + Because this is used only in specific cases, we don't need to + check all the things that `on_failure_jump' does, to make + sure the right things get saved on the stack. Hence we don't + share its code. The only reason to push anything on the + stack at all is that otherwise we would have to change + `anychar's code to do something besides goto fail in this + case; that seems worse than this. */ + case on_failure_keep_string_jump: + DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump"); + + EXTRACT_NUMBER_AND_INCR (mcnt, p); +#ifdef _LIBC + DEBUG_PRINT3 (" %d (to %p):\n", mcnt, p + mcnt); +#else + DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt); +#endif + + PUSH_FAILURE_POINT (p + mcnt, NULL, -2); + break; + + + /* Uses of on_failure_jump: + + Each alternative starts with an on_failure_jump that points + to the beginning of the next alternative. Each alternative + except the last ends with a jump that in effect jumps past + the rest of the alternatives. (They really jump to the + ending jump of the following alternative, because tensioning + these jumps is a hassle.) + + Repeats start with an on_failure_jump that points past both + the repetition text and either the following jump or + pop_failure_jump back to this on_failure_jump. */ + case on_failure_jump: + on_failure: + DEBUG_PRINT1 ("EXECUTING on_failure_jump"); + + EXTRACT_NUMBER_AND_INCR (mcnt, p); +#ifdef _LIBC + DEBUG_PRINT3 (" %d (to %p)", mcnt, p + mcnt); +#else + DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt); +#endif + + /* If this on_failure_jump comes right before a group (i.e., + the original * applied to a group), save the information + for that group and all inner ones, so that if we fail back + to this point, the group's information will be correct. + For example, in \(a*\)*\1, we need the preceding group, + and in \(zz\(a*\)b*\)\2, we need the inner group. */ + + /* We can't use `p' to check ahead because we push + a failure point to `p + mcnt' after we do this. */ + p1 = p; + + /* We need to skip no_op's before we look for the + start_memory in case this on_failure_jump is happening as + the result of a completed succeed_n, as in \(a\)\{1,3\}b\1 + against aba. */ + while (p1 < pend && (re_opcode_t) *p1 == no_op) + p1++; + + if (p1 < pend && (re_opcode_t) *p1 == start_memory) + { + /* We have a new highest active register now. This will + get reset at the start_memory we are about to get to, + but we will have saved all the registers relevant to + this repetition op, as described above. */ + highest_active_reg = *(p1 + 1) + *(p1 + 2); + if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) + lowest_active_reg = *(p1 + 1); + } + + DEBUG_PRINT1 (":\n"); + PUSH_FAILURE_POINT (p + mcnt, d, -2); + break; + + + /* A smart repeat ends with `maybe_pop_jump'. + We change it to either `pop_failure_jump' or `jump'. */ + case maybe_pop_jump: + EXTRACT_NUMBER_AND_INCR (mcnt, p); + DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt); + { + register UCHAR_T *p2 = p; + + /* Compare the beginning of the repeat with what in the + pattern follows its end. If we can establish that there + is nothing that they would both match, i.e., that we + would have to backtrack because of (as in, e.g., `a*a') + then we can change to pop_failure_jump, because we'll + never have to backtrack. + + This is not true in the case of alternatives: in + `(a|ab)*' we do need to backtrack to the `ab' alternative + (e.g., if the string was `ab'). But instead of trying to + detect that here, the alternative has put on a dummy + failure point which is what we will end up popping. */ + + /* Skip over open/close-group commands. + If what follows this loop is a ...+ construct, + look at what begins its body, since we will have to + match at least one of that. */ + while (1) + { + if (p2 + 2 < pend + && ((re_opcode_t) *p2 == stop_memory + || (re_opcode_t) *p2 == start_memory)) + p2 += 3; + else if (p2 + 2 + 2 * OFFSET_ADDRESS_SIZE < pend + && (re_opcode_t) *p2 == dummy_failure_jump) + p2 += 2 + 2 * OFFSET_ADDRESS_SIZE; + else + break; + } + + p1 = p + mcnt; + /* p1[0] ... p1[2] are the `on_failure_jump' corresponding + to the `maybe_finalize_jump' of this case. Examine what + follows. */ + + /* If we're at the end of the pattern, we can change. */ + if (p2 == pend) + { + /* Consider what happens when matching ":\(.*\)" + against ":/". I don't really understand this code + yet. */ + p[-(1+OFFSET_ADDRESS_SIZE)] = (UCHAR_T) + pop_failure_jump; + DEBUG_PRINT1 + (" End of pattern: change to `pop_failure_jump'.\n"); + } + + else if ((re_opcode_t) *p2 == exactn +#ifdef MBS_SUPPORT + || (re_opcode_t) *p2 == exactn_bin +#endif + || (bufp->newline_anchor && (re_opcode_t) *p2 == endline)) + { + register UCHAR_T c + = *p2 == (UCHAR_T) endline ? '\n' : p2[2]; + + if (((re_opcode_t) p1[1+OFFSET_ADDRESS_SIZE] == exactn +#ifdef MBS_SUPPORT + || (re_opcode_t) p1[1+OFFSET_ADDRESS_SIZE] == exactn_bin +#endif + ) && p1[3+OFFSET_ADDRESS_SIZE] != c) + { + p[-(1+OFFSET_ADDRESS_SIZE)] = (UCHAR_T) + pop_failure_jump; +#ifdef WCHAR + DEBUG_PRINT3 (" %C != %C => pop_failure_jump.\n", + (wint_t) c, + (wint_t) p1[3+OFFSET_ADDRESS_SIZE]); +#else + DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n", + (char) c, + (char) p1[3+OFFSET_ADDRESS_SIZE]); +#endif + } + +#ifndef WCHAR + else if ((re_opcode_t) p1[3] == charset + || (re_opcode_t) p1[3] == charset_not) + { + int not = (re_opcode_t) p1[3] == charset_not; + + if (c < (unsigned) (p1[4] * BYTEWIDTH) + && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) + not = !not; + + /* `not' is equal to 1 if c would match, which means + that we can't change to pop_failure_jump. */ + if (!not) + { + p[-3] = (unsigned char) pop_failure_jump; + DEBUG_PRINT1 (" No match => pop_failure_jump.\n"); + } + } +#endif /* not WCHAR */ + } +#ifndef WCHAR + else if ((re_opcode_t) *p2 == charset) + { + /* We win if the first character of the loop is not part + of the charset. */ + if ((re_opcode_t) p1[3] == exactn + && ! ((int) p2[1] * BYTEWIDTH > (int) p1[5] + && (p2[2 + p1[5] / BYTEWIDTH] + & (1 << (p1[5] % BYTEWIDTH))))) + { + p[-3] = (unsigned char) pop_failure_jump; + DEBUG_PRINT1 (" No match => pop_failure_jump.\n"); + } + + else if ((re_opcode_t) p1[3] == charset_not) + { + int idx; + /* We win if the charset_not inside the loop + lists every character listed in the charset after. */ + for (idx = 0; idx < (int) p2[1]; idx++) + if (! (p2[2 + idx] == 0 + || (idx < (int) p1[4] + && ((p2[2 + idx] & ~ p1[5 + idx]) == 0)))) + break; + + if (idx == p2[1]) + { + p[-3] = (unsigned char) pop_failure_jump; + DEBUG_PRINT1 (" No match => pop_failure_jump.\n"); + } + } + else if ((re_opcode_t) p1[3] == charset) + { + int idx; + /* We win if the charset inside the loop + has no overlap with the one after the loop. */ + for (idx = 0; + idx < (int) p2[1] && idx < (int) p1[4]; + idx++) + if ((p2[2 + idx] & p1[5 + idx]) != 0) + break; + + if (idx == p2[1] || idx == p1[4]) + { + p[-3] = (unsigned char) pop_failure_jump; + DEBUG_PRINT1 (" No match => pop_failure_jump.\n"); + } + } + } +#endif /* not WCHAR */ + } + p -= OFFSET_ADDRESS_SIZE; /* Point at relative address again. */ + if ((re_opcode_t) p[-1] != pop_failure_jump) + { + p[-1] = (UCHAR_T) jump; + DEBUG_PRINT1 (" Match => jump.\n"); + goto unconditional_jump; + } + /* Note fall through. */ + + + /* The end of a simple repeat has a pop_failure_jump back to + its matching on_failure_jump, where the latter will push a + failure point. The pop_failure_jump takes off failure + points put on by this pop_failure_jump's matching + on_failure_jump; we got through the pattern to here from the + matching on_failure_jump, so didn't fail. */ + case pop_failure_jump: + { + /* We need to pass separate storage for the lowest and + highest registers, even though we don't care about the + actual values. Otherwise, we will restore only one + register from the stack, since lowest will == highest in + `pop_failure_point'. */ + active_reg_t dummy_low_reg, dummy_high_reg; + UCHAR_T *pdummy = NULL; + const CHAR_T *sdummy = NULL; + + DEBUG_PRINT1 ("EXECUTING pop_failure_jump.\n"); + POP_FAILURE_POINT (sdummy, pdummy, + dummy_low_reg, dummy_high_reg, + reg_dummy, reg_dummy, reg_info_dummy); + } + /* Note fall through. */ + + unconditional_jump: +#ifdef _LIBC + DEBUG_PRINT2 ("\n%p: ", p); +#else + DEBUG_PRINT2 ("\n0x%x: ", p); +#endif + /* Note fall through. */ + + /* Unconditionally jump (without popping any failure points). */ + case jump: + EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ + DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt); + p += mcnt; /* Do the jump. */ +#ifdef _LIBC + DEBUG_PRINT2 ("(to %p).\n", p); +#else + DEBUG_PRINT2 ("(to 0x%x).\n", p); +#endif + break; + + + /* We need this opcode so we can detect where alternatives end + in `group_match_null_string_p' et al. */ + case jump_past_alt: + DEBUG_PRINT1 ("EXECUTING jump_past_alt.\n"); + goto unconditional_jump; + + + /* Normally, the on_failure_jump pushes a failure point, which + then gets popped at pop_failure_jump. We will end up at + pop_failure_jump, also, and with a pattern of, say, `a+', we + are skipping over the on_failure_jump, so we have to push + something meaningless for pop_failure_jump to pop. */ + case dummy_failure_jump: + DEBUG_PRINT1 ("EXECUTING dummy_failure_jump.\n"); + /* It doesn't matter what we push for the string here. What + the code at `fail' tests is the value for the pattern. */ + PUSH_FAILURE_POINT (NULL, NULL, -2); + goto unconditional_jump; + + + /* At the end of an alternative, we need to push a dummy failure + point in case we are followed by a `pop_failure_jump', because + we don't want the failure point for the alternative to be + popped. For example, matching `(a|ab)*' against `aab' + requires that we match the `ab' alternative. */ + case push_dummy_failure: + DEBUG_PRINT1 ("EXECUTING push_dummy_failure.\n"); + /* See comments just above at `dummy_failure_jump' about the + two zeroes. */ + PUSH_FAILURE_POINT (NULL, NULL, -2); + break; + + /* Have to succeed matching what follows at least n times. + After that, handle like `on_failure_jump'. */ + case succeed_n: + EXTRACT_NUMBER (mcnt, p + OFFSET_ADDRESS_SIZE); + DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); + + assert (mcnt >= 0); + /* Originally, this is how many times we HAVE to succeed. */ + if (mcnt > 0) + { + mcnt--; + p += OFFSET_ADDRESS_SIZE; + STORE_NUMBER_AND_INCR (p, mcnt); +#ifdef _LIBC + DEBUG_PRINT3 (" Setting %p to %d.\n", p - OFFSET_ADDRESS_SIZE + , mcnt); +#else + DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p - OFFSET_ADDRESS_SIZE + , mcnt); +#endif + } + else if (mcnt == 0) + { +#ifdef _LIBC + DEBUG_PRINT2 (" Setting two bytes from %p to no_op.\n", + p + OFFSET_ADDRESS_SIZE); +#else + DEBUG_PRINT2 (" Setting two bytes from 0x%x to no_op.\n", + p + OFFSET_ADDRESS_SIZE); +#endif /* _LIBC */ + +#ifdef WCHAR + p[1] = (UCHAR_T) no_op; +#else + p[2] = (UCHAR_T) no_op; + p[3] = (UCHAR_T) no_op; +#endif /* WCHAR */ + goto on_failure; + } + break; + + case jump_n: + EXTRACT_NUMBER (mcnt, p + OFFSET_ADDRESS_SIZE); + DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); + + /* Originally, this is how many times we CAN jump. */ + if (mcnt) + { + mcnt--; + STORE_NUMBER (p + OFFSET_ADDRESS_SIZE, mcnt); + +#ifdef _LIBC + DEBUG_PRINT3 (" Setting %p to %d.\n", p + OFFSET_ADDRESS_SIZE, + mcnt); +#else + DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p + OFFSET_ADDRESS_SIZE, + mcnt); +#endif /* _LIBC */ + goto unconditional_jump; + } + /* If don't have to jump any more, skip over the rest of command. */ + else + p += 2 * OFFSET_ADDRESS_SIZE; + break; + + case set_number_at: + { + DEBUG_PRINT1 ("EXECUTING set_number_at.\n"); + + EXTRACT_NUMBER_AND_INCR (mcnt, p); + p1 = p + mcnt; + EXTRACT_NUMBER_AND_INCR (mcnt, p); +#ifdef _LIBC + DEBUG_PRINT3 (" Setting %p to %d.\n", p1, mcnt); +#else + DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p1, mcnt); +#endif + STORE_NUMBER (p1, mcnt); + break; + } + +#if 0 + /* The DEC Alpha C compiler 3.x generates incorrect code for the + test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of + AT_WORD_BOUNDARY, so this code is disabled. Expanding the + macro and introducing temporary variables works around the bug. */ + + case wordbound: + DEBUG_PRINT1 ("EXECUTING wordbound.\n"); + if (AT_WORD_BOUNDARY (d)) + break; + goto fail; + + case notwordbound: + DEBUG_PRINT1 ("EXECUTING notwordbound.\n"); + if (AT_WORD_BOUNDARY (d)) + goto fail; + break; +#else + case wordbound: + { + boolean prevchar, thischar; + + DEBUG_PRINT1 ("EXECUTING wordbound.\n"); + if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) + break; + + prevchar = WORDCHAR_P (d - 1); + thischar = WORDCHAR_P (d); + if (prevchar != thischar) + break; + goto fail; + } + + case notwordbound: + { + boolean prevchar, thischar; + + DEBUG_PRINT1 ("EXECUTING notwordbound.\n"); + if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) + goto fail; + + prevchar = WORDCHAR_P (d - 1); + thischar = WORDCHAR_P (d); + if (prevchar != thischar) + goto fail; + break; + } +#endif + + case wordbeg: + DEBUG_PRINT1 ("EXECUTING wordbeg.\n"); + if (!AT_STRINGS_END (d) && WORDCHAR_P (d) + && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1))) + break; + goto fail; + + case wordend: + DEBUG_PRINT1 ("EXECUTING wordend.\n"); + if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1) + && (AT_STRINGS_END (d) || !WORDCHAR_P (d))) + break; + goto fail; + +#ifdef emacs + case before_dot: + DEBUG_PRINT1 ("EXECUTING before_dot.\n"); + if (PTR_CHAR_POS ((unsigned char *) d) >= point) + goto fail; + break; + + case at_dot: + DEBUG_PRINT1 ("EXECUTING at_dot.\n"); + if (PTR_CHAR_POS ((unsigned char *) d) != point) + goto fail; + break; + + case after_dot: + DEBUG_PRINT1 ("EXECUTING after_dot.\n"); + if (PTR_CHAR_POS ((unsigned char *) d) <= point) + goto fail; + break; + + case syntaxspec: + DEBUG_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt); + mcnt = *p++; + goto matchsyntax; + + case wordchar: + DEBUG_PRINT1 ("EXECUTING Emacs wordchar.\n"); + mcnt = (int) Sword; + matchsyntax: + PREFETCH (); + /* Can't use *d++ here; SYNTAX may be an unsafe macro. */ + d++; + if (SYNTAX (d[-1]) != (enum syntaxcode) mcnt) + goto fail; + SET_REGS_MATCHED (); + break; + + case notsyntaxspec: + DEBUG_PRINT2 ("EXECUTING notsyntaxspec %d.\n", mcnt); + mcnt = *p++; + goto matchnotsyntax; + + case notwordchar: + DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.\n"); + mcnt = (int) Sword; + matchnotsyntax: + PREFETCH (); + /* Can't use *d++ here; SYNTAX may be an unsafe macro. */ + d++; + if (SYNTAX (d[-1]) == (enum syntaxcode) mcnt) + goto fail; + SET_REGS_MATCHED (); + break; + +#else /* not emacs */ + case wordchar: + DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.\n"); + PREFETCH (); + if (!WORDCHAR_P (d)) + goto fail; + SET_REGS_MATCHED (); + d++; + break; + + case notwordchar: + DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n"); + PREFETCH (); + if (WORDCHAR_P (d)) + goto fail; + SET_REGS_MATCHED (); + d++; + break; +#endif /* not emacs */ + + default: + abort (); + } + continue; /* Successfully executed one pattern command; keep going. */ + + + /* We goto here if a matching operation fails. */ + fail: + if (!FAIL_STACK_EMPTY ()) + { /* A restart point is known. Restore to that state. */ + DEBUG_PRINT1 ("\nFAIL:\n"); + POP_FAILURE_POINT (d, p, + lowest_active_reg, highest_active_reg, + regstart, regend, reg_info); + + /* If this failure point is a dummy, try the next one. */ + if (!p) + goto fail; + + /* If we failed to the end of the pattern, don't examine *p. */ + assert (p <= pend); + if (p < pend) + { + boolean is_a_jump_n = false; + + /* If failed to a backwards jump that's part of a repetition + loop, need to pop this failure point and use the next one. */ + switch ((re_opcode_t) *p) + { + case jump_n: + is_a_jump_n = true; + case maybe_pop_jump: + case pop_failure_jump: + case jump: + p1 = p + 1; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + p1 += mcnt; + + if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n) + || (!is_a_jump_n + && (re_opcode_t) *p1 == on_failure_jump)) + goto fail; + break; + default: + /* do nothing */ ; + } + } + + if (d >= string1 && d <= end1) + dend = end_match_1; + } + else + break; /* Matching at this starting point really fails. */ + } /* for (;;) */ + + if (best_regs_set) + goto restore_best_regs; + + FREE_VARIABLES (); + + return -1; /* Failure to match. */ +} /* re_match_2 */ + +/* Subroutine definitions for re_match_2. */ + + +/* We are passed P pointing to a register number after a start_memory. + + Return true if the pattern up to the corresponding stop_memory can + match the empty string, and false otherwise. + + If we find the matching stop_memory, sets P to point to one past its number. + Otherwise, sets P to an undefined byte less than or equal to END. + + We don't handle duplicates properly (yet). */ + +static boolean +PREFIX(group_match_null_string_p) (p, end, reg_info) + UCHAR_T **p, *end; + PREFIX(register_info_type) *reg_info; +{ + int mcnt; + /* Point to after the args to the start_memory. */ + UCHAR_T *p1 = *p + 2; + + while (p1 < end) + { + /* Skip over opcodes that can match nothing, and return true or + false, as appropriate, when we get to one that can't, or to the + matching stop_memory. */ + + switch ((re_opcode_t) *p1) + { + /* Could be either a loop or a series of alternatives. */ + case on_failure_jump: + p1++; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + + /* If the next operation is not a jump backwards in the + pattern. */ + + if (mcnt >= 0) + { + /* Go through the on_failure_jumps of the alternatives, + seeing if any of the alternatives cannot match nothing. + The last alternative starts with only a jump, + whereas the rest start with on_failure_jump and end + with a jump, e.g., here is the pattern for `a|b|c': + + /on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6 + /on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3 + /exactn/1/c + + So, we have to first go through the first (n-1) + alternatives and then deal with the last one separately. */ + + + /* Deal with the first (n-1) alternatives, which start + with an on_failure_jump (see above) that jumps to right + past a jump_past_alt. */ + + while ((re_opcode_t) p1[mcnt-(1+OFFSET_ADDRESS_SIZE)] == + jump_past_alt) + { + /* `mcnt' holds how many bytes long the alternative + is, including the ending `jump_past_alt' and + its number. */ + + if (!PREFIX(alt_match_null_string_p) (p1, p1 + mcnt - + (1 + OFFSET_ADDRESS_SIZE), + reg_info)) + return false; + + /* Move to right after this alternative, including the + jump_past_alt. */ + p1 += mcnt; + + /* Break if it's the beginning of an n-th alternative + that doesn't begin with an on_failure_jump. */ + if ((re_opcode_t) *p1 != on_failure_jump) + break; + + /* Still have to check that it's not an n-th + alternative that starts with an on_failure_jump. */ + p1++; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + if ((re_opcode_t) p1[mcnt-(1+OFFSET_ADDRESS_SIZE)] != + jump_past_alt) + { + /* Get to the beginning of the n-th alternative. */ + p1 -= 1 + OFFSET_ADDRESS_SIZE; + break; + } + } + + /* Deal with the last alternative: go back and get number + of the `jump_past_alt' just before it. `mcnt' contains + the length of the alternative. */ + EXTRACT_NUMBER (mcnt, p1 - OFFSET_ADDRESS_SIZE); + + if (!PREFIX(alt_match_null_string_p) (p1, p1 + mcnt, reg_info)) + return false; + + p1 += mcnt; /* Get past the n-th alternative. */ + } /* if mcnt > 0 */ + break; + + + case stop_memory: + assert (p1[1] == **p); + *p = p1 + 2; + return true; + + + default: + if (!PREFIX(common_op_match_null_string_p) (&p1, end, reg_info)) + return false; + } + } /* while p1 < end */ + + return false; +} /* group_match_null_string_p */ + + +/* Similar to group_match_null_string_p, but doesn't deal with alternatives: + It expects P to be the first byte of a single alternative and END one + byte past the last. The alternative can contain groups. */ + +static boolean +PREFIX(alt_match_null_string_p) (p, end, reg_info) + UCHAR_T *p, *end; + PREFIX(register_info_type) *reg_info; +{ + int mcnt; + UCHAR_T *p1 = p; + + while (p1 < end) + { + /* Skip over opcodes that can match nothing, and break when we get + to one that can't. */ + + switch ((re_opcode_t) *p1) + { + /* It's a loop. */ + case on_failure_jump: + p1++; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + p1 += mcnt; + break; + + default: + if (!PREFIX(common_op_match_null_string_p) (&p1, end, reg_info)) + return false; + } + } /* while p1 < end */ + + return true; +} /* alt_match_null_string_p */ + + +/* Deals with the ops common to group_match_null_string_p and + alt_match_null_string_p. + + Sets P to one after the op and its arguments, if any. */ + +static boolean +PREFIX(common_op_match_null_string_p) (p, end, reg_info) + UCHAR_T **p, *end; + PREFIX(register_info_type) *reg_info; +{ + int mcnt; + boolean ret; + int reg_no; + UCHAR_T *p1 = *p; + + switch ((re_opcode_t) *p1++) + { + case no_op: + case begline: + case endline: + case begbuf: + case endbuf: + case wordbeg: + case wordend: + case wordbound: + case notwordbound: +#ifdef emacs + case before_dot: + case at_dot: + case after_dot: +#endif + break; + + case start_memory: + reg_no = *p1; + assert (reg_no > 0 && reg_no <= MAX_REGNUM); + ret = PREFIX(group_match_null_string_p) (&p1, end, reg_info); + + /* Have to set this here in case we're checking a group which + contains a group and a back reference to it. */ + + if (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE) + REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret; + + if (!ret) + return false; + break; + + /* If this is an optimized succeed_n for zero times, make the jump. */ + case jump: + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + if (mcnt >= 0) + p1 += mcnt; + else + return false; + break; + + case succeed_n: + /* Get to the number of times to succeed. */ + p1 += OFFSET_ADDRESS_SIZE; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + + if (mcnt == 0) + { + p1 -= 2 * OFFSET_ADDRESS_SIZE; + EXTRACT_NUMBER_AND_INCR (mcnt, p1); + p1 += mcnt; + } + else + return false; + break; + + case duplicate: + if (!REG_MATCH_NULL_STRING_P (reg_info[*p1])) + return false; + break; + + case set_number_at: + p1 += 2 * OFFSET_ADDRESS_SIZE; + + default: + /* All other opcodes mean we cannot match the empty string. */ + return false; + } + + *p = p1; + return true; +} /* common_op_match_null_string_p */ + + +/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN + bytes; nonzero otherwise. */ + +static int +PREFIX(bcmp_translate) (s1, s2, len, translate) + const CHAR_T *s1, *s2; + register int len; + RE_TRANSLATE_TYPE translate; +{ + register const UCHAR_T *p1 = (const UCHAR_T *) s1; + register const UCHAR_T *p2 = (const UCHAR_T *) s2; + while (len) + { +#ifdef WCHAR + if (((*p1<=0xff)?translate[*p1++]:*p1++) + != ((*p2<=0xff)?translate[*p2++]:*p2++)) + return 1; +#else /* BYTE */ + if (translate[*p1++] != translate[*p2++]) return 1; +#endif /* WCHAR */ + len--; + } + return 0; +} + + +#else /* not INSIDE_RECURSION */ + +/* Entry points for GNU code. */ + +/* re_compile_pattern is the GNU regular expression compiler: it + compiles PATTERN (of length SIZE) and puts the result in BUFP. + Returns 0 if the pattern was valid, otherwise an error string. + + Assumes the `allocated' (and perhaps `buffer') and `translate' fields + are set in BUFP on entry. + + We call regex_compile to do the actual compilation. */ + +const char * +re_compile_pattern (pattern, length, bufp) + const char *pattern; + size_t length; + struct re_pattern_buffer *bufp; +{ + reg_errcode_t ret; + + /* GNU code is written to assume at least RE_NREGS registers will be set + (and at least one extra will be -1). */ + bufp->regs_allocated = REGS_UNALLOCATED; + + /* And GNU code determines whether or not to get register information + by passing null for the REGS argument to re_match, etc., not by + setting no_sub. */ + bufp->no_sub = 0; + + /* Match anchors at newline. */ + bufp->newline_anchor = 1; + +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + ret = wcs_regex_compile (pattern, length, re_syntax_options, bufp); + else +# endif + ret = byte_regex_compile (pattern, length, re_syntax_options, bufp); + + if (!ret) + return NULL; + return gettext (re_error_msgid + re_error_msgid_idx[(int) ret]); +} +#ifdef _LIBC +weak_alias (__re_compile_pattern, re_compile_pattern) +#endif + +/* Entry points compatible with 4.2 BSD regex library. We don't define + them unless specifically requested. */ + +#if defined _REGEX_RE_COMP || defined _LIBC + +/* BSD has one and only one pattern buffer. */ +static struct re_pattern_buffer re_comp_buf; + +char * +#ifdef _LIBC +/* Make these definitions weak in libc, so POSIX programs can redefine + these names if they don't use our functions, and still use + regcomp/regexec below without link errors. */ +weak_function +#endif +re_comp (s) + const char *s; +{ + reg_errcode_t ret; + + if (!s) + { + if (!re_comp_buf.buffer) + return gettext ("No previous regular expression"); + return 0; + } + + if (!re_comp_buf.buffer) + { + re_comp_buf.buffer = (unsigned char *) malloc (200); + if (re_comp_buf.buffer == NULL) + return (char *) gettext (re_error_msgid + + re_error_msgid_idx[(int) REG_ESPACE]); + re_comp_buf.allocated = 200; + + re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH); + if (re_comp_buf.fastmap == NULL) + return (char *) gettext (re_error_msgid + + re_error_msgid_idx[(int) REG_ESPACE]); + } + + /* Since `re_exec' always passes NULL for the `regs' argument, we + don't need to initialize the pattern buffer fields which affect it. */ + + /* Match anchors at newlines. */ + re_comp_buf.newline_anchor = 1; + +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + ret = wcs_regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); + else +# endif + ret = byte_regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); + + if (!ret) + return NULL; + + /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ + return (char *) gettext (re_error_msgid + re_error_msgid_idx[(int) ret]); +} + + +int +#ifdef _LIBC +weak_function +#endif +re_exec (s) + const char *s; +{ + const int len = strlen (s); + return + 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0); +} + +#endif /* _REGEX_RE_COMP */ + +/* POSIX.2 functions. Don't define these for Emacs. */ + +#ifndef emacs + +/* regcomp takes a regular expression as a string and compiles it. + + PREG is a regex_t *. We do not expect any fields to be initialized, + since POSIX says we shouldn't. Thus, we set + + `buffer' to the compiled pattern; + `used' to the length of the compiled pattern; + `syntax' to RE_SYNTAX_POSIX_EXTENDED if the + REG_EXTENDED bit in CFLAGS is set; otherwise, to + RE_SYNTAX_POSIX_BASIC; + `newline_anchor' to REG_NEWLINE being set in CFLAGS; + `fastmap' to an allocated space for the fastmap; + `fastmap_accurate' to zero; + `re_nsub' to the number of subexpressions in PATTERN. + + PATTERN is the address of the pattern string. + + CFLAGS is a series of bits which affect compilation. + + If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we + use POSIX basic syntax. + + If REG_NEWLINE is set, then . and [^...] don't match newline. + Also, regexec will try a match beginning after every newline. + + If REG_ICASE is set, then we considers upper- and lowercase + versions of letters to be equivalent when matching. + + If REG_NOSUB is set, then when PREG is passed to regexec, that + routine will report only success or failure, and nothing about the + registers. + + It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for + the return codes and their meanings.) */ + +int +regcomp (preg, pattern, cflags) + regex_t *preg; + const char *pattern; + int cflags; +{ + reg_errcode_t ret; + reg_syntax_t syntax + = (cflags & REG_EXTENDED) ? + RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; + + /* regex_compile will allocate the space for the compiled pattern. */ + preg->buffer = 0; + preg->allocated = 0; + preg->used = 0; + + /* Try to allocate space for the fastmap. */ + preg->fastmap = (char *) malloc (1 << BYTEWIDTH); + + if (cflags & REG_ICASE) + { + unsigned i; + + preg->translate + = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE + * sizeof (*(RE_TRANSLATE_TYPE)0)); + if (preg->translate == NULL) + return (int) REG_ESPACE; + + /* Map uppercase characters to corresponding lowercase ones. */ + for (i = 0; i < CHAR_SET_SIZE; i++) + preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; + } + else + preg->translate = NULL; + + /* If REG_NEWLINE is set, newlines are treated differently. */ + if (cflags & REG_NEWLINE) + { /* REG_NEWLINE implies neither . nor [^...] match newline. */ + syntax &= ~RE_DOT_NEWLINE; + syntax |= RE_HAT_LISTS_NOT_NEWLINE; + /* It also changes the matching behavior. */ + preg->newline_anchor = 1; + } + else + preg->newline_anchor = 0; + + preg->no_sub = !!(cflags & REG_NOSUB); + + /* POSIX says a null character in the pattern terminates it, so we + can use strlen here in compiling the pattern. */ +# ifdef MBS_SUPPORT + if (MB_CUR_MAX != 1) + ret = wcs_regex_compile (pattern, strlen (pattern), syntax, preg); + else +# endif + ret = byte_regex_compile (pattern, strlen (pattern), syntax, preg); + + /* POSIX doesn't distinguish between an unmatched open-group and an + unmatched close-group: both are REG_EPAREN. */ + if (ret == REG_ERPAREN) ret = REG_EPAREN; + + if (ret == REG_NOERROR && preg->fastmap) + { + /* Compute the fastmap now, since regexec cannot modify the pattern + buffer. */ + if (re_compile_fastmap (preg) == -2) + { + /* Some error occurred while computing the fastmap, just forget + about it. */ + free (preg->fastmap); + preg->fastmap = NULL; + } + } + + return (int) ret; +} +#ifdef _LIBC +weak_alias (__regcomp, regcomp) +#endif + + +/* regexec searches for a given pattern, specified by PREG, in the + string STRING. + + If NMATCH is zero or REG_NOSUB was set in the cflags argument to + `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at + least NMATCH elements, and we set them to the offsets of the + corresponding matched substrings. + + EFLAGS specifies `execution flags' which affect matching: if + REG_NOTBOL is set, then ^ does not match at the beginning of the + string; if REG_NOTEOL is set, then $ does not match at the end. + + We return 0 if we find a match and REG_NOMATCH if not. */ + +int +regexec (preg, string, nmatch, pmatch, eflags) + const regex_t *preg; + const char *string; + size_t nmatch; + regmatch_t pmatch[]; + int eflags; +{ + int ret; + struct re_registers regs; + regex_t private_preg; + int len = strlen (string); + boolean want_reg_info = !preg->no_sub && nmatch > 0; + + private_preg = *preg; + + private_preg.not_bol = !!(eflags & REG_NOTBOL); + private_preg.not_eol = !!(eflags & REG_NOTEOL); + + /* The user has told us exactly how many registers to return + information about, via `nmatch'. We have to pass that on to the + matching routines. */ + private_preg.regs_allocated = REGS_FIXED; + + if (want_reg_info) + { + regs.num_regs = nmatch; + regs.start = TALLOC (nmatch * 2, regoff_t); + if (regs.start == NULL) + return (int) REG_NOMATCH; + regs.end = regs.start + nmatch; + } + + /* Perform the searching operation. */ + ret = re_search (&private_preg, string, len, + /* start: */ 0, /* range: */ len, + want_reg_info ? ®s : (struct re_registers *) 0); + + /* Copy the register information to the POSIX structure. */ + if (want_reg_info) + { + if (ret >= 0) + { + unsigned r; + + for (r = 0; r < nmatch; r++) + { + pmatch[r].rm_so = regs.start[r]; + pmatch[r].rm_eo = regs.end[r]; + } + } + + /* If we needed the temporary register info, free the space now. */ + free (regs.start); + } + + /* We want zero return to mean success, unlike `re_search'. */ + return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; +} +#ifdef _LIBC +weak_alias (__regexec, regexec) +#endif + + +/* Returns a message corresponding to an error code, ERRCODE, returned + from either regcomp or regexec. We don't use PREG here. */ + +size_t +regerror (errcode, preg, errbuf, errbuf_size) + int errcode; + const regex_t *preg; + char *errbuf; + size_t errbuf_size; +{ + const char *msg; + size_t msg_size; + + if (errcode < 0 + || errcode >= (int) (sizeof (re_error_msgid_idx) + / sizeof (re_error_msgid_idx[0]))) + /* Only error codes returned by the rest of the code should be passed + to this routine. If we are given anything else, or if other regex + code generates an invalid error code, then the program has a bug. + Dump core so we can fix it. */ + abort (); + + msg = gettext (re_error_msgid + re_error_msgid_idx[errcode]); + + msg_size = strlen (msg) + 1; /* Includes the null. */ + + if (errbuf_size != 0) + { + if (msg_size > errbuf_size) + { +#if defined HAVE_MEMPCPY || defined _LIBC + *((char *) __mempcpy (errbuf, msg, errbuf_size - 1)) = '\0'; +#else + memcpy (errbuf, msg, errbuf_size - 1); + errbuf[errbuf_size - 1] = 0; +#endif + } + else + memcpy (errbuf, msg, msg_size); + } + + return msg_size; +} +#ifdef _LIBC +weak_alias (__regerror, regerror) +#endif + + +/* Free dynamically allocated space used by PREG. */ + +void +regfree (preg) + regex_t *preg; +{ + if (preg->buffer != NULL) + free (preg->buffer); + preg->buffer = NULL; + + preg->allocated = 0; + preg->used = 0; + + if (preg->fastmap != NULL) + free (preg->fastmap); + preg->fastmap = NULL; + preg->fastmap_accurate = 0; + + if (preg->translate != NULL) + free (preg->translate); + preg->translate = NULL; +} +#ifdef _LIBC +weak_alias (__regfree, regfree) +#endif + +#endif /* not emacs */ + +#endif /* not INSIDE_RECURSION */ + + +#undef STORE_NUMBER +#undef STORE_NUMBER_AND_INCR +#undef EXTRACT_NUMBER +#undef EXTRACT_NUMBER_AND_INCR + +#undef DEBUG_PRINT_COMPILED_PATTERN +#undef DEBUG_PRINT_DOUBLE_STRING + +#undef INIT_FAIL_STACK +#undef RESET_FAIL_STACK +#undef DOUBLE_FAIL_STACK +#undef PUSH_PATTERN_OP +#undef PUSH_FAILURE_POINTER +#undef PUSH_FAILURE_INT +#undef PUSH_FAILURE_ELT +#undef POP_FAILURE_POINTER +#undef POP_FAILURE_INT +#undef POP_FAILURE_ELT +#undef DEBUG_PUSH +#undef DEBUG_POP +#undef PUSH_FAILURE_POINT +#undef POP_FAILURE_POINT + +#undef REG_UNSET_VALUE +#undef REG_UNSET + +#undef PATFETCH +#undef PATFETCH_RAW +#undef PATUNFETCH +#undef TRANSLATE + +#undef INIT_BUF_SIZE +#undef GET_BUFFER_SPACE +#undef BUF_PUSH +#undef BUF_PUSH_2 +#undef BUF_PUSH_3 +#undef STORE_JUMP +#undef STORE_JUMP2 +#undef INSERT_JUMP +#undef INSERT_JUMP2 +#undef EXTEND_BUFFER +#undef GET_UNSIGNED_NUMBER +#undef FREE_STACK_RETURN + +# undef POINTER_TO_OFFSET +# undef MATCHING_IN_FRST_STRING +# undef PREFETCH +# undef AT_STRINGS_BEG +# undef AT_STRINGS_END +# undef WORDCHAR_P +# undef FREE_VAR +# undef FREE_VARIABLES +# undef NO_HIGHEST_ACTIVE_REG +# undef NO_LOWEST_ACTIVE_REG + +# undef CHAR_T +# undef UCHAR_T +# undef COMPILED_BUFFER_VAR +# undef OFFSET_ADDRESS_SIZE +# undef CHAR_CLASS_SIZE +# undef PREFIX +# undef ARG_PREFIX +# undef PUT_CHAR +# undef BYTE +# undef WCHAR + +# define DEFINED_ONCE diff -uprN binutils-2.11.90.0.23/opcodes/ChangeLog binutils-2.11.90.0.24/opcodes/ChangeLog --- binutils-2.11.90.0.23/opcodes/ChangeLog Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/opcodes/ChangeLog Sat Jul 14 10:29:16 2001 @@ -1,3 +1,26 @@ +2001-07-12 Jeff Johnston + + * cgen-asm.in: Include "xregex.h" always to enable the libiberty + regex support. + (@arch@_cgen_build_insn_regex): New routine from Graydon. + (@arch@_cgen_assemble_insn): Add Graydon's code to use regex + to verify if it is worth parsing the insn as insn "x". Also update + error message when insn is not a recognized format of the insn vs + when the insn is completely unrecognized. + +2001-07-11 Frank Ch. Eigler + + * cgen-dis.in (print_insn): Use cgen_get_insn_value instead of + bfd_get_bits. + * cgen-opc.c (cgen_get_insn_value, cgen_put_insn_value): Respect + non-zero CGEN_CPU_DESC->insn_chunk_bitsize. + +2001-07-09 Andreas Jaeger , Karsten Keil + + * i386-dis.c (set_op): Handle 64 bit and 32 bit mode. + (OP_J): Use bfd_vma for mask to work properly with 64 bits. + (op_address,op_riprel): Use bfd_vma to handle 64 bits. + 2001-07-05 Ben Elliston * Makefile.am (CPUDIR): Define. diff -uprN binutils-2.11.90.0.23/opcodes/cgen-asm.in binutils-2.11.90.0.24/opcodes/cgen-asm.in --- binutils-2.11.90.0.23/opcodes/cgen-asm.in Fri Mar 23 13:52:17 2001 +++ binutils-2.11.90.0.24/opcodes/cgen-asm.in Sat Jul 14 10:29:16 2001 @@ -34,6 +34,7 @@ along with this program; if not, write t #include "@prefix@-desc.h" #include "@prefix@-opc.h" #include "opintl.h" +#include "xregex.h" #undef min #define min(a,b) ((a) < (b) ? (a) : (b)) @@ -45,6 +46,104 @@ static const char * parse_insn_normal /* -- assembler routines inserted here */ + +/* + Regex construction routine. + + This translates an opcode syntax string into a regex string, + by replacing any non-character syntax element (such as an + opcode) with the pattern '.*' + + It then compiles the regex and stores it in the opcode, for + later use by @arch@_cgen_assemble_insn + + returns NULL for success, an error message for failure +*/ + +char * +@arch@_cgen_build_insn_regex (insn) + CGEN_INSN *insn; +{ + CGEN_OPCODE *opc = CGEN_INSN_OPCODE (insn); + const char *mnem = CGEN_INSN_MNEMONIC (insn); + int mnem_len; + char rxbuf[CGEN_MAX_RX_ELEMENTS]; + char *rx = rxbuf; + const CGEN_SYNTAX_CHAR_TYPE *syn; + int reg_err; + + syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); + + /* Mnemonics come first in the syntax string */ + if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) return "missing mnemonic in syntax string"; + ++syn; + + /* copy the literal mnemonic out of the insn */ + memset (rx, 0, CGEN_MAX_RX_ELEMENTS); + mnem_len = strlen(mnem); + memcpy (rx, mnem, mnem_len); + rx += mnem_len; + + /* copy any remaining literals from the syntax string into the rx */ + for(; * syn != 0 && rx < rxbuf + (CGEN_MAX_RX_ELEMENTS - 9); ++syn, ++rx) + { + if (CGEN_SYNTAX_CHAR_P (* syn)) + { + char tmp = CGEN_SYNTAX_CHAR (* syn); + switch (tmp) + { + /* escape any regex metacharacters in the syntax */ + case '.': case '[': case '\\': + case '*': case '^': case '$': + +#ifdef CGEN_ESCAPE_EXTENDED_REGEX + case '?': case '{': case '}': + case '(': case ')': case '*': + case '|': case '+': case ']': +#endif + + * rx++ = '\\'; + break; + } + /* insert syntax char into rx */ + * rx = tmp; + } + else + { + /* replace non-syntax fields with globs */ + * rx = '.'; + * ++rx = '*'; + } + } + + /* trailing whitespace ok */ + * rx++ = '['; + * rx++ = ' '; + * rx++ = '\t'; + * rx++ = ']'; + * rx++ = '*'; + + /* but anchor it after that */ + * rx++ = '$'; + * rx = '\0'; + + CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); + reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB|REG_ICASE); + + if (reg_err == 0) + return NULL; + else + { + static char msg[80]; + regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); + regfree ((regex_t *) CGEN_INSN_RX (insn)); + free (CGEN_INSN_RX (insn)); + (CGEN_INSN_RX (insn)) = NULL; + return msg; + } +} + + /* Default insn parser. The syntax string is scanned and operands are parsed and stored in FIELDS. @@ -211,6 +310,7 @@ const CGEN_INSN * CGEN_INSN_LIST *ilist; const char *parse_errmsg = NULL; const char *insert_errmsg = NULL; + int recognized_mnemonic = 0; /* Skip leading white space. */ while (isspace (* str)) @@ -226,6 +326,7 @@ const CGEN_INSN * for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) { const CGEN_INSN *insn = ilist->insn; + recognized_mnemonic = 1; #ifdef CGEN_VALIDATE_INSN_SUPPORTED /* not usually needed as unsupported opcodes shouldn't be in the hash lists */ @@ -242,6 +343,11 @@ const CGEN_INSN * str = start; + /* skip this insn if str doesn't look right lexically */ + if (CGEN_INSN_RX (insn) != NULL && + regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) + continue; + /* Allow parse/insert handlers to obtain length of insn. */ CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); @@ -269,6 +375,7 @@ const CGEN_INSN * Failing that, use parse_errmsg */ tmp_errmsg = (insert_errmsg ? insert_errmsg : parse_errmsg ? parse_errmsg : + recognized_mnemonic ? _("unrecognized form of instruction") : _("unrecognized instruction")); if (strlen (start) > 50) diff -uprN binutils-2.11.90.0.23/opcodes/cgen-dis.in binutils-2.11.90.0.24/opcodes/cgen-dis.in --- binutils-2.11.90.0.23/opcodes/cgen-dis.in Mon May 21 10:34:08 2001 +++ binutils-2.11.90.0.24/opcodes/cgen-dis.in Sat Jul 14 10:29:16 2001 @@ -229,12 +229,12 @@ print_insn (cd, pc, info, buf, buflen) char *buf; int buflen; { - unsigned long insn_value; + CGEN_INSN_INT insn_value; const CGEN_INSN_LIST *insn_list; CGEN_EXTRACT_INFO ex_info; /* Extract base part of instruction, just in case CGEN_DIS_* uses it. */ - insn_value = bfd_get_bits (buf, buflen * 8, info->endian == BFD_ENDIAN_BIG); + insn_value = cgen_get_insn_value (cd, buf, buflen * 8); /* Fill in ex_info fields like read_insn would. Don't actually call read_insn, since the incoming buffer is already read (and possibly diff -uprN binutils-2.11.90.0.23/opcodes/cgen-opc.c binutils-2.11.90.0.24/opcodes/cgen-opc.c --- binutils-2.11.90.0.23/opcodes/cgen-opc.c Fri Jul 6 09:26:11 2001 +++ binutils-2.11.90.0.24/opcodes/cgen-opc.c Sat Jul 14 10:29:16 2001 @@ -391,7 +391,35 @@ cgen_get_insn_value (cd, buf, length) unsigned char *buf; int length; { - return bfd_get_bits (buf, length, cd->insn_endian == CGEN_ENDIAN_BIG); + int big_p = (cd->insn_endian == CGEN_ENDIAN_BIG); + int insn_chunk_bitsize = cd->insn_chunk_bitsize; + CGEN_INSN_INT value = 0; + + if (insn_chunk_bitsize != 0 && insn_chunk_bitsize < length) + { + /* We need to divide up the incoming value into insn_chunk_bitsize-length + segments, and endian-convert them, one at a time. */ + int i; + + /* Enforce divisibility. */ + if ((length % insn_chunk_bitsize) != 0) + abort (); + + for (i = 0; i < length; i += insn_chunk_bitsize) /* NB: i == bits */ + { + int index; + bfd_vma this_value; + index = i; /* NB: not dependent on endianness; opposite of cgen_put_insn_value! */ + this_value = bfd_get_bits (& buf[index / 8], insn_chunk_bitsize, big_p); + value = (value << insn_chunk_bitsize) | this_value; + } + } + else + { + value = bfd_get_bits (buf, length, cd->insn_endian == CGEN_ENDIAN_BIG); + } + + return value; } /* Cover function to store an insn value properly byteswapped. */ @@ -403,8 +431,31 @@ cgen_put_insn_value (cd, buf, length, va int length; CGEN_INSN_INT value; { - bfd_put_bits ((bfd_vma) value, buf, length, - cd->insn_endian == CGEN_ENDIAN_BIG); + int big_p = (cd->insn_endian == CGEN_ENDIAN_BIG); + int insn_chunk_bitsize = cd->insn_chunk_bitsize; + + if (insn_chunk_bitsize != 0 && insn_chunk_bitsize < length) + { + /* We need to divide up the incoming value into insn_chunk_bitsize-length + segments, and endian-convert them, one at a time. */ + int i; + + /* Enforce divisibility. */ + if ((length % insn_chunk_bitsize) != 0) + abort (); + + for (i = 0; i < length; i += insn_chunk_bitsize) /* NB: i == bits */ + { + int index; + index = (length - insn_chunk_bitsize - i); /* NB: not dependent on endianness! */ + bfd_put_bits ((bfd_vma) value, & buf[index / 8], insn_chunk_bitsize, big_p); + value >>= insn_chunk_bitsize; + } + } + else + { + bfd_put_bits ((bfd_vma) value, buf, length, big_p); + } } /* Look up instruction INSN_*_VALUE and extract its fields. diff -uprN binutils-2.11.90.0.23/opcodes/i386-dis.c binutils-2.11.90.0.24/opcodes/i386-dis.c --- binutils-2.11.90.0.23/opcodes/i386-dis.c Tue Jun 19 11:57:59 2001 +++ binutils-2.11.90.0.24/opcodes/i386-dis.c Sat Jul 14 10:29:16 2001 @@ -67,7 +67,7 @@ static bfd_vma get64 PARAMS ((void)); static bfd_signed_vma get32 PARAMS ((void)); static bfd_signed_vma get32s PARAMS ((void)); static int get16 PARAMS ((void)); -static void set_op PARAMS ((unsigned int, int)); +static void set_op PARAMS ((bfd_vma, int)); static void OP_REG PARAMS ((int, int)); static void OP_IMREG PARAMS ((int, int)); static void OP_I PARAMS ((int, int)); @@ -1786,8 +1786,8 @@ prefix_name (pref, sizeflag) static char op1out[100], op2out[100], op3out[100]; static int op_ad, op_index[3]; -static unsigned int op_address[3]; -static unsigned int op_riprel[3]; +static bfd_vma op_address[3]; +static bfd_vma op_riprel[3]; static bfd_vma start_pc; @@ -3238,12 +3238,21 @@ get16 () static void set_op (op, riprel) - unsigned int op; + bfd_vma op; int riprel; { op_index[op_ad] = op_ad; - op_address[op_ad] = op; - op_riprel[op_ad] = riprel; + if (mode_64bit) + { + op_address[op_ad] = op; + op_riprel[op_ad] = riprel; + } + else + { + /* Mask to get a 32-bit address. */ + op_address[op_ad] = op & 0xffffffff; + op_riprel[op_ad] = riprel & 0xffffffff; + } } static void @@ -3515,7 +3524,7 @@ OP_J (bytemode, sizeflag) int sizeflag; { bfd_vma disp; - int mask = -1; + bfd_vma mask = -1; switch (bytemode) {