This directory contains shell scripts used as tests for Info.  They are
listed in "info/Makefile.am" and will usually be run with "make check".
It should be possible to run a test on its own by running it at the
command line and checking the return value, e.g.

$ t/goto-quoted.sh
$ echo $?
1

shows a test failure.

The tests fall into two categories: tests of non-interactive and
interactive operation.

New tests can be created by copying existing ones.  Each test should
start with the lines

srcdir=${srcdir:-.}
. $srcdir/t/Init-test.inc

to allow running at the command line, or with either an in-source or
out-of-source build with "make check".

Any interactive test should also finish with

cleanup

to cleanup temporary files and spawned processes.  'cleanup' will
exit the script with a status of $retval, which the test should set
to indicate success or failure.

Non interactive tests can run 'info' as $ginfo.  The interactive tests
use the run_ginfo function.

Tests of interactive operation
------------------------------

These use the "pseudotty" program in the info subdirectory, which
creates a pseudo-terminal for the input and output of the program.
This allows the program to happily enter interactive operation (its standard
file descriptors pass the isatty library function) and avoids affecting
the output of the terminal the test was invoked from.  pseudotty reads and
discards all input on its stdin, and passes through any bytes read on 
its control channel into the pseudo-terminal.  It prints the name of the 
pseudoterminal slave device on standard output.

(test script)
 ^  |
 |  `-control----> pseudotty (master) <---> (slave) stdin/stdout ginfo
 `-name of slave-----'

pseudotty will stop running either when killed, or when it exits after 
its control channel is closed.

Since ginfo is reading to and writing from the pseudoterminal slave 
device, if pseudotty exits before ginfo does, ginfo will exit with an 
I/O error.

Many of the tests of interactive operation try to position the cursor 
on a cross-reference by various means, follow the reference, and dump the node
reached to a file.  (It can be compared with a target file in
info/t/node-target, to check that we ended up where we thought we would.)