Re: Batch of testsuite patches - Part 1: pathseps

[Akim Demaille]

"Tim Van Holder" <address@hidden> writes:

> 2001-02-11    Tim Van Holder <address@hidden>
> 
>       * tests/aclocal.m4: Fix a few typos.  Add PATH_SEPARATOR
>       to the list of envvars to ignore.
>       * tests/semantics.at: Use ac_path_separator where appropriate.
>       * tests/atconfig.in: Set up PATH for tests; moved from...
>       * tests/atgeneral.m4: ... here.

Hi Tim,

This is good, but I still don't have the right to apply your patch, so
it will remain in the queue until the FSF awakes.  Don't lose it.


> -# Save the environment, but the variables we are allowed to touch.
> +# Save the environment, except for those variables we are allowed to touch.

Wow, thanks!  I'll have to study `but' again :(


> Index: semantics.at

> -path=`echo "1:2:3:4:5:6" | sed -e 's,\([[0-9]]\),'"$pwd"'/path/\1,g'`
> +p="1${ac_path_separator}2${ac_path_separator}3${ac_path_separator}4${ac_pat
> h_separator}5${ac_path_separator}6"

Please, don't cut and paste patches as the result is almost always
wrong, and implies more work for the person who applies the patch.
And actually, patches for patch -p0 are preferred (Index: tests/semantics.at).


> diff -u -r1.8 atconfig.in
> --- atconfig.in       2001/01/16 10:25:58     1.8
> +++ atconfig.in       2001/02/11 19:48:29
> @@ -1,4 +1,5 @@
> -# @configure_input@                                  -*- shell-script -*-
> +# -*- shell-script -*-
> +# @configure_input@
>  # Configurable variable values for building test suites.
>  # Copyright 2000 Free Software Foundation, Inc.
> 
> @@ -58,6 +59,23 @@
>  srcdir='@srcdir@'
>  top_srcdir='@top_srcdir@'
>  AUTOTEST_PATH='@AUTOTEST_PATH@'
> +
> +# Set this to have PATH behave predictably on DJGPP
> +PATH_SEPARATOR="@PATH_SEPARATOR@"
> +
> +# Don't take risks: use absolute path names.
> +at_path=`pwd`
> +at_IFS_save=$IFS
> +IFS="@PATH_SEPARATOR@"
> +for at_dir in $AUTOTEST_PATH $PATH; do
> +  # There might be directories that don't exist, but don't redirect
> +  # builtins' (eg., cd) stderr directly: Ultrix's sh hates that.
> +  at_dir=`(cd "$at_dir" && pwd) 2>/dev/null`
> +  test -n "$at_dir" && at_path="address@hidden@$at_dir"
> +done
> +IFS=$at_IFS_save
> +PATH=$at_path
> +export PATH

Err, I don't like this, atconfig is meant to be only a pool of
variables to set, it is up to atgeneral.m4 to do such things.  Why did
you do this?