Re: patch: contrib/compare_tests

[Paul Eggert]

Andrew Pinski <address@hidden> writes:

> On May 25, 2006, at 12:25 AM, Paul Eggert wrote:
>> Can somebody please reopen bug #27434 and apply its patch to GCC?
>> Unlike the patches proposed for 14251, the 27434 patch shouldn't break
>> builds on older (pre-POSIX) systems.
>
> Are you 100% sure?

For gcc/Makefile.in, I am more sure that the patched version will
work, than the unpatched one.  The unpatched version is clearly buggy
on pre-POSIX systems, since it invokes "tail +16c" and not "tail +17c".

For ltmain.sh and libjava/classpath/ltmain.sh the patch is merely
propagating fixes that I contributed quite some time ago to Libtool.
These fixes have been distributed and exercised extensively on many
platforms.  (I am a bit surprised that GCC does not simply track
Libtool in this area, but that's a different matter.)

The contrib/compare_tests fix, which fixes the bug that started this
thread, uses the same basic idea as the Libtool patch.  Given its
pedigree, it should be quite reliable.

> It is not pre-POSIX system, it is pre new POSIX systems also.

The patched version should work on all the kinds of systems that
you're worried about.  I considered those systems when I wrote the
patch.

> I am wondering why coreutils follow POSIX instead of the psedo-
> standard of what everyone else does?

That issue was discussed previously in this thread; please see
<http://gcc.gnu.org/ml/gcc-patches/2006-05/msg00140.html>, in the text
starting "The remaining problem in this area".