Re: build/894: GDB 5.3 incompatibilities with coreutils 4.5.4 in POSIX-2001 mode

[Michael Elizabeth Chastain <mec at shout dot net>]

The following reply was made to PR build/894; it has been noted by GNATS.

From: Michael Elizabeth Chastain <mec at shout dot net>
To: gdb-gnats at sources dot redhat dot com
Cc:  
Subject: Re: build/894: GDB 5.3 incompatibilities with coreutils 4.5.4 in POSIX-2001 mode
Date: Tue, 25 Feb 2003 18:03:26 -0600

 I'm going to suspend this PR.
 
 The affected files are:
 
   ltcf-c.sh
   ltcf-gcj.sh
   ltmain.sh
 
 These files are imported from gcc.
 
 In the gcc repository, these files are created by libtool.
 
 gcc still uses a branched version of libtool, the
 'multi-language-branch'.  Unfortunately for gcc, the libtool people
 declared the multi-language-branch version of libtool dead as of
 2001-06-24.
 
 So ... I can't propagate changes upstream from gcc to libtool because
 the upstream libtool version is a dead version.  The gcc versions of
 these files are orphaned.
 
 This problem will be fixed eventually when gcc switches from the dead
 multi-language-branch version of libtool to a released version of
 libtool or something cut from libtool HEAD -- some version of libtool
 that the libtool people will accept patches for.
 
 For more background information, see:
 
   Alexandre Oliva - On libtool patches for 3.0.1
   http://gcc.gnu.org/ml/gcc/2001-07/msg00918.html
 
 The next step is to make the POSIX correctness changes in the current
 development version of libtool.  I see in the libtool ChangeLogs that
 Paul Eggert has done that (ChangeLog entries of 2002-11-19 and
 2002-05-06).  Good job.
 
 After that, we wait for gcc to switch from multi-language-branch libtool
 to a current version of libtool.  That will take indefinite time.  And
 then gdb can copy gcc's libtool files as usual, and I can re-test with
 posix-strict textutils and/or coreutils, per Paul's directions.
 
 The alternative would be: push gcc to make local changes to their
 orphaned files.  I've done that before, in a case where ltmain.sh was
 crapping out on hppa-hp-hpux11.00.  I'm willing to do it again for
 issues where actual platforms are actually failing.  But for this issue,
 I prefer to wait for gcc to switch to the new libtool.  Again I note
 that the posix-dirty code is in host-specific sections, cygwin|mingw and
 rhapsody|darwin, and we're not getting user reports about actual
 failures.
 
 Michael C