[patch] AC_SYS_LARGEFILE

[patch] AC_SYS_LARGEFILE

[Jan Kratochvil]

Hi,

other sourceware tree parts already contain AC_SYS_LARGEFILE affecting ABI
compatibility on 32bit hosts.

While I was grepping for `off_t' before found now there is now already an ABI
incompatiblity for `struct stat' also affected by AC_SYS_LARGEFILE which is
declared in gdb/rs6000-nat.c but filled through bfd_stat() in bfd/ .  It does
corrupt the gdb stack.  Non-gdb/ sourceware parts were patched by me before.

For gdb/ the patch is already being tested for some time
	http://cvs.fedora.redhat.com/viewvc/rpms/gdb/devel/gdb-6.8-bz457187-largefile.patch?content-type=text%2Fplain&view=co
so submitting this patch also for gdb/ as a prerequisite to fix another
FSF GDB regression from me:
	Recent separate debug file warning caused Debian regressions
	http://sourceware.org/ml/gdb-patches/2009-11/msg00034.html

The patch was originally introduced to read elf64-i386 Linux kernel core dumps
on PAE systems (>4GB RAM).  But the existing distro testcase is not applicable
for FSF GDB as it depends on GDB configure options not guaranteeable for FSF
GDB.
	http://cvs.fedora.redhat.com/viewvc/rpms/gdb/devel/gdb-6.8-bz457187-largefile-test.patch?content-type=text%2Fplain&view=co

Regression tested {x86_64,x86_64-m32,i686}-fedora12-linux-gnu.


Thanks,
Jan


gdb/
2009-03-02  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* configure.ac: Call AC_SYS_LARGEFILE.
	* config.in: Regenerate.
	* configure: Regenerate.

--- gdb-6.8.50.20090909.orig/gdb/configure.ac	2009-09-09 20:08:04.000000000 +0200
+++ gdb-6.8.50.20090909/gdb/configure.ac	2009-09-09 20:11:04.000000000 +0200
@@ -28,6 +28,7 @@ AM_MAINTAINER_MODE
 AC_PROG_CC
 AC_USE_SYSTEM_EXTENSIONS
 gl_EARLY
+AC_SYS_LARGEFILE
 AM_PROG_CC_STDC
 
 AC_CONFIG_AUX_DIR(..)

Re: [patch] AC_SYS_LARGEFILE

[Joel Brobecker]

> gdb/
> 2009-03-02  Jan Kratochvil  <jan.kratochvil@redhat.com>
> 
> 	* configure.ac: Call AC_SYS_LARGEFILE.
> 	* config.in: Regenerate.
> 	* configure: Regenerate.

Almost OK.

This macro ran a bell, and I found:
http://www.sourceware.org/ml/binutils/2008-11/msg00248.html

Basically, we cannot use AC_SYS_LARGEFILE on Solaris.  The little
piece of code that I wrote to disable it on solaris-native should
work well in this case too.

-- 
Joel

Re: [patch] AC_SYS_LARGEFILE

[Tom Tromey]

>>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:

Joel> This macro ran a bell, and I found:
Joel> http://www.sourceware.org/ml/binutils/2008-11/msg00248.html

Joel> Basically, we cannot use AC_SYS_LARGEFILE on Solaris.  The little
Joel> piece of code that I wrote to disable it on solaris-native should
Joel> work well in this case too.

libiberty uses AC_SYS_LARGEFILE without any Solaris-specific code.  Why
does that work?

If this code is still needed, then let me suggest the more painful but
also nicer approach of a new .m4 file in config/.

Tom

Re: [patch] AC_SYS_LARGEFILE

[Tom Tromey]

>>>>> "Joel" == Tom Tromey <tromey@redhat.com> writes:

Joel> This macro ran a bell, and I found:
Joel> http://www.sourceware.org/ml/binutils/2008-11/msg00248.html

Joel> Basically, we cannot use AC_SYS_LARGEFILE on Solaris.  The little
Joel> piece of code that I wrote to disable it on solaris-native should
Joel> work well in this case too.

Tom> libiberty uses AC_SYS_LARGEFILE without any Solaris-specific code.  Why
Tom> does that work?

Actually, never mind, I think I understand it now.

Tom

Re: [patch] AC_SYS_LARGEFILE

[Jan Kratochvil]

On Tue, 03 Nov 2009 17:54:47 +0100, Tom Tromey wrote:
> >>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:
> 
> Joel> This macro ran a bell, and I found:
> Joel> http://www.sourceware.org/ml/binutils/2008-11/msg00248.html
> 
> Joel> Basically, we cannot use AC_SYS_LARGEFILE on Solaris.  The little
> Joel> piece of code that I wrote to disable it on solaris-native should
> Joel> work well in this case too.
> 
> libiberty uses AC_SYS_LARGEFILE without any Solaris-specific code.  Why
> does that work?

Because the Solaris exception is required only for bfd/ and libiberty has no
ABI dependencies related to the largefile support.  bfd/ interface has such
ABI dependencies (at least `struct stat' of bfd_stat()).
	http://gcc.gnu.org/ml/gcc-patches/2009-11/msg00289.html


> If this code is still needed, then let me suggest the more painful but
> also nicer approach of a new .m4 file in config/.

There is now config/largefile.m4.


This mail thread is now closed by the checked-in ACX_LARGEFILE macro usage:
	http://sourceware.org/ml/gdb-cvs/2009-11/msg00079.html


Regards,
Jan