Fix ACX_LARGEFILE to use canonical host/target names

Fix ACX_LARGEFILE to use canonical host/target names

[Joseph S. Myers]

The ACX_LARGEFILE macro checks the values of $host and $target, which
may or may not be canonical names at this point depending on which
other macros have been called first; it should require
AC_CANONICAL_HOST and AC_CANONICAL_TARGET to ensure they are both
canonical.  This shows up as build failures for GDB configured
--build=i686-pc-linux-gnu --host=i686-solaris2.10
--target=i686-solaris2.10.  OK to commit this patch (GCC and src),
with regeneration of the GDB and binutils configure scripts that use
ACX_LARGEFILE?

2009-11-20  Joseph Myers  <joseph@codesourcery.com>

	* largefile.m4 (ACX_LARGEFILE): Require AC_CANONICAL_HOST and
	AC_CANONICAL_TARGET.

Index: config/largefile.m4
===================================================================
RCS file: /cvs/src/src/config/largefile.m4,v
retrieving revision 1.2
diff -u -r1.2 largefile.m4
--- config/largefile.m4	9 Nov 2009 23:19:11 -0000	1.2
+++ config/largefile.m4	20 Nov 2009 22:33:02 -0000
@@ -5,6 +5,11 @@
 
 AC_DEFUN([ACX_LARGEFILE],[dnl
 
+# The tests for host and target for $enable_largefile require
+# canonical names.
+AC_REQUIRE([AC_CANONICAL_HOST])
+AC_REQUIRE([AC_CANONICAL_TARGET])
+
 # As the $enable_largefile decision depends on --enable-plugins we must set it
 # even in directories otherwise not depending on the $plugins option.
 

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: Fix ACX_LARGEFILE to use canonical host/target names

[Jan Kratochvil]

On Fri, 20 Nov 2009 23:38:16 +0100, Joseph S. Myers wrote:
> The ACX_LARGEFILE macro checks the values of $host and $target, which
> may or may not be canonical names at this point depending on which
> other macros have been called first;

Missed that, thanks for catching it.


> it should require
> AC_CANONICAL_HOST and AC_CANONICAL_TARGET to ensure they are both
> canonical.  This shows up as build failures for GDB configured
> --build=i686-pc-linux-gnu --host=i686-solaris2.10 --target=i686-solaris2.10.

and so the config/largefile.m4 pattern
  sparc-*-solaris*|i[3-7]86-*-solaris*)
would not match `i686-solaris2.10'.  Canonical form of `i686-solaris2.10'
after AC_CANONICAL_HOST or AC_CANONICAL_TARGET is `i686-pc-solaris2.10' which
would already match the config/largefile.m4 pattern.

On native OpenSolaris 2009.06 snv_111b X86 tried that
  ./configure --host=i386-solaris2.11 --target=i386-solaris2.11; make
will fail due to it with:
  /usr/include/sys/old_procfs.h:57:2: #error "Cannot use procfs in the large file compilation environment"
(despite the OS is 64bit the default build targets are 32bit there; as on ppc).

While all the subdirectories calling ACX_LARGEFILE already contain
AC_CANONICAL_TARGET (equivalent form AC_CANONICAL_SYSTEM is not present there)
none of them explicitly calls AC_CANONICAL_HOST, bfd/ already contains it (by
../ltoptions.m4) but gdb/ does not - which causes the compilation error.


> with regeneration of the GDB and binutils configure scripts that use
> ACX_LARGEFILE?

I would like to get regenerated all the subdirectories calling ACX_LARGEFILE
(bfd/ binutils/ gas/ gprof/ ld/ gdb/) as the common config/largefile.m4 file
purpose was exactly to get all the subdirectories ABI compatible, as described
in the mails introducing ACX_LARGEFILE.



Thanks,
Jan

Ping Re: Fix ACX_LARGEFILE to use canonical host/target names

[Joseph S. Myers]

Ping.  This patch 
<http://sourceware.org/ml/gdb-patches/2009-11/msg00452.html> is pending 
review.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: Fix ACX_LARGEFILE to use canonical host/target names

[Nick Clifton]

Hi Joseph,

> OK to commit this patch (GCC and src),
> with regeneration of the GDB and binutils configure scripts that use
> ACX_LARGEFILE?
> 
> 2009-11-20  Joseph Myers  <joseph@codesourcery.com>
> 
> 	* largefile.m4 (ACX_LARGEFILE): Require AC_CANONICAL_HOST and
> 	AC_CANONICAL_TARGET.

OK for binutils.

Cheers
   Nick

Re: Ping Re: Fix ACX_LARGEFILE to use canonical host/target names

[Paolo Bonzini]

On 11/28/2009 09:11 PM, Joseph S. Myers wrote:
> Ping.  This patch
> <http://sourceware.org/ml/gdb-patches/2009-11/msg00452.html>  is pending
> review.

Ok.

Paolo