[PATCH] [perl bindings] Bizarre copy of UNKNOWN in subroutine

Le 20 juin 2011 15:28:32, Stefan Sperling a écrit :
> On Mon, Jun 20, 2011 at 09:54:12AM -0400, Stéphane Gaudreault wrote:
> > Le 19 juin 2011 13:53:12, Stefan Sperling a écrit :
> > > On Sun, Jun 19, 2011 at 07:43:31PM +0200, Otto Allmendinger wrote:
> > > > So does this qualify as a proper bug? Can I add this to the issue
> > > > tracker?
> > >
> > > Yes, please add it.
> > >
> > > Someone will need to pin down where the problem is coming from.
> > > Is it SWIG? Is it Perl? Is it Subversion?
> > >
> > > Can you try to reproduce the problem with an earlier version of
> > > SWIG and/or Perl?
> >
> > Hi,
> >
> > We think that only 32 bits systems are affected by this bug because
> >
> > cd subversion-1.6.17/subversion/bindings/swig/perl/native; make test
> >
> > fails on i686, but works on x86_64[1].
> >
> > We have that problem with either swig 2.0.3 or 2.0.4. We had no problem
> > with perl v5.12.3. The problem was noticed after the upgrade to v5.14.0.
> > Someone suggested that the problem could be related to the use of 64bits
> > offset by perl [2].
> >
> > Regards,
> >
> > Stéphane Gaudreault
> >
> > [1] https://bugs.archlinux.org/task/24540
> > [2] http://www.gossamer-threads.com/lists/perl/porters/263222
>
> Are you sure the test failures referenced in [2] are related to
> the "bizarre copy of UNKNOWN" problem? I don't see that error
> appearing in [2].
>
> But if your are sure it's related, the link at [2] clearly explains
> that perl and extensions were compiled in an incompatible way:
>
> "I doubt there's anything crucial about the particular flag, but rather
> it's the fact that you're building extensions using flags that give
> you code that is binary incompatible with the perl binary it's being
> built against."
>
> "With options like -D_LARGEFILE_SOURCE and -D_FILE_OFFSET_BITS=64' used
> to build Perl but dropped when testing extension building, you could
> be getting a different and incompatible stat structure or other binary
> incompatible differences between the extension and the Perl core. "
>
> Which is right. Any software that shares data structures needs the
> data structures to be compatible. Else it crashes and whatnot.
>
> So is this "bizarre copy of UNKNOWN" problem showing anywhere else than
> Debian and Arch Linux? Maybe it's a problem with how these distributions
> compile perl and related software? Maybe Perl is compiled with support
> for large files but Subversion is not, or something like that?

Hi,

Our collegue Marcela Mašláňová from the Red Hat team suggested that the
problem might be in in Makefile.PL where ExtUtils::MakeMaker overwrite the
CCFLAGS. The following patch fix the problem for us. Could you please apply it
?

Regards,

--
Stéphane Gaudreault
ArchLinux developer
===========================================
diff -Naur 
subversion-1.6.17.ori/subversion/bindings/swig/perl/native/Makefile.PL.in 
subversion-1.6.17/subversion/bindings/swig/perl/native/Makefile.PL.in
--- subversion-1.6.17.ori/subversion/bindings/swig/perl/native/Makefile.PL.in   
2010-11-24 20:42:16.000000000 +0000
+++ subversion-1.6.17/subversion/bindings/swig/perl/native/Makefile.PL.in       
2011-07-01 20:16:16.520892074 +0000
@@ -43,7 +43,7 @@
 my %config = (
     ABSTRACT => 'Perl bindings for Subversion',
     DEFINE => $cppflags,
-    CCFLAGS => $cflags,
+    CCFLAGS => $Config{ccflags},
     INC  => join(' ',$apr_cflags, $apu_cflags, 
                  " -I$swig_srcdir/perl/libsvn_swig_perl",
                  " -I$svnlib_srcdir/include",