Re: target/6569: sparc-sun-solaris2.7 C testsuite regression in compile/20011119-2.c

[Franz Sirl <Franz.Sirl-kernel@lauterbach.com>]

The following reply was made to PR target/6569; it has been noted by GNATS.

From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>
To: "Kaveh R. Ghazi" <ghazi@caip.rutgers.edu>
Cc: davem@gcc.gnu.org,davem@redhat.com,gcc-bugs@gcc.gnu.org,
 gcc-gnats@gcc.gnu.org,gcc-prs@gcc.gnu.org,jakub@redhat.com,
 mark@codesourcery.com,nobody@gcc.gnu.org,rth@redhat.com
Subject: Re: target/6569: sparc-sun-solaris2.7 C testsuite regression
  in compile/20011119-2.c
Date: Tue, 07 May 2002 15:14:49 +0200

 At 14:45 07.05.2002, Kaveh R. Ghazi wrote:
 >  > From: Mark Mitchell <mark@codesourcery.com>
 >  >
 >  > >> Can you try the attached patch? It seems to work for me, but the
 >  > >> bootstrap hasn't completed yet. While I was at it, I improved the
 >  > >> placing of the warning messages. I'm still a bit confused, cause it
 >  > >> sometimes seems olddecl and newdecl appeared to be swapped compared to
 >  > >> their sourcefile ordering.
 >  > >
 >  > > FYI, bootstrap+regtest on x86-linux-gnu completed successfully.
 >  >
 >  > OK; once you have confirmation of the SPARC results go ahead and check it
 >  > in.
 >
 >Unfortunately, the patch did not solve the problem listed in the PR.
 >I still get the same 'as' errors from compile/20011119-2.c
 >
 >compile/20011119-2.c:3: warning: weak declaration of `foo' after first use 
 >results in unspecified behavior
 >/usr/ccs/bin/as: "/var/tmp//ccQOIg1d.s", line 51: error: invalid operand
 >--------------------------------------------------^^^^^
 
 Yeah, I managed to get access to a solaris-2.8 machine, even though I 
 wasn't able to bootstrap with Solaris as/ld (see below), I was able to 
 reproduce the failure. Re-checking on x86-linux-gnu revealed that it even 
 doesn't fix the problem there, so I must have mixed something up yesterday. 
 Frankly, I'm a bit at a loss here, cause I've tried several strategies 
 yesterday and they either didn't fix the testcase or caused some of the 
 weak tests to fail. Especially I tried to use the TREE_USED flag of the 
 WEAK_DECLS TREE_LIST to mark when a weak already had been assembled, but it 
 didn't work out :-(. I seem to misunderstand something about how the tree 
 structures are handled.
 
 Franz.
 
 PS:
 The bootstrap failure with solaris-2.8 as/ld and configured with:
 
           --enable-shared \
           --enable-threads \
           --with-system-zlib \
           --with-libiconv-prefix=/usr/local \
           sparc-sun-solaris2.8
 
 This results in:
 
 ...
 strcpy                              0x42c 
 .libs/libstdc++.lax/libsupc++convenience.a/dyn-string.o
 strdup                              0x694 
 .libs/libstdc++.lax/libsupc++convenience.a/cxa_demangle.o
 __iob                               0x45e8 
 .libs/libstdc++.lax/libsupc++convenience.a/cxa_demangle.o
 __iob                               0x4634 
 .libs/libstdc++.lax/libsupc++convenience.a/cxa_demangle.o
 __iob                               0x4644 
 .libs/libstdc++.lax/libsupc++convenience.a/cxa_demangle.o
 __iob                               0x4654 
 .libs/libstdc++.lax/libsupc++convenience.a/cxa_demangle.o
 ld: fatal: relocations remain against allocatable but non-writable sections
 collect2: ld returned 1 exit status
 make[4]: *** [libstdc++.la] Error 1
 make[4]: Leaving directory 
 `/home/gnu/obj/gcc31/sparc-sun-solaris2.8/libstdc++-v3/src'