Re: Bug in GNU ld at elf_i386_copy_indirect_symbol with gcc-3.3

[H. J. Lu]

On Mon, Jun 02, 2003 at 11:56:06AM +0200, Martin Aumueller wrote:
> Hi,
> 
> I'm sorry I can't manage to reproduce it on Redhat 9.
> 
> A collegue of mine (the only one using RH9 right now) patched the
> libFL.a distributed by SGI with OpenInventor to the version included
> in the last tar file: he renamed all references to __c_type* to
> address@hidden because the libc of R9 no longer includes the
> unversioned symbol. Somehow, this patched libFL.a with an unpatched
> libXm.a caused this problem. When trying to reproduce it on RH9, I
> did also this renaming trick for the libXm.a from RH8, and then the
> application linked. With this patched libXm.a the application also

That is the right thing to do.

> linked on RH8. With the unpatched SGI libFL.a and the unpatched
> RH8 libXM.a, the application does link correctly. If I mix the
> unpatched libFL.a with the patched libXm.a, the same problem
> occurs.
> 
> If it is still of interest, I put all the libraries and object
> files into another tar file: 
> http://www.hlrs.de/people/aumueller/bug-ld-elf32-i386-complete.tar.gz
> There you will find patched and unpatched versions of both libXm.a
> and libFL.a for reproducing the problem with RH8.
> 

I downloaded your tar ball. But I still can't reproduce the linker
assert on RedHat 9. Do you have a tar ball I can use to reproduce
the linker assert in binutils 2.14.90.0.4 on RedHat 9 + errata?


H.J.