rebuild of glibc-2.3.0 for kfreeBSD (packages to test)

Aurelien Jarno aurelien at aurel32.net
Fri Nov 18 15:44:25 UTC 2005


Robert Millan a écrit :
> On Fri, Nov 18, 2005 at 10:07:36AM +0100, Aurelien Jarno wrote:
> 
>>They mainly fix problem of ET_DYN/ET_EXEC so that we could now get rid
>>of the ugly patch in binutils, the fact that ld.so was mapped at
>>0x00000000, rpcgen which is now compatible with gcc-4.0, the sys_errlist
>>symbol which was not available, plus some minor fixes.
> 
> 
> Note that we aren't really supposed to provide sys_errlist.  Newer ports of
> Glibc (as registered in upstream's shlibs file) stop providing deprecated
> symbols like this one.  GNU/kFreeBSD is already registered (don't panic ;),
> but it came later, and upstream will never accept a patch that gives us old
> symbols.
> 
> OTOH, I think new ports like amd64 already get rid of this.  So it's not a big
> concern not having it.

I did that because at least 30 packages (and probably more if we check 
all the logs) fail to build because this symbol is missing. Moreover, 
amd64 has this symbols:

[farad:~]$ objdump -T /lib/libc.so.6 | grep sys_err
0000000000000000 l    d  .gnu.warning.sys_errlist       0000000000000000
0000000000223fe0 g    DO .data  00000000000003f0  GLIBC_2.3   sys_errlist
0000000000223fe0 g    DO .data  00000000000003e8 (GLIBC_2.2.5) _sys_errlist
0000000000223fe0 g    DO .data  00000000000003f0  GLIBC_2.3   _sys_errlist
0000000000223fe0 g    DO .data  00000000000003e8 (GLIBC_2.2.5) sys_errlist

So I think we need that.

>>If you could, please test these packages. Note that you need a kernel
>>5.4 (which supports ET_DYN and ET_EXEC linker).
> 
> 
> I would add a Conflicts or something, to prevent people from screwing their
> system if they are still running 5.3.
> 
> How about:
> 
> Conflicts: kfreebsd-image-5.3-1 (<< 5.3-5), kfreebsd5
> 
> (I've verified the versions)
> 
I'll do that in the next upload, probably this night.

Aurelien



More information about the Glibc-bsd-devel mailing list