On Tue, Mar 07, 2006 at 09:19:06AM +0100, Petr Salinger wrote: > >I still have the problem with a gcc-4.0 rebuilt from source. I have > >uploaded it to the archive, as it is a newer version. > > > >Moreover the problem does not seem to come from binutils, as using the .o > >files from GNU/Linux fixes the problem. On the contrary I see the same > >problem using the .o files from GNU/kFreeBSD on GNU/Linux. > > It still can be bug in binutils - in gas. > > Please could you compile attached lib.c by commads bellow > on both linux-amd64 and kfreebsd-amd64 > and compare .s and .o, especially "readelf -s" ? > > gcc -fPIC -O2 -Wall -S lib.c -o lib1.s -DFIRST > gcc -fPIC -O2 -Wall -S lib.c -o lib2.s > gcc -fPIC -O2 -Wall -c lib1.s > gcc -fPIC -O2 -Wall -c lib2.s > gcc -shared -o liblib.so lib1.o lib2.o I have done you test, and I compared the resulting files. The .s files are identical, but the .o and .so files are different. readelf -s show the same sections (and a few difference than can be explained), but the assembly code is a bit different. Sorry I can tell you more, however I have attached the files to the mail. > BTW, glibc 2.4 in Debian might be closer than we think > http://sourceware.org/ml/libc-alpha/2006-03/msg00050.html Well, we have started a glibc 2.4 branch that will be targeted for experimental, however, we still don't know if it will be enough stable on all architectures to be considered for etch (the freeze of the toolchain will occur in July). Thanks for you help, Aurelien -- .''`. Aurelien Jarno | GPG: 1024D/F1BCDB73 : :' : Debian developer | Electrical Engineer `. `' aurel32@debian.org | aurelien@aurel32.net `- people.debian.org/~aurel32 | www.aurel32.net
Attachment:
visibility_amd64.tar.gz
Description: Binary data