|
From: Diego 'F. <fla...@gm...> - 2007-06-02 09:30:09
Attachments:
signature.asc
|
Hello, Is there any chance that a new version of valgrind is released to support glibc 2.6? Seems to me like a lot of people might actually need valgrind for this, at least xine now started to crash mysteriously inside malloc() calls, and valgrind would be quite useful to get a hold of the problem. I see that -trunk at least has support for 2.6 from revision 6744, although it seems to me like the suppression file glibc-2.6.supp is wrong (object names are -2.5 still). --=20 Diego "Flameeyes" Petten=C3=B2 http://farragut.flameeyes.is-a-geek.org/ |
|
From: Julian S. <js...@ac...> - 2007-06-02 09:43:30
|
> Is there any chance that a new version of valgrind is released to > support glibc 2.6? Sure, this could be pushed into 3.2.4 if it does not cause any big complications. > I see that -trunk at least has support for 2.6 from revision 6744, > although it seems to me like the suppression file glibc-2.6.supp is > wrong (object names are -2.5 still). Uh, ok. Can you send a patch to glibc-2.6.supp that makes it work properly for you? Note, due to build system problems, every time you edit that file, you need to do "./autogen.sh && ./configure ... && make install" before you can use the new suppressions. What is the status of glibc-2.6 ? Is it released yet? How big are the changes relative to glibc-2.5 ? J |
|
From: Diego 'F. <fla...@gm...> - 2007-06-02 11:18:10
Attachments:
signature.asc
|
On Sat, 2 Jun 2007 10:42:36 +0100 Julian Seward <js...@ac...> wrote: > Sure, this could be pushed into 3.2.4 if it does not cause any big > complications. =20 I've jsut tried backporting that patch to valgrind 3.2.3, but the result is the same as the previous valgrind (built against 2.5) running on 2.6: valgrind: the 'impossible' happened: repeated section! =3D=3D9029=3D=3D at 0x38018B2C: report_and_quit (m_libcassert.c:136) =3D=3D9029=3D=3D by 0x38018C44: panic (m_libcassert.c:210) =3D=3D9029=3D=3D by 0x38018CB2: vgPlain_core_panic_at (m_libcassert.c:21= 5) =3D=3D9029=3D=3D by 0x38018CD1: vgPlain_core_panic (m_libcassert.c:220) =3D=3D9029=3D=3D by 0x380657B4: vgModuleLocal_read_elf_debug_info (readelf.c:1193)=20 =3D=3D9029=3D=3D by 0x3803644E: vgPlain_di_notify_mmap (debuginfo.c:207)=20 =3D=3D9029=3D=3D by 0x3804917D: vgModuleLocal_generic_PRE_sys_mmap (syswrap-generic.c:1918) =3D=3D9029=3D=3D by 0x3805D95A: vgSysWrap_amd64_linux_sys_mmap_before (syswrap-amd64-linux.c:944)=20 =3D=3D9029=3D=3D by 0x3804D2B2: vgPlain_client_syscall (syswrap-main.c:719)=20 =3D=3D9029=3D=3D by 0x380387C5: vgPlain_scheduler (scheduler.c:721)=20 =3D=3D9029=3D=3D by 0x3805A86F: run_a_thread_NORETURN (syswrap-linux.c:87) > Uh, ok. Can you send a patch to glibc-2.6.supp that makes it work > properly for you? Note, due to build system problems, every time you > edit that file, you need to do "./autogen.sh && ./configure ... && > make install" before you can use the new suppressions. =20 Uhm trunk doesn't seem good: --19309-- DWARF2 CFI reader: unhandled DW_OP_ opcode 0x6 valgrind: ../../coregrind/m_debuginfo/readdwarf.c:2118 (copy_convert_CfiExpr_tree): Assertion 'srcix >=3D 0 && srcix < VG_(sizeXA)(srcxa)' failed. =3D=3D19309=3D=3D at 0x380192F4: ??? sched status: running_tid=3D0 > What is the status of glibc-2.6 ? Is it released yet? How big are > the changes relative to glibc-2.5 ? =20 It is released (at least it's available on GNU's mirrors) since 17 May; the NEWS file is quite short, but seems like some things not noted there were changed in malloc's code, as both emacs had troubles with the new glibc, and now xine is having strange hiccups during calloc() calls. --=20 Diego "Flameeyes" Petten=C3=B2 http://farragut.flameeyes.is-a-geek.org/ |
|
From: Tom H. <to...@co...> - 2007-06-02 13:08:08
|
In message <200...@en...>
Diego 'Flameeyes' Pettenò <fla...@gm...> wrote:
> On Sat, 2 Jun 2007 10:42:36 +0100
> Julian Seward <js...@ac...> wrote:
>
> > Sure, this could be pushed into 3.2.4 if it does not cause any big
> > complications.
>
> I've jsut tried backporting that patch to valgrind 3.2.3, but the
> result is the same as the previous valgrind (built against 2.5) running
> on 2.6:
>
> valgrind: the 'impossible' happened:
> repeated section!
>
> ==9029== at 0x38018B2C: report_and_quit (m_libcassert.c:136)
> ==9029== by 0x38018C44: panic (m_libcassert.c:210)
> ==9029== by 0x38018CB2: vgPlain_core_panic_at (m_libcassert.c:215)
> ==9029== by 0x38018CD1: vgPlain_core_panic (m_libcassert.c:220)
> ==9029== by 0x380657B4: vgModuleLocal_read_elf_debug_info
> (readelf.c:1193)
> ==9029== by 0x3803644E: vgPlain_di_notify_mmap
> (debuginfo.c:207)
> ==9029== by 0x3804917D:
> vgModuleLocal_generic_PRE_sys_mmap (syswrap-generic.c:1918) ==9029==
> by 0x3805D95A: vgSysWrap_amd64_linux_sys_mmap_before
> (syswrap-amd64-linux.c:944)
> ==9029== by 0x3804D2B2:
> vgPlain_client_syscall (syswrap-main.c:719)
> ==9029== by 0x380387C5:
> vgPlain_scheduler (scheduler.c:721)
> ==9029== by 0x3805A86F:
> run_a_thread_NORETURN (syswrap-linux.c:87)
Ah, that's nothing to do with glibc as such - it's probably the same
problem that somebody reported on the bugtracker the other week where
they had a debuginfo RPM installed and some of the ELF sections that
had been moved to the debuginfo object hadn't been stripped from the
main object properly so they were duplicated.
Look for "repeated section" in bugzilla and you should find it.
Tom
--
Tom Hughes (to...@co...)
http://www.compton.nu/
|
|
From: Diego 'F. <fla...@gm...> - 2007-06-02 13:13:17
Attachments:
signature.asc
|
On Sat, 02 Jun 2007 14:08:03 +0100 Tom Hughes <to...@co...> wrote: > Ah, that's nothing to do with glibc as such - it's probably the same > problem that somebody reported on the bugtracker the other week where > they had a debuginfo RPM installed and some of the ELF sections that > had been moved to the debuginfo object hadn't been stripped from the > main object properly so they were duplicated. Oh that might explain, I've updated binutils almost at the same time, and if I recall correctly the current snapshot of binutils 2.17.50 doesn't strip out everything that gets moved to debuginfo from the original file with --only-keep-debug. I'll see to work this around, and then check again if the patch lets valgrind to build/work fine with 2.6. --=20 Diego "Flameeyes" Petten=C3=B2 http://farragut.flameeyes.is-a-geek.org/ |
|
From: Diego 'F. <fla...@gm...> - 2007-06-02 18:57:02
Attachments:
signature.asc
valgrind-trunk-glibc-2.6-suppressions.patch
|
On Sat, 2 Jun 2007 10:42:36 +0100 Julian Seward <js...@ac...> wrote: > Can you send a patch to glibc-2.6.supp that makes it work > properly for you? Here it comes then :) -- Diego "Flameeyes" Pettenò http://farragut.flameeyes.is-a-geek.org/ |