|
From: <sv...@va...> - 2011-02-10 09:09:20
|
Author: tom
Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011)
New Revision: 11534
Log:
Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964.
Modified:
trunk/configure.in
Modified: trunk/configure.in
===================================================================
--- trunk/configure.in 2011-02-09 12:47:23 UTC (rev 11533)
+++ trunk/configure.in 2011-02-10 09:09:09 UTC (rev 11534)
@@ -749,6 +749,13 @@
DEFAULT_SUPP="glibc-2.34567-NPTL-helgrind.supp ${DEFAULT_SUPP}"
DEFAULT_SUPP="glibc-2.X-drd.supp ${DEFAULT_SUPP}"
;;
+ 2.13)
+ AC_MSG_RESULT(2.13 family)
+ AC_DEFINE([GLIBC_2_13], 1, [Define to 1 if you're using glibc 2.13.x])
+ DEFAULT_SUPP="glibc-2.X.supp ${DEFAULT_SUPP}"
+ DEFAULT_SUPP="glibc-2.34567-NPTL-helgrind.supp ${DEFAULT_SUPP}"
+ DEFAULT_SUPP="glibc-2.X-drd.supp ${DEFAULT_SUPP}"
+ ;;
aix5)
AC_MSG_RESULT(AIX 5.1 or 5.2 or 5.3)
AC_DEFINE([AIX5_LIBC], 1, [Define to 1 if you're using AIX 5.1 or 5.2 or 5.3])
@@ -762,7 +769,7 @@
*)
AC_MSG_RESULT([unsupported version ${GLIBC_VERSION}])
- AC_MSG_ERROR([Valgrind requires glibc version 2.2 - 2.12])
+ AC_MSG_ERROR([Valgrind requires glibc version 2.2 - 2.13])
AC_MSG_ERROR([or AIX 5.1 or 5.2 or 5.3 GLIBC_VERSION])
AC_MSG_ERROR([or Darwin libc])
;;
|
|
From: Tom H. <to...@co...> - 2011-02-10 09:16:06
|
On 10/02/11 09:09, sv...@va... wrote: > Author: tom > Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011) > New Revision: 11534 > > Log: > Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964. > This is another one that should probably be pulled into the release. Tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: Julian S. <js...@ac...> - 2011-02-10 09:55:48
|
On Thursday, February 10, 2011, Tom Hughes wrote: > On 10/02/11 09:09, sv...@va... wrote: > > Author: tom > > Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011) > > New Revision: 11534 > > > > Log: > > Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964. > > This is another one that should probably be pulled into the release. Noted, thx. I'll start merging fixes into 3_6_BRANCH sometime tomorrow or the weekend. J |
|
From: Tom H. <to...@co...> - 2011-02-10 11:13:16
|
On 10/02/11 09:55, Julian Seward wrote: > On Thursday, February 10, 2011, Tom Hughes wrote: >> On 10/02/11 09:09, sv...@va... wrote: >>> Author: tom >>> Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011) >>> New Revision: 11534 >>> >>> Log: >>> Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964. >> >> This is another one that should probably be pulled into the release. > > Noted, thx. I'll start merging fixes into 3_6_BRANCH sometime tomorrow > or the weekend. It looks like 2.13 is provoking a new complaint in ld.so that we should probably suppress: ==26778== Conditional jump or move depends on uninitialised value(s) ==26778== at 0x613E7AAB: index (strchr.S:59 in /lib64/ld-2.13.so) ==26778== by 0x613D8084: expand_dynamic_string_token (dl-load.c:324 in /lib64/ld-2.13.so) ==26778== by 0x613D88E9: _dl_map_object (dl-load.c:2179 in /lib64/ld-2.13.so) ==26778== by 0x613D226D: map_doit (rtld.c:629 in /lib64/ld-2.13.so) ==26778== by 0x613DEC65: _dl_catch_error (dl-error.c:178 in /lib64/ld-2.13.so) ==26778== by 0x613D2186: do_preload (rtld.c:813 in /lib64/ld-2.13.so) ==26778== by 0x613D4FC8: dl_main (rtld.c:1696 in /lib64/ld-2.13.so) ==26778== by 0x613E565A: _dl_sysdep_start (dl-sysdep.c:244 in /lib64/ld-2.13.so) ==26778== by 0x613D58C0: _dl_start (rtld.c:334 in /lib64/ld-2.13.so) Tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: Bart V. A. <bva...@ac...> - 2011-02-10 11:14:04
|
On Thu, Feb 10, 2011 at 10:55 AM, Julian Seward <js...@ac...> wrote: > On Thursday, February 10, 2011, Tom Hughes wrote: > > On 10/02/11 09:09, sv...@va... wrote: > > > Author: tom > > > Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011) > > > New Revision: 11534 > > > > > > Log: > > > Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964. > > > > This is another one that should probably be pulled into the release. > > Noted, thx. I'll start merging fixes into 3_6_BRANCH sometime tomorrow > or the weekend. > I've just noticed that DRD is broken on Darwin (which has been made worse by r11461). Please wait until that has been fixed. Bart. |
|
From: Julian S. <js...@ac...> - 2011-02-10 11:19:33
|
On Thursday, February 10, 2011, Tom Hughes wrote: > It looks like 2.13 is provoking a new complaint in ld.so that we should > probably suppress: > > ==26778== Conditional jump or move depends on uninitialised value(s) > ==26778== at 0x613E7AAB: index (strchr.S:59 in /lib64/ld-2.13.so) Add a hardwired redirect for it, in m_redir.c, as we do for x86_linux, you mean? I'd prefer that to suppressing it. J |
|
From: Tom H. <to...@co...> - 2011-02-10 11:28:10
|
On 10/02/11 11:19, Julian Seward wrote: > On Thursday, February 10, 2011, Tom Hughes wrote: > >> It looks like 2.13 is provoking a new complaint in ld.so that we should >> probably suppress: >> >> ==26778== Conditional jump or move depends on uninitialised value(s) >> ==26778== at 0x613E7AAB: index (strchr.S:59 in /lib64/ld-2.13.so) > > Add a hardwired redirect for it, in m_redir.c, as we do for x86_linux, > you mean? I'd prefer that to suppressing it. Actually it looks like dl-hack3-cond-1 already gets it, but I was using a valgrind built with glibc 2.12 so my suppressions weren't working. A redir might be better though? Tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: Bart V. A. <bva...@ac...> - 2011-02-11 11:32:48
|
On Thu, Feb 10, 2011 at 12:13 PM, Bart Van Assche <bva...@ac...>wrote: > On Thu, Feb 10, 2011 at 10:55 AM, Julian Seward <js...@ac...> wrote: > >> On Thursday, February 10, 2011, Tom Hughes wrote: >> > On 10/02/11 09:09, sv...@va... wrote: >> > > Author: tom >> > > Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011) >> > > New Revision: 11534 >> > > >> > > Log: >> > > Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964. >> > >> > This is another one that should probably be pulled into the release. >> >> Noted, thx. I'll start merging fixes into 3_6_BRANCH sometime tomorrow >> or the weekend. >> > > I've just noticed that DRD is broken on Darwin (which has been made worse > by r11461). Please wait until that has been fixed. > Since I don't have access anymore to an OS/X system, it would be appreciated if someone who has access to such a system could verify the following: - Whether running the DRD regression tests with the code on the 3.6.1 branch triggers any misaligned_stack_error messages. - If so, whether merging r11536 from the trunk makes these misaligned_stack_error messages go away. Thanks, Bart. |
|
From: Rich C. <rc...@wi...> - 2011-02-13 18:19:57
|
On darwin OSX 10.6, the build of 'make regtest' is currently broken because
of code in (at least these, there may be others):
none/tests/amd64/cmpxchg.c
none/tests/amd64/fxsave-amd64.c
The problems stem from
- cannot do signed 4 byte relocation
- 32-bit absolute addressing is not supported for x86-64
I researched the second issue, and darwin only supports relative RIP
addressing for 64bit builds.
Rich
On Fri, 11 Feb 2011 12:32:22 +0100
Bart Van Assche <bva...@ac...> wrote:
> On Thu, Feb 10, 2011 at 12:13 PM, Bart Van Assche <bva...@ac...>wrote:
>
> > On Thu, Feb 10, 2011 at 10:55 AM, Julian Seward <js...@ac...> wrote:
> >
> >> On Thursday, February 10, 2011, Tom Hughes wrote:
> >> > On 10/02/11 09:09, sv...@va... wrote:
> >> > > Author: tom
> >> > > Date: 2011-02-10 09:09:09 +0000 (Thu, 10 Feb 2011)
> >> > > New Revision: 11534
> >> > >
> >> > > Log:
> >> > > Add glibc 2.13 support. Patch from Dmitry Djachenko in bug #265964.
> >> >
> >> > This is another one that should probably be pulled into the release.
> >>
> >> Noted, thx. I'll start merging fixes into 3_6_BRANCH sometime tomorrow
> >> or the weekend.
> >>
> >
> > I've just noticed that DRD is broken on Darwin (which has been made worse
> > by r11461). Please wait until that has been fixed.
> >
>
> Since I don't have access anymore to an OS/X system, it would be appreciated
> if someone who has access to such a system could verify the following:
> - Whether running the DRD regression tests with the code on the 3.6.1 branch
> triggers any misaligned_stack_error messages.
> - If so, whether merging r11536 from the trunk makes these
> misaligned_stack_error messages go away.
>
> Thanks,
>
> Bart.
--
Rich Coe rc...@wi...
|
|
From: Julian S. <js...@ac...> - 2011-02-11 12:29:15
|
On Thursday, February 10, 2011, Tom Hughes wrote: > >> ==26778== Conditional jump or move depends on uninitialised value(s) > >> ==26778== at 0x613E7AAB: index (strchr.S:59 in /lib64/ld-2.13.so) > Actually it looks like dl-hack3-cond-1 already gets it, but I was using > a valgrind built with glibc 2.12 so my suppressions weren't working. So this is a non-problem and I can stop thinking about it. Correct? J |
|
From: Tom H. <to...@co...> - 2011-02-11 12:30:18
|
On 11/02/11 12:28, Julian Seward wrote: > On Thursday, February 10, 2011, Tom Hughes wrote: > >>>> ==26778== Conditional jump or move depends on uninitialised value(s) >>>> ==26778== at 0x613E7AAB: index (strchr.S:59 in /lib64/ld-2.13.so) > >> Actually it looks like dl-hack3-cond-1 already gets it, but I was using >> a valgrind built with glibc 2.12 so my suppressions weren't working. > > So this is a non-problem and I can stop thinking about it. Correct? Pretty much, yes. tom -- Tom Hughes (to...@co...) http://compton.nu/ |
|
From: Julian S. <js...@ac...> - 2011-02-11 12:54:53
|
On Friday, February 11, 2011, Bart Van Assche wrote: > following: - Whether running the DRD regression tests with the code on the > 3.6.1 branch triggers any misaligned_stack_error messages. > - If so, whether merging r11536 from the trunk makes these > misaligned_stack_error messages go away. Both 11535 and 11536 fail the same way for me (for drd running /bin/ls on OSX 10.6.whatever). none, memcheck etc run it without problems. r11535 ==40776== Process terminating with default action of signal 11 (SIGSEGV) ==40776== General Protection Fault ==40776== at 0x10018E299: dyld_stub_binder (in /usr/lib/libSystem.B.dylib) ==40776== by 0x100247007: ??? (in /usr/lib/libSystem.B.dylib) ==40776== by 0x10001B4A9: pthread_once (drd_pthread_intercepts.c:525) ==40776== by 0x1000DA18E: compat_mode (in /usr/lib/libSystem.B.dylib) ==40776== by 0x1000022FD: ??? (in /bin/ls) ==40776== by 0x100002E53: ??? (in /bin/ls) ==40776== by 0x1000014AB: ??? (in /bin/ls) r11536 ==42993== Process terminating with default action of signal 11 (SIGSEGV) ==42993== General Protection Fault ==42993== at 0x10018A299: dyld_stub_binder (in /usr/lib/libSystem.B.dylib) ==42993== by 0x100243007: ??? (in /usr/lib/libSystem.B.dylib) ==42993== by 0x100018469: pthread_once_intercept (drd_pthread_intercepts.c:528) ==42993== by 0x1000D618E: compat_mode (in /usr/lib/libSystem.B.dylib) ==42993== by 0x1000022FD: ??? (in /bin/ls) ==42993== by 0x100002E53: ??? (in /bin/ls) ==42993== by 0x1000014AB: ??? (in /bin/ls) |
|
From: Bart V. A. <bva...@ac...> - 2011-02-13 07:44:24
|
2011/2/11 Julian Seward <js...@ac...> > On Friday, February 11, 2011, Bart Van Assche wrote: > > following: - Whether running the DRD regression tests with the code on > the > > 3.6.1 branch triggers any misaligned_stack_error messages. > > - If so, whether merging r11536 from the trunk makes these > > misaligned_stack_error messages go away. > > Both 11535 and 11536 fail the same way for me (for drd running /bin/ls > on OSX 10.6.whatever). none, memcheck etc run it without problems. > > r11535 > > ==40776== Process terminating with default action of signal 11 (SIGSEGV) > ==40776== General Protection Fault > ==40776== at 0x10018E299: dyld_stub_binder (in > /usr/lib/libSystem.B.dylib) > ==40776== by 0x100247007: ??? (in /usr/lib/libSystem.B.dylib) > ==40776== by 0x10001B4A9: pthread_once (drd_pthread_intercepts.c:525) > ==40776== by 0x1000DA18E: compat_mode (in /usr/lib/libSystem.B.dylib) > ==40776== by 0x1000022FD: ??? (in /bin/ls) > ==40776== by 0x100002E53: ??? (in /bin/ls) > ==40776== by 0x1000014AB: ??? (in /bin/ls) > > r11536 > > ==42993== Process terminating with default action of signal 11 (SIGSEGV) > ==42993== General Protection Fault > ==42993== at 0x10018A299: dyld_stub_binder (in > /usr/lib/libSystem.B.dylib) > ==42993== by 0x100243007: ??? (in /usr/lib/libSystem.B.dylib) > ==42993== by 0x100018469: pthread_once_intercept > (drd_pthread_intercepts.c:528) > ==42993== by 0x1000D618E: compat_mode (in /usr/lib/libSystem.B.dylib) > ==42993== by 0x1000022FD: ??? (in /bin/ls) > ==42993== by 0x100002E53: ??? (in /bin/ls) > ==42993== by 0x1000014AB: ??? (in /bin/ls) > Hi Julian, Thanks to a kind soul I have been able to run a few quick tests on Darwin. What I learned is: - Disabling the pthread_once() intercept in DRD is sufficient to let /bin/ls run fine under DRD. - Changing the pthread_once() intercept in DRD to a trivial intercept (just invoke pthread_once()) didn't help. - Disabling optimization (-O2 -> -O0) for drd_pthread_intercepts.c didn't help either. So it looks like the only alternative left to find out what is causing these GPFs (probably stack misalignment) is to analyze the assembler code generated by gcc for the DRD intercepts manually. Sorry but I won't have the time to do that before the 3.6.1 release. Bart. |