valgrinding mozilla on wine

[Eric Pouech]

Dan Kegel wrote:
> On 3/7/06, Dan Kegel <dank at kegel.com> wrote:
> 
>>I'm trying to track down a wine crash in firefox
>>(http://bugs.winehq.org/show_bug.cgi?id=4762)
>>and thought valgrind might be helpful.  So, following
>>http://wiki.winehq.org/Wine_and_Valgrind
> 
> 
> When I posted the above, I hadn't yet applied the patch
> mentioned in the wiki page.  When I did, valgrind itself
> crapped out:
> 
> ==18736== Warning: client switching stacks?  SP change: 0xBE857B4C --> 0x48DFFF0
> ==18736==          to suppress, use: --max-stackframe=1174963364 or greater
> --18707-- INTERNAL ERROR: Valgrind received a signal 11 (SIGSEGV) - exiting
> --18707-- si_code=1;  Faulting address: 0x57E58959;  sp: 0x62527DA4
> 
> valgrind: the 'impossible' happened:
>    Killed by fatal signal
> ==18707==    at 0xB001AB5C: vgPlain_get_StackTrace2 (m_stacktrace.c:147)
> ==18707==    by 0xB001AC31: vgPlain_get_StackTrace (m_stacktrace.c:331)
> ==18707==    by 0xB000DDCF: vgPlain_record_ExeContext (m_execontext.c:199)
> ==18707==    by 0xB0001116: create_MAC_Chunk (mac_malloc_wrappers.c:139)
> ==18707==    by 0xB0001B17: vgMAC_malloc (mac_malloc_wrappers.c:203)
> ==18707==    by 0xB002D9C3: do_client_request (scheduler.c:987)
> ==18707==    by 0xB002D498: vgPlain_scheduler (scheduler.c:721)
> ==18707==    by 0xB003BB99: thread_wrapper (syswrap-linux.c:86)
> ==18707==    by 0xB003BC5D: run_a_thread_NORETURN (syswrap-linux.c:119)
> 
> so I guess there's definitely some work left to do on valgrind...
yes :-/
actually, I haven't updated the wiki page lately

- the good news: most of the points related to false positives is either 
in latest Wine and will be present in yet to come 3.1.1 (RC1 is for 
Friday this week)

- the bad news: the SEH support is still deficient, and we still have 
stack issues (what you get). I updated the VG patch on the Wiki that 
fixes (for me) some errors that look very close to what you describe 
(the stack ones). As of today, the VG folks are rather sluggish to look 
at those issues and to integrate some of the patches. So, they won't be 
present for 3.1.1, perhaps for 3.2 (somewhere in April).

A+

-- 
Eric Pouech