> The cause of this crash is that in revision 7291 the behavior of > VG_(record_ExeContext)() has been changed: in revision 7290 and before > it was allowed to record the exe context from within > VG_TRACK(pre_thread_create)(), but from revision 7291 on this results > in a crash.
As per http://www.mail-archive.com/valgrind-developers%40lists.sourceforge.net/msg01282.html and also http://www.mail-archive.com/valgrind-developers%40lists.sourceforge.net/msg01324.html the cause of the crash is that drd is calling VG_(record_ExeContext) for a thread which all the registers are zero, leading to the unwinder crashing. Not sure what 7291 has to do with it. 7291 simply created a copy of trunk 7290 post release. My guess is the important change is this, in 7300: --- trunk/coregrind/m_stacktrace.c 2007-12-12 11:42:33 UTC (rev 7299) +++ trunk/coregrind/m_stacktrace.c 2007-12-15 22:13:05 UTC (rev 7300) @@ -97,11 +97,9 @@ /* Assertion broken before main() is reached in pthreaded programs; the * offending stack traces only have one item. --njn, 2002-aug-16 */ /* vg_assert(fp_min <= fp_max);*/ - - if (fp_min + VG_(clo_max_stackframe) <= fp_max) { - /* If the stack is ridiculously big, don't poke around ... but - don't bomb out either. Needed to make John Regehr's - user-space threads package work. JRS 20021001 */ + if (fp_min + 512 >= fp_max) { + /* If the stack limits look bogus, don't poke around ... but + don't bomb out either. */ ips[0] = ip; return 1; } That change gets rid of an essentially bogus check and replaces it with something more reasonable. Unfortunately I think the bogus check caught the case where drd passes SP=0 (etc) and so had the unintended side effect of avoiding the segfault. Basically you can't call VG_(record_ExeContext) with a zero stack pointer as is happening now. So don't do the call -- wait till you have a valid SP and only then call it. In any case there is no loss since the previous arrangement would have gotten you only a 1-element stack trace with the top/only IP value being zero. J ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Valgrind-developers mailing list Valgrind-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/valgrind-developers
