https://bugs.documentfoundation.org/show_bug.cgi?id=99784
--- Comment #58 from Michael Meeks <[email protected]> ---
Great trace Telesto, and nice to have in text too - no need to have the 's'
stepping after the nice symbols backtrace though:
frame #2: 0x00007fff8d6eb6ad
libsystem_pthread.dylib`_pthread_mutex_lock_slow + 285
frame #3: 0x00007fff783f95de CoreFoundation`CFRunLoopPerformBlock + 670
frame #4: 0x00007fff76295053 AppKit`-[NSEvent _postAtStart:] + 191
frame #5: 0x0000000114e09c4d libvcllo.dylib`ImplSalStartTimer(nMS=41) + 701
at saltimer.cxx:84
frame #6: 0x0000000114e09f39
libvcllo.dylib`AquaSalTimer::Start(this=0x0000600000003820, nMS=41) + 25 at
saltimer.cxx:126
if( pEvent )
[NSApp postEvent: pEvent atStart: YES];
deadlocks against:
frame #10: 0x00007fff779a8c26
HIToolbox`_BlockUntilNextEventMatchingListInModeWithFilter + 71
frame #11: 0x00007fff76092b79 AppKit`_DPSNextEvent + 1093
frame #12: 0x00007fff767a81c3 AppKit`-[NSApplication(NSEvent)
_nextEventMatchingEventMask:untilDate:inMode:dequeue:] + 1637
frame #13: 0x0000000114e02ef6
libvcllo.dylib`AquaSalInstance::DoYield(this=0x00006080000be8a0, bWait=false,
bHandleAllCurrentEvents=false, nReleased=0) + 998 at salinst.cxx:628
frame #14: 0x0000000114c928f0 libvcllo.dylib`ImplYield(i_bWait=false,
i_bAllEvents=false, nReleased=0) + 1744 at svapp.cxx:505
Which is somewhat amazing ...
Of course the gymnastics going on here to implement the sal timer callback API
is pretty horrible, although since timers are protected by the solar mutex -
perhaps posting an event to the main thread in order to set a timer event does
make some sense (?) The:
if( pSalData->mpFirstInstance->isNSAppThread() )
... do the real stuff ...
} else {
... post this event ...
}
It does seem extraordinary that ::Yield() when called by two separate threads
on the Mac hangs this way.
Norbert - any ideas ? =)
--
You are receiving this mail because:
You are the assignee for the bug._______________________________________________
Libreoffice-bugs mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs
