Closed Bug 553009 Opened 16 years ago Closed 16 years ago

as3/sampling/Callback hangs intermittently on mac 32bit

Categories

(Tamarin Graveyard :: Virtual Machine, defect, P2)

x86
macOS
defect

Tracking

(Not tracked)

VERIFIED DUPLICATE of bug 546407
flash10.1

People

(Reporter: brbaker, Assigned: treilly)

Details

This testcase has hung a couple of times in the build system and I have been able to reproduce the hang about 1 in 10 times on my mac. Run the release-debugger 32bit shell and run the as3/sampling/Callback.abc $AVM acceptance/as3/sampling/Callback.abc Sampling api: callback callback1 called at 40064 memory summary: allocs=120324 deallocs=78584 The test should execute with output like: $AVM as3/sampling/Callback.abc Sampling api: callback callback1 called at 40065 callback1 called at 79998 memory summary: allocs=240129 deallocs=158344 callback2 called at 119929 callback2 called at 159980 Callback: callback called at least once callback#=2 = true PASSED! Callback: callback with return type called at least once callback#=2 = true PASED! Callback: function is not defined = ReferenceError: Error #1065 PASSED! I have been able to reproduce the issue with argo rev 3832 and 3812
Flags: in-testsuite+
Flags: flashplayer-triage+
Flags: flashplayer-qrb?
I can repro on my mac at office using release_debugger and running abc file multiple times, 1 out of 10 repros, seen only on mac-intel 32bit so far. NOTE: I am able to reproduce this issue using last weeks merge, this is NOT an injection for this week Similar bugs that show sampling/Callback to be flakey, but not exactly the same behavior: https://bugzilla.mozilla.org/show_bug.cgi?id=507685 https://bugzilla.mozilla.org/show_bug.cgi?id=464957
any change with jit/interp? is the cpu pegged when it's hung?
hybrid -> cpu pegs -Dinterp -> unable to repo (ran 30 times) -Ojit -> unable to repo (ran 30 times)
Assignee: nobody → treilly
Status: NEW → ASSIGNED
Flags: flashplayer-qrb? → flashplayer-qrb+
Priority: -- → P2
Target Milestone: --- → flash10.1
https://bugzilla.mozilla.org/show_bug.cgi?id=546407 Looks like deadlock, callstack doesn't make sense but I think the patch in 546407 should fix it (gdb) bt #0 0xffff0279 in __spin_lock () #1 0x00043d2d in MMgc::FixedMalloc::OutOfLineAlloc () #2 0x0005f464 in operator new () #3 0x0008cdd5 in avmplus::InvokerCompiler::InvokerCompiler () #4 0x0008d8b1 in avmplus::InvokerCompiler::jitInvokerNow () #5 0x000cbda6 in avmplus::Sampler::sampleSpaceCheck () #6 0x000cc7be in avmplus::Sampler::recordAllocationSample () #7 0x000ccd2e in avmplus::recordAllocationSample () #8 0x000433f4 in MMgc::FixedAlloc::Alloc () #9 0x00043d43 in MMgc::FixedMalloc::OutOfLineAlloc () #10 0x0005f464 in operator new () #11 0x0008cdd5 in avmplus::InvokerCompiler::InvokerCompiler () #12 0x0008d8b1 in avmplus::InvokerCompiler::jitInvokerNow () #13 0x00119343 in avmplus::callprop_b<avmplus::Toplevel*> () #14 0x000ad1d0 in avmplus::interpBoxed () #15 0x00073e14 in avmplus::AvmCore::handleActionPool () #16 0x00073f26 in avmplus::AvmCore::handleActionBlock () #17 0x0002d860 in avmshell::ShellCore::handleArbitraryExecutableContent () #18 0x0002dcae in avmshell::ShellCore::evaluateFile () #19 0x00022714 in avmshell::Shell::singleWorkerHelper () #20 0x00024832 in avmshell::Shell::singleWorker () #21 0x00024a2b in avmshell::Shell::run () #22 0x00036c2c in main ()
Duplicate of bug 546407 per Tom Reilly.
Status: ASSIGNED → RESOLVED
Closed: 16 years ago
Resolution: --- → DUPLICATE
Ran with the unapproved patch in 546407 50 times and did not hang, without patch it hung on the 3rd try running the abc.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.