Closed Bug 832437 Opened 12 years ago Closed 11 years ago

Assertion failure: &entry == &stack->mStack[stack->stackSize() - 1], at tools/profiler/TableTicker.cpp:747

Categories

(Core :: JavaScript Engine: JIT, defect)

x86
macOS
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla30

People

(Reporter: Irving, Assigned: jandem)

References

Details

Attachments

(2 files)

Debug nightly build, with all profiling enabled; browser was in the background but had recently launched a tab containing http://www.marriedtothesea.com/archives/2006/Feb/ In a similar location to bug 776025, but I'm not certain they're related. The crash was not recorded in about:crashes.
I'm working on enabling TI and IonMonkey for chrome and we hit this crash on Try, in Mochitest-BC: https://tbpl.mozilla.org/?tree=Try&rev=ab22d82dc5b5 It seems Linux Debug only and so far I'm unable to reproduce it locally. BenWa, any idea what's going on here? Any pointers on how to debug this (on Try)? This is the last issue blocking bug 929374.
Blocks: 929374
Flags: needinfo?(bgirard)
When the JS engine pushes a js frame to the pseudostack it doesn't have to fill in the JSPC until it pushes another js frame pseudostack. This assertion is catching that a js frame on the pseudostack should but doesn't have a js pc. My advice is to get this to reproduce locally instead of Try. I know irving hit it a few times but most people don't profile with debug builds (because the performance numbers are off) so these assertion are often missed.
Flags: needinfo?(bgirard)
Component: Gecko Profiler → JavaScript Engine: JIT
Attached patch PatchSplinter Review
I added some extra debug checks, then ran all JIT tests with the SPS profiler + extra assertions enabled. This caught some problems where we could have JS frames with a NullPCIndex. This patch fixes them. And indeed, I have 5 beautiful green b-c runs now with Ion and TI enabled for chrome, yay: https://tbpl.mozilla.org/?tree=Try&rev=c641fd52bec4
Assignee: nobody → jdemooij
Status: NEW → ASSIGNED
Attachment #8374815 - Flags: review?(kvijayan)
Comment on attachment 8374815 [details] [diff] [review] Patch Review of attachment 8374815 [details] [diff] [review]: ----------------------------------------------------------------- Ah, VM calls from Ion ICs. Nice job tracking this down.
Attachment #8374815 - Flags: review?(kvijayan) → review+
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla30
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: