587277 - TM: ExecutablePool release race

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Chris Leary [:cdleary] (not checking bugmail)]

ASSERTION FAILED: m_refCount != 0
(/builds/slave/tracemonkey-linux64-debug/build/js/src/assembler/jit/ExecutableAllocator.h:115 void JSC::ExecutablePool::release())
NEXT ERROR TEST-UNEXPECTED-FAIL | /tests/dom/src/threads/test/test_suspend.html | Exited with code 11 during test run

Comment 1

[Chris Leary [:cdleary] (not checking bugmail)]

Ultimately we want to put the ExecutableAllocator instance per-compartment (note that compartments can migrate across threads, so putting it in ThreadData isn't appropriate). As a stopgap solution to avoid races, we'll do atomic refcounts on the ExecutablePools.

Comment 2

[Chris Leary [:cdleary] (not checking bugmail)]

Attachment: Atomic refcount of ExecutablePools.

Comment 3

[Robert Sayre]

Comment on attachment 465954 [details] [diff] [review]
Atomic refcount of ExecutablePools.

File a follow-up to deal with this issue. I will check this into TM now.

Comment 4

[Chris Leary [:cdleary] (not checking bugmail)]

(In reply to comment #3)
> File a follow-up to deal with this issue.

Bug 587288 filed.

Comment 5

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/9825a728b2cb