+++ This bug was initially created as a clone of Bug #678830 +++ Currently we do not report JS-engine leaks when releasing JS runtime. In particularly, js_FinishGC silently releases all compartments and GC chunks when at that point any compartment or chunk containing unlocked live GC things indicate a leak. It would be nice to report that similar to the way the malloc leaks are reported.

Igor, do you still think this might be practical? Shutdown leak detection is our most heavily used test for memory regressions, so it would be nice if we could get better coverage for JS there.

(In reply to Andrew McCreight [:mccr8] from comment #1) > Igor, do you still think this might be practical? I tried to hack a patch this weak and try to see how useful that would be.

This would be useful. Igor, any progress?

(In reply to Nicholas Nethercote [:njn] from comment #3) > This would be useful. Igor, any progress? I have a rotted patch that did some reporting, however I have not found it useful - pure JS leaks are rare and the leaks that comes via XPCOM are visible via existing tools in any case. I plan to update the patch after we remove all the support for multithreaded JS runtime currently scheduled for landing in 2-3 weeks.

I don't think this will go anywhere, and comment 4 makes it sound like it wasn't that useful.