Closed Bug 1046191 Opened 8 years ago Closed 7 years ago
_COUNT _CTOR,DTOR for GMPProcess Parent
Right now we seem to only seem to detect leaks like bug 1046165 with LSan, which is a crummy user experience. These all seem rooted in GMPProcessParent, so we should add MOZ_COUNT_CTOR stuff so we detect these leaks in all debug builds.
> which is a crummy user experience I really meant "which is a crummy Gecko developer experience".
Hrmph, this didn't actually turn the tree orange when landed on top of jld's leak, which is what I would expect: https://tbpl.mozilla.org/?tree=Try&rev=644c880e291f
(In reply to Andrew McCreight [:mccr8] from comment #2) > Hrmph, this didn't actually turn the tree orange when landed on top of jld's > leak, which is what I would expect: > https://tbpl.mozilla.org/?tree=Try&rev=644c880e291f The memory leak (bug 1046539) was provoked by a child process crash, and that crash (bug 1046538) occurred only on Sanitizer builds. I suggest trying it again with a MOZ_CRASH() somewhere near the PR_LoadLibrary in GMPChild.
Oh, right, I meant to retest that, thanks.
GMPProcessParent inherits from GeckoChildProcessHost which already has MOZ_COUNT_CTOR, so hopefully this just works. I'm not sure how I missed that before. So I'm just going to assume this happened due to the broken leak checking infrastructure...
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.