This small leak shows up in a few different mochitests. leaked 1 Accessible leaked 1 HyperTextAccessible (There are a few other things leaking, but they show up in all win32 content processes.) It seems like maybe this is showing up during this test: accessible/tests/browser/e10s/browser_caching_description.js Or maybe this is related to the preallocated process manager or something, because that's an early test (but not the first). There's a similar thing with accessible/tests/browser/tree/browser_aria_owns.js and as I said, other tests.
Yura, any ideas about this one?
(In reply to Marco Zehe (:MarcoZ) from comment #1) > Yura, any ideas about this one? I think this is the same issue we looked at a month ago or so, related to querying interface method https://dxr.mozilla.org/mozilla-central/source/ipc/mscom/Interceptor.cpp#610-626 Here's earlier discussion in IRC: http://logs.glob.uno/?c=mozilla%23accessibility&s=25+Aug+2017&e=25+Aug+2017#c161332 Aaron had a patch for me to try but iirc it did not fix the issue. ni? aaron in case he might have some thoughts.
Flags: needinfo?(yzenevich) → needinfo?(aklotz)
Whiteboard: [MemShrink] → [MemShrink:P2]
status-firefox57: --- → wontfix
status-firefox58: --- → fix-optional
Priority: -- → P2
status-firefox58: fix-optional → wontfix
status-firefox59: --- → ?
You need to log in before you can comment on or make changes to this bug.