Found that in 690469 I dropped BasicManager()->DestroyedThebesBuffer(BasicManager()->Hold(this), mBackBuffer); which was collecting dying buffers and deallocating it. Looks like that part need to be moved back, otherwise we are leaking thebes buffers
Created attachment 565907 [details] [diff] [review] Collect and drop dying thebes backBuffers
This fix looks OK, but why wasn't the leak caught by any of our tests? Were buffers actually leaking during reftest-ipc but not being caught, or do none of our tests exercise the path that was leading to a leak? We need to make sure this doesn't happen again.
yep, that is interesting question... that leak is very easily visible in about:memory reporter.... but reason is that we don't run IPC-reftests on debug linux builds... only on opt builds... that is why we don't run leak enabled test-run...
Sigh. Filed bug 693478.
This patch need to be landed in couple with patch from bug 690469 in order 1-st bug 690469, second 693282