Despite the landing bug 1000539, I'm still seeing leaks under stacks in AudioOutputObserver::InsertFarEnd while running Mochitest-3. I'm now seeing about 1mb of leaks under M3.

Noticed this isn't a new'd object; we should balance moz_xmalloc() with moz_free(). Also, Clear() was popping but not freeing, and that was likely the source of the leaks.

now actually tested and working

Comment on attachment 8435615 [details] [diff] [review] ensure no error paths lose far-end audio chunks in getUserMedia one more try...

Comment on attachment 8435615 [details] [diff] [review] ensure no error paths lose far-end audio chunks in getUserMedia Review of attachment 8435615 [details] [diff] [review]: ----------------------------------------------------------------- Thanks! It might be nice to do MOZ_COUNT_CTOR/MOZ_COUNT_DTOR when you allocate and free these chunks, so that we would detect leaks in regular debug builds. I guess it would be a little tedious to put them every place these blocks are allocated and freed, as there is no actual CTOR/DTOR for them.

This patch fixed the leaks I was seeing with LSAN.