Comment on attachment 8650724 [details] [diff] [review] mse-memory-reporter this is almost the same as bug 1190592, and like it, this isn't thread-safe. But unlike bug 1190592, we don't have the same requirements to use MallocSizeOf. You don't need TrackBuffersManagers::SizeOfResources() there's already a member with the cummulative size of all samples: TrackData::mSizeBuffer Check the MediaSourceDemuxer::GetMozDebugReaderData you need mVideoTrack->mVideoTracks.mSizeBuffer and mAudioTrack->mAudioTracks.mSizeBuffer mSizeBuffer will have to be made atomic.

Seems like we should just combine efforts here rather than have two separate mechanisms. Reporting via MediaMemoryTracker::CollectReports rather than nsWindowMemoryReporter is *much* preferred. With bug 1194555 we will be able to have asynchronous reporters which means we'll be able to measure off main thread. I can put together of proof of concept to show how it could work.

(In reply to Eric Rahm [:erahm] from comment #3) > Seems like we should just combine efforts here rather than have two separate > mechanisms. Reporting via MediaMemoryTracker::CollectReports rather than > nsWindowMemoryReporter is *much* preferred. > > With bug 1194555 we will be able to have asynchronous reporters which means > we'll be able to measure off main thread. I can put together of proof of > concept to show how it could work. Yeah, I agree. I hadn't seen the other bug, and I'm happy to go with that one instead.

(In reply to Matt Woodrow (:mattwoodrow) from comment #4) > (In reply to Eric Rahm [:erahm] from comment #3) > > Seems like we should just combine efforts here rather than have two separate > > mechanisms. Reporting via MediaMemoryTracker::CollectReports rather than > > nsWindowMemoryReporter is *much* preferred. > > > > With bug 1194555 we will be able to have asynchronous reporters which means > > we'll be able to measure off main thread. I can put together of proof of > > concept to show how it could work. > > Yeah, I agree. I hadn't seen the other bug, and I'm happy to go with that > one instead. > > *** This bug has been marked as a duplicate of bug 1190592 *** I'll probably use your version as it's easier to implement, but we can do the work in the other bug.