See bug 1065536 for context.

This was r+'d in bug 1065536, but it is causing Windows-only test failures: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=43a1611643d8

We marked this P3 because it seems like this isn't really a leak, but just some per-process singleton we're not tearing down properly in leak checking builds. The main goal here is to remove the noise from the leak logs so we can detect real leaks. Hopefully I'll get back to this at some point, but not in the near term.

Comment on attachment 8675165 [details] [diff] [review] Shut down ImageBridgeChild in content processes. Nical, is there some reason we can't just call Shutdown in the child process, too? I don't know what all is going on in the code, but a try run was green, and it seems to fix the ImageBridgeChild leak. https://treeherder.mozilla.org/#/jobs?repo=try&revision=0f75253cd574

Comment on attachment 8675165 [details] [diff] [review] Shut down ImageBridgeChild in content processes. Review of attachment 8675165 [details] [diff] [review]: ----------------------------------------------------------------- In short: I don't remember. The longer version is: I looked into hg logs and this comes from bug 774388 which attempted to fix a whole bunch of issues with the shutdown and turned into an epic saga. The reason is probably buried in there, I assume that we had to cut some corners to get the thing landed. I am keen on trying this, although you should watch out for intermittent issues after you land this because we had many of those in bug 774388 and it could be the reason ImageBridgeChild ended up leaked in content processes. You definitely want a try push with all platforms for this one!

roc is fixing this in bug 1215265.