A test I added recently (TestWindowHandles.test_tab_and_window_handles) hangs pseudo-deterministically with e10s turned on. getOuterWindowWithId in Marionette:register is throwing. I don't think it's supposed to, but we don't do much with the listenerWindow reference in the current code once we get it (and don't care if it's null), so maybe we're ok to catch and carry on.

I'm having a lot of trouble reproducing this now, but I was able to determine earlier that the error is NS_ERROR_XPC_BAD_CONVERT_NATIVE, and getOuterWindowWithId is called with the id of a freshly created window in the child process.

/r/2557 - Bug 1120809 - Don't fail when trying to retrieve a window from the child process in "Marionette:register".;r=jgriffin Pull down this commit: hg pull review -r ee82b1602297101c057c0aa409bff145bdeba2ac

This is what I mean by "pseudo-deterministic": https://treeherder.mozilla.org/#/jobs?repo=try&revision=33b7abc6d85f I'm waiting on try results with the fix in comment 2 here: https://treeherder.mozilla.org/#/jobs?repo=try&revision=187007fd94b9 Ultimately we might have to abandon the idea of having a reference to the listenerWindow. Looking through the uses on elementManager.seenItems, I don't think we're getting much out of it anyways.

I guess those might not be scheduled at this point. Re-push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=5be6a543bed4

/r/2557 - Bug 1120809 - Don't fail when trying to retrieve a window from the child process in "Marionette:register".;r=jgriffin /r/2575 - Bug 1120809 - Unskip test_tab_and_window_handles with e10s enabled.;r=jgriffin Pull down these commits: hg pull review -r fce8fccc2bdb252a6d785849e757fcede3cb2344

https://reviewboard.mozilla.org/r/2555/#review1747 Looks reasonable.

https://hg.mozilla.org/integration/mozilla-inbound/rev/eb735f61a1af https://hg.mozilla.org/integration/mozilla-inbound/rev/b0734bca62e4

https://hg.mozilla.org/mozilla-central/rev/eb735f61a1af https://hg.mozilla.org/mozilla-central/rev/b0734bca62e4