This would be good to do and also would help with runnable labeling (since I think we don't need to bother labeling idle runnables for now).

Need to use idle + timeout for this too to ensure objects are released at some point

I don't really grok Olli's comment 1, but I think that could be deferred to a later point anyway?

This bug is or was about idle dispatch, and if idle dispatch is used, need to use idle + timeout to ensure objects are released soon enough in case there isn't any idle time.

Oh, oh, I see. Instead of labeling the runnable, we idle dispatch it, and then its unlabeledness doesn't really matter. What is a reasonable timeout value, then? Do we have guidelines on choosing those, or is it just "whatever sounds good and the reviewer thinks is OK?"

Comment on attachment 8896383 [details] [diff] [review] label IncrementalFinalizeRunnable dispatches Review of attachment 8896383 [details] [diff] [review]: ----------------------------------------------------------------- I think we should just use idle dispatch. It's not much more work. This code, which is for a similar purpose, uses a 2.5 second timeout: http://searchfox.org/mozilla-central/rev/e5b13e6224dbe3182050cf442608c4cb6a8c5c55/js/xpconnect/src/XPCJSRuntime.cpp#143 Maybe we should just do that?

Address the suggestion in comment 5. (Try to share some loading. Hope you don't mind, Nathan) Treeherder result looks fine: https://treeherder.mozilla.org/#/jobs?repo=try&revision=637c30a701d95a6484b521b9d5eceaf94d4b01e7

(In reply to Bevis Tseng [:bevis][:btseng] from comment #6) > Address the suggestion in comment 5. > > (Try to share some loading. Hope you don't mind, Nathan) That's quite all right, thank you!

Pushed by btseng@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/6668a2b3f65a Use idle dispatch for IncrementalFinalizeRunnable. r=billm