mozilla :: #e10s

17 Apr 2017
20:02bkellymrbkap: are you aware of any issues where calling opening a window immediately after a content process has started triggers an infinite loop in the parent process?
20:09mconleyjimm: great breakdown in your "Thread Separation Research" doc
20:11bkellymrbkap: never mind I found the problem... it was me
20:13bkellywhen implementing an actor dealloc function it helps not to call that function recursively
20:17mconleyjimm: now that you've done that breakdown... what are the next steps?
20:40jimmmconley: we need to make a decision whether or not it's worth the investment today considering the risk.. I'm going to chat with selena about it tomorrow and I need to get some UX feedback as well.
20:40jimmone concern I have is the amount of time that window might be visible (based on telemetry data)
20:40jimmI wonder if ux is taking that into account
21:09felipejimm/mconley: "Thread Separation Research" doc... where can I find that, as a curious bystander?
21:10* felipe requested access
21:43jimmI'll open it up
21:44jimmfelipe: ok, open to all
22:13firebotBug 1349363 NEW, [e10s-multi] Beta 54 experiment
22:13elanhow are you doing? do you think it will be testable by the beginning of business tomorrow morning?
22:22mrbkapelan: Yes.
23:27mrbkapfelipe: are you around by any chance?
23:37felipemrbkap: I am
23:38mrbkapfelipe: Hey, is there any non-manual way to test the rollout addon?
23:38mrbkapfelipe: I've written some code and verified it's valid JS.
23:38mrbkapfelipe: it's pretty hairy though :-/
23:38* mrbkap would like to step through it a few times before asking for review.
23:39felipemrbkap: you mean automated tests?
23:39mrbkapfelipe: well, how do I run the code? :)
23:40felipeJust a sec, let me send you a patch
23:40felipeThat I apply to run it
23:41felipeUnfortunately it's not pretty from an automated way, but it's simple
23:41mrbkapfelipe: that's all I need.
23:45felipemrbkap: if you build from central, this is all that's needed:
23:46felipeif you build from mozilla-beta (such that #ifdef BETA is defined), you don't need any code changes, but you need --enable-update-channel=beta... But I just find applying the above patch easier
23:46mrbkapfelipe: ah
23:46mrbkapfelipe: btw, I kind of ignored ADDON_ROLLOUT_POLICY
23:46mrbkapfelipe: is that OK/
23:47mrbkapfelipe: or, do you want to take a quick look at my patch to bootstrap.js before I attach it to the bug?
23:47felipemrbkap: yep, I think so.. If your code falls back to it when it doesn't qualify for e10s-multi, I'm pretty sure your new stuff can totally ignore ADDON_ROLLOUT_POLICY
23:47felipemrbkap: yeah I can look
23:48mrbkapfelipe: well, I did things the other way around
23:48mrbkapfelipe: e10s-multi qualification is checked after e10s qualification.
23:48felipeah, ok
23:48felipebecause e10s-multi is more strict, right?
23:48mrbkapfelipe: right
23:49felipemrbkap: looks reasonable!
23:50mrbkapfelipe: Phew.
23:50felipethe ".web" prefix is because it means "processes for web content"?
23:50mrbkapfelipe: It's a 3-patch series.
23:50mrbkapfelipe: Right.
23:50felipewhat are the other 2 patches?
23:51mrbkapfelipe: Centralizing the "how many content processes should we use" code and then using it throughout the codebase.
23:51mrbkapfelipe: pushing to mozreview right now
23:52felipemrbkap: alright.. I will probably only look at it tomorrow, but in the meantime I guess you can test it a bit and see if everything works as expected
23:53mrbkapfelipe: Yep, sounds great.
18 Apr 2017
No messages
Last message: 157 days and 23 hours ago