mozilla :: #e10s

16 Mar 2017
01:06Caspy7This tweet suggests that "cookie manager style addons" are unable to work with e10s
01:06Caspy7is that true?
01:11The_8472the cookie addons i'm using work fine in e10s. he'd have to be more specific
01:28Caspy7The_8472: thanks
01:28Caspy7I'll point them to
01:28Caspy7where there are obviously compatible cookie managers
15:41blasseydid you see me move your 1:1?
15:41blasseymrbkap: ^
15:41mrbkapblassey: Yeah, to 11 my time?
15:41blasseyyea, does that work for you?
15:41blassey11:30 also works for me, so which ever is better for you
15:42mrbkapblassey: 11 is slightly better.
18:01mrbkapblassey: 1:1?
18:04blasseymrbkap: in my room
18:09firebotBug 1306458 NEW, Figure out how to make gOpenPopupSpamCount work with multiple content processes
18:15mrbkapjimm: ping?
18:17jimmmrbkap: pong
18:17mrbkapjimm: Hey, I got an intermittent orange in a try push that might be sandboxing related (on Linux) -- are you interested?
18:18jimmsure, file a bug and add sb? to the whiteboard, we'll find it
18:18mrbkapjimm: ok.
18:18mrbkapjimm: it might not be related, I just saw something in the log that said "Sandbox: Unexpected EOF, op 0 flags 0302 path /dev/shm/org.chromium.33gBmx"
18:19mrbkapjimm: So maybe the sandbox is just reporting that something went wrong/
18:19jimmhmm, I don't recognize that log output.
18:19jimmgcp: ^
18:20gcpMeans a process died most likely
18:21gcpBroker or broker client saw eof on its connection
18:26RyanVMmrbkap: holy crap++ @ bug 1208957
18:26firebot FIXED, Intermittent Assertion failure: 0 == rv, nsprpub/pr/src/pthreads/ptthread.c:288, PROCESS-CRASH | Mai
18:26mrbkapRyanVM: how so?
18:27RyanVMmrbkap: those have been around forever
18:27mrbkapRyanVM: yeah.
18:27mrbkapRyanVM: it was gal's fault :P
18:34mconleybillm: hey - sanity check: layers acceleration is necessary in order e10s to work properly, right?
18:34billmmconley: no, it's not
18:34billmmconley: the basic compositor is sufficient
18:35billmmconley: and I think that's what we planned to ship on XP
18:35mconleybillm: so I'm looking at bug 1344839, and it's kinda a hell hole
18:35firebot NEW, Themes causing browsing failure
18:35mconleybut I think I've found a clue:
18:35mconleyI believe our style calculator is deciding that a background-color set to a CSS variable of "#" means the window should go transparent
18:36mconleyso we enter here:
18:36mconleyand we call something called CleanLayerManagerRecursive
18:36mconleywhich... looks like it blows away the layer manager for the window
18:37billmmconley: I don't think transparency is expected to work for remote browsers
18:37mconleyand, maybe my mental model of all of this isn't 100% accurate, but that destruction is probably not something that the content process is prepared for
18:37billmmconley: I can't say I really understand this code. you could ask dvander.
18:38mconleySounds good, thanks
18:38mconleydvander: ping
18:38billmmconley: but I remember mattwoodrow telling me a long time ago that transparent remote <browser>s won&#39;t work
18:38billmor maybe it was roc
18:38mconleybillm: I think we successfully set opacity: 0 on them for things like the spinner
18:39mconleybut I guess that&#39;s transparency at a different level?
18:39billmmconley: er, I guess I mean transparent background
18:39billmmconley: maybe that&#39;s not what&#39;s happening here?
18:39billmmaking them invisible is fine
18:39mconleyI think it&#39;s the destruction of the layer manager that&#39;s confusing everything here
18:40mconleyI&#39;m 80% convinced of that, but still trying to get some hard proof
18:40billmmconley: yeah, I dunno. dvander is a better person to ask.
18:40mconleybillm: k, no worries
18:41billmmconley: in any case, can we just ask the theme author to fix their theme? it sounds broken.
18:41mconleyyeah, for sure
18:41mconleybut jaws was concerned that that would wallpaper over a larger issue here
18:42mconleyyeah, it&#39;s the destruction of the layer manager that&#39;s doing it
18:42mconleyI&#39;ll update the bug
18:46* jaws perks up
18:48jawsbillm / mconley: okay i just read scroll back. we can definitely fix this in the theme parser code to throw away properties that use &quot;#&quot;. i think before they got disregarded anyways
18:48jawsif we think the root bug is not worth the cost to fix it, then we can fix it in the front-end. it would be very easy to fix it in the front-end
18:49mconleystill, this looks like a footgun in our GTK backend. If setting transparency to true for GTK completely breaks the browser in e10s mode, we should probably disable that ability.
18:49mconleyI&#39;ll let our widget folk decide though
19:05jawsmconley: the good news here is that with complete theme removals the only way that something third-party can set this property is by going through our LWT code
19:33dvandermconley: pong
19:33mconleydvander: hiya! I needinfo&#39;d you. :)
19:52dvandermconley: replied
19:53mconleydvander: much appreciated!
17 Mar 2017
No messages
Last message: 9 days and 15 hours ago