mozilla :: #devtools

12 Sep 2017
07:06gasolin_Honza, ping
07:19Honzagasolin_: Hi!
07:22gasolin_Honza, we'll have a mozilla dev conference in Taiwan, Ricky and I would like to share recent Devtool updates. We like to sync with you if we have right things to show. Do you have time to meet tomorrow?
07:23gasolin_Honza, the date is at 9/22
07:23Honzagasolin_: sure!
07:24Honzagasolin_: could be sometime around this time now, but tomorrow?
07:24gasolin_how about the usual time 3:30pm (UTC+8) tomorrow?
07:25gasolin_(just around this time)
07:25Honzagasolin_: sounds good
09:26mib_csh061Hi
09:26mib_csh061any one know what API should use instead of getResources for devtools ?
09:31mib_csh061getResources is for chrome
09:32mib_csh061what should I use for firefox?
09:32mib_csh061:(
09:32araiwhat is getResources ?
09:33mib_csh061this is method for devtools in chrome
09:33mib_csh061https://developer.chrome.com/extensions/devtools_inspectedWindow
09:34araibug 1361121 seems to be related
09:34firebothttps://bugzil.la/1361121 NEW, nobody@mozilla.org Implement devtools.inspectedWindow.getResources
09:37mib_csh061I saw this but I cant understand how to fix this
09:41mib_csh061I got this error:
09:41mib_csh061TypeError: browser.devtools.inspectedWindow.getResources is not a function
09:54nchevobbemib_csh061: it's not in Firefox yet. Maybe rpl can give you better status on that bug
12:05solezer0: do you want to do the nosdk update on the weekly meeting today?
12:14zer0sole: you can do it, I did it twice in a row, remember? :) Also, I think the most important bug to talk about is the one you're working on.
12:14soleI think I did that last week?
12:14pbroyes, I think it was Sole
12:14pbrofrom what I remember
12:14soleI'm checking, I'm not even sure
12:15soleBut I'd swear I did because I talked about the try build
12:15soleI also have to leave early today
12:20solezer0: it was definitely me
12:20soleprego...!
12:20zer0sole: it was you, but it before I did twice. :)
12:20zer0Maybe I missed once, but you've already did it twice too?
12:21soleDoes it really matter? I'm surely not keeping count of this
12:22zer0sole: I'm just saying that since you told me if I could do it twice that time, and you will catch up later, that's it. :)
12:24zer0sole: of course if you can't you don't have to worry, I'll do it!
12:25soleI moved that to the top and I'll do that, then I'll leave
12:26zer0sole, no really it's fine, I'll do it no worries. :)
12:26zer0You don't have to attend just for that!
12:53soleThat was an interesting way of wasting people's time, but thanks
12:58soleochameau: I'm looking into your patch! sorry for the delay
12:59ochameausole: thanks, note that there is no hurry as it is only docs.
13:00soleI care about docs :D
13:00jdescottes"only docs"
13:00jdescottesahahah
13:00jdescottesI was expecting a reaction there :)
13:00soleYeah I am not sending you the Docs police... yet
13:23ochameaudocs are going faster, they are already out of the slow release cycle ;)
13:23ochameauso no need to rush for a merge date for them
15:10smaugis mozbrowser used in devtools?
15:15pbrosmaug: I believe it is how our Responsive Design Mode works
15:16smaugok, thanks
16:03nchevobbesole: is this the "this is fine" character ?
16:03nchevobbelooks awesome
16:04solenchevobbe: yes! It helps me cope with life
16:04solehahaha
16:04soleI hug it from time to time whenever things are terrible
16:04pbro"Jason Debugger
16:04nchevobbelol
16:04pbrothat's your last name now
16:11solesorry all, I have to leave
16:24digitaraldsole ochameau: I filed https://github.com/mozilla/platform-health/issues/9
16:24ochameaupbro: I opened a meta bug regarding DAMP fixes, but the main issue is this one:
16:24ochameauhttps://bugzilla.mozilla.org/show_bug.cgi?id=1397745
16:24firebotBug 1397745 NEW, nobody@mozilla.org DAMP doesn't report significant regression/improvement because it only care at summary results
16:25pbroochameau: thanks
16:30ochameaudigitarald: I have questions for you on that bug --^
16:31ochameauI already forked your dashboard to build a DAMP one
16:31nchevobbeI really want to have that dashboard
16:32nchevobbeThe screenshot Alex linked to already shows stuff
16:33ochameaunchevobbe: here is a shot with console and debugger: https://screenshots.firefox.com/FlbvMqwz4m3YF5EK/127.0.0.1
16:34nchevobbeochameau: is there a tooltip or something so we can link a point in time to a patch ?
16:39digitaraldochameau: ha, you really stripped all the things for the devtools dashboard
16:42ochameaudigitarald: yes, it was really slow to build ad update and I've been radical
16:42ochameauhaving said that, the devtools file should just work if you drop it in upstream
18:34digitaraldmaybe some more eyes on these profiles: https://github.com/devtools-html/debugger.html/issues/3988 < ochameau mikeratcliffe
19:39digitaraldmikeratcliffe: looking at https://perfht.ml/2wn6RYD ; a lot of panel do react updates even when they are in background
19:40digitaraldI see console and netmonitor high up there and they are not in foreground.
19:41digitaraldgregtatum: you worked on panel frontend, maybe you got ideas? ^
19:48mconleygregtatum: ping
19:49mconleyor mstange: ping
19:55gregtatumdevtools client code runs in the main thread, so they may be responding to events, even if they are in the background
19:56gregtatumthose that looks like the console is responding to messages and re-rendering
19:56gregtatumi don&#39;t really have any context though
19:56gregtatumbatchedMessageAdd/throttledDispatchTimeout< is the throttling that the console does for new messages coming into the frontend
19:57gregtatumdigitarald: mconley: ^
19:58gregtatumit also looks like console has on network request logging
19:59mconleygregtatum: hi - sorry, I think I&#39;ve answered my own question over here. Sorry for the noise!
19:59gregtatummconley: no worries
19:59mconleygregtatum: wait, no, I have questions now
19:59mconleygregtatum: I was wrong. :) So, it looks like perf.html tries to prevent uploads of profiles until symbolication has completed
20:00gregtatummconley: ah i was conflating the pings
20:01mconleyThat&#39;s super great. Unfortunately, I&#39;m working with a user that seems to somehow be falling through the cracks. 99% of the profiles they&#39;ve attached have been unsymbolicated. I&#39;m reasonably certain they&#39;re using official Nightly&#39;s... is there some kind of timeout that a user can hit if they don&#39;t get the symbols in time?
20:01mconleyand if so, what is that timeout?
20:01gregtatummconley: I don&#39;t think there is a timeout, I didn&#39;t review that patch, hold on let me check
20:01mconleymstange: unping
20:01mconleygregtatum: thanks!
20:03gregtatumoh i did review it, i just have no long term memory, lol
20:06gregtatummconley: i wonder if we can change the behavior to display a warning instead, but still let the user upload a profile. In the meantime you might be able to right click, inspect element, and remove the disabled attribute
20:09mconleygregtatum: is there anyway I can symbolicate the profile after the fact?
20:10digitaraldmconley: we rolled out the disable upload as a quickfix until we have a way to symbolicate using snappy after the fact
20:11digitaraldmconley: did those profiles get recorded before the patch maybe?
20:11gregtatummconley: https://github.com/devtools-html/perf.html/issues/253
20:11gregtatummconley: it&#39;s at the top of our triage list
20:11digitaraldmconley: is that the issue though, are they uploaded?
20:12gregtatummconley: i&#39;m currently focused on data transformations of the call tree, but we can shift around some priorities and probably get someone working on it
20:13gregtatummconley: https://github.com/devtools-html/perf.html/projects/2#card-3576662 here is our triage list, and bug order, we prioritize every friday and issues bubble up to be worked on as we finish them
20:15mconleyCool thanks!
20:16gregtatumdigitarald: maybe for the profile you shared above we should look into implementing some wrapper component for the visibilitychange component https://developer.mozilla.org/en-US/docs/Web/Events/visibilitychange
20:17gregtatumyou could defer component updates until the page was visible again. Hopefully that event works correctly for devtools panels
20:18digitaraldgregtatum: agreed. there should be no react happening in background panels
20:18digitaraldI file an issue
20:18gregtatumit&#39;d be a good shared component
20:29digitaraldgregtatum: filed https://bugzilla.mozilla.org/show_bug.cgi?id=1399242
20:29firebotBug 1399242 NEW, nobody@mozilla.org Background panels cause massive overhead with React rendering
20:29digitaraldmaybe add more ideas
20:46gregtatumdigitarald: done
20:54ochameaudigitarald: fyi, I&#39;m only looking at toolbox opening performance, and plan to focus on that for the few days of FF57 we have left
20:55ochameaudigitarald: expect to see a significative speedup for the inspector
20:58jdescottesjryans: ping
20:59jryansjdescottes: pong
20:59jdescotteshi! regarding the generic solution I mentioned in the lazyloading bug...
20:59jdescottesthe thing I have in mind is that all specs/fronts should support the loading mechanism ochameau introduces in this patch
21:00jdescottesI don&#39;t really like having this only for 6 actor types
21:00jryansjdescottes: ah, so &quot;generic&quot; means, &quot;do the same for the rest of them&quot;?
21:01jdescottesjryans: yes, but I&#39;m open to better ideas :)
21:01jryansjdescottes: haha, that seems fine :) i wasn&#39;t sure if the &quot;generic&quot; thing had different design.
21:02ochameauI&#39;ll sure plan to apply that to everything, but I would like to get the inspector faster sooner than later. and it gives a pretty good idea of how it would look like
21:02jdescottesochameau: yep understood, I r+&#39;d based on this assumption
21:03jryansochameau: yeah, seems reasonable to me, since we know there&#39;s gains to be had from it for the inspectro
21:03ochameauI also have a list of followups to cleanup protocol.js, I may wait FF58 cycle for that
21:03jryansochameau: i can only imagine, it&#39;s quite a soup
21:08ochameaujryans: I just read your second review. we discussed with jdescottes to also move front require path to this &quot;allSpecs&quot; object (to be better named). so that all the module paths are in this registry object
21:08jryansochameau: okay, i think that seems reasonable!
21:09ochameauI imagine we would also do the same for actors, and remove the the big addBrowserActors/addTabActors
21:09ochameaus/would/could/
21:10jdescottesahah- just commented that having it in a separate file would indeed be nice :)
21:11jdescottesto be clear, I&#39;d like everything to follow the same pattern. Then having specs+fronts+actors in the same registry object or in 3 different ones, I don&#39;t see a big difference
21:11ochameaujdescottes: jryans thanks for the co-review. I wasn&#39;t expecting to see so much help about protocol.js :p
21:12gljdescottes: let&#39;s go with revision 2 of Bug 1399028
21:12jdescottescome on it&#39;s the best file in devtools
21:12firebothttps://bugzil.la/1399028 ASSIGNED, jdescottes@mozilla.com Photon update - DevTools tabs colors
21:12jryanshaha, fun with mystical infrastructure files!
21:13jdescottesgl: hmmmm let me revisit rev2, I&#39;m pretty sure it was terrible :)
21:13jdescottesgl: oh no never mind
21:13jdescottesrev2 is great, amazing even
21:14gli am ready to r+, but i think we just want to move the icon filters
21:14gldown to a darker place
21:14glwhere we don&#39;t have to look at it again
21:14glthese are the variables i am talking about
21:14jdescottesgl did you see my comment in the bug ? we should be able to use fill-context to clean that up when we have more time
21:14glya
21:15jdescottesI just don&#39;t think it worth updating all our svgs on such short notice
21:15glit would be nice to not have to do these filter things
21:15jdescottesyeah I don&#39;t like them either
21:15jdescottesok I&#39;m repushing rev2, you let me know if you want me to tune/move some stuff
21:16gljdescottes: I can move the variables and land it for you
21:16glsaves you a bit of time
21:16gland effort
21:16jdescottesgl: as you want, it&#39;s fine by me :)
21:17jdescottesgl: I&#39;ll let you do it, I can use one less context switch actually, and this will be one round trip less
21:17jdescottesthanks!
21:18gljdescottes: just let me know when you repush
21:36jdescottesgl: pushed + addressed comments! it&#39;s yours
23:05ntimgl: https://github.com/JanitorTechnology/janitor/pull/159
13 Sep 2017
No messages
   
Last message: 8 days and 4 hours ago