mozilla :: #fx-team

25 Jul 2017
00:25jawsmconley: patch is on bug. i put up the patch only 9 minutes after taking it :)
01:40dolskehmm, last couple Nightlies have been kind of sluggish and giving me the E10S throbber fairly frequently
03:03Mossopdolske: I noticed that too but thought it was due to reinstalling my machine
03:04dolskeyeah, your machine reinstall is probably not causing my issue. :)
03:05dolskealso seems like gmaps is kinda slow, but that's so hard to tell apart from network problems.
14:07Gijsmak: in I'm asking a much dumber question than the one you're answering :)
14:07firebotBug 1354533 ASSIGNED, Update the 'History' view in the Library for photon
14:08makGijs: hmmm?!
14:08Gijsmak: it sounds like you want me to create a method on some places object - is that going to be PlacesViewBase or something else?
14:08* Gijs doesn't know as much about places views as you and mdeboer
14:10makGijs: PlacesViewBase may be fine. I just don't know if this code is also speaking with other views that are not based off that
14:10Gijsno, I think it's only about the history view
14:10Gijsin the menu
14:11makwell, this is a generic "onSubViewShowing" method
14:11Gijsyes, but it only gets called when its own subview is being shown
14:11Gijsie the history subview
14:12makyes, what I'm saying is that if then we apply this to another view that is not Places-based, then it won't have our magic cleanup method
14:12makthis code looks like it's there to be reused
14:12makah I see what you mean
14:12makit's history-panelmenu
14:13makthis code is so tall it's hard to tell what it refers to
14:13makwell, ok then
14:13makI thought it was some generic helper to show subviews in any photon view :p
14:16Gijsmak: ah, hm, so do you still want it to be refactored into somewhere esle?
14:16Gijs*else, even
14:17makyes, PlacesViewBase is fine
14:31Gijsmconley: can you remind me how I suggest someone look at specific profiles on a bug?
14:31Gijs(for reference, bug 1384101 )
14:31firebot UNCONFIRMED, v55b11 seems slow again. (Adminer 4.3.1)
14:38jryanshmm, one of my browser windows has a tiny back and reload icon, is it expected?
14:38Gijsjryans: sounds like ?
14:38firebotBug 1383879 NEW, Chrome icons are too small
14:38Gijsjryans: do you have a multi-dpi setup?
14:39jryansGijs: ah, yes, i guess so... i have a MBP with external monitor, probably they are different...
14:40Gijsyeah, that looks like the other screenshot in that bug
14:40Gijsjryans: what version of OSX?
14:40jryansGijs: 10.12.6
14:40Gijsjryans: commenting in that bug to that effect is probably helpful...
14:40jryansGijs: the odd thing is that only one of two windows on that screen is affected...
14:41Gijsinteresting :)
14:50Gijsmak: OK, I tried... if you can let me know if what I put up makes sense that'd be helpful :)
14:50makI'll throw a 16 sided dice and decide your fate
15:24GijsOsmose: did you figure things out yesterday?
15:33OsmoseGijs: For the about page? Not yet, I was working on my add-on issues, which I think I have figured out, at least.
15:33OsmoseI think NetUtils is caching file:// URLs
15:34GijsOsmose: OK. I'm around for the next ~1.5 hour or so, though I have meetings from 5-6pm (ie in 30-60 minutes
15:34GijsOsmose: then dinner, then I should be back later tonight
15:34OsmoseAnd I think nsIFile.createUnique is creating two test XPIs with the same name
15:34GijsOsmose: happy to help debug stuff
15:34Gijsthat doesn't sound very unique!
15:34OsmoseIt's after the file is deleted
15:34Gijsyeah, that seems plausible
15:35OsmoseSo while I was trying to create a test add-on with one ID, it was using the ID from another test that reused the same "unique" filename
15:35OsmoseBecause NetUtils was returning a cached result
15:37Gijshow did this tie into the about:studies thing again?
15:38OsmoseThis was for another PR
15:38OsmoseThat hooks up the study storage so that it's installs and uninstalls add-ons properly
15:38Gijsah right
16:13aswanOsmose: just going through the backlog here, i don't know about NetUtils but if you're loading xpi files, the zip reader definitely does cache those.
16:33dolskeGijs / dao: meeting ping, you're late!
16:34daoI'm there
16:55Osmoseaswan: Is there any way to disable that caching?
16:56aswanOsmose: i don't know of a way to disable it. you can flush individual files with something like Services.obs.notifyObservers(file, "flush-cache-entry")
16:57aswanwhere file is an nsIFile
16:57Osmoseaswan: Ohhh, maybe. I'll try that, thanks!
16:58aswanyeah, if adding that to your test works, i would suggest landing the test with that but also either let me know or file a bug, ideally you shouldn't have to flush manually
17:00mconleyGijs: generally speaking, setting [qf] in the whiteboard will get it on the right people's radar
17:01mconleyGijs: although, looking at those two profiles... I'm not confident much can be drawn from it.
17:02Gijsmconley: ah.
17:02mrbkapGijs: ping?
17:03Gijsmrbkap: pong-ish, going to head out for dinner soon - what's up?
17:03mconleyGijs: I'll coment.
17:03mconleyer, comment
17:03mrbkapGijs: Hey, I was wondering if you had any time to look at bug 1375244
17:03firebot NEW, Remove sync PContent::Msg_KeywordToURI IPC message
17:03Gijsmrbkap: I have WIP patches
17:03Gijsbut they're unfinished
17:03sfosterriddle me this. How does disabling a pref controlling cosmetic animations break a test which checks XHR responseText functionality:
17:04Gijsmrbkap: if you want to drive them over the line that would be *awesome*
17:04* Gijs has other photon stuff to work on :(
17:04mrbkapGijs: Yeah, I'd be happy to.
17:04GijsI think when I started with those patches, I didn't think this would snowball as much as it did
17:04Gijsmrbkap: see also some backscroll in #developers about how to delegate keyword lookups to browser code etc.
17:04mrbkapGijs: Yeah, I was a bit worried when I did my initial searches :P
17:04sfosterwhile you are at it, how does changing download notification animations break browser scrolling?
17:04firebotBug 1376519 ASSIGNED, Add new Photon-themed download notification animations
17:05mrbkapGijs: I honestly have half a mind to fix the feed code first and leave the rest for a later date.
17:05* sfoster feels like the last Jenga player
17:05Gijsmrbkap: yeah
17:05mrbkapGijs: but I'll take your patches and see how far I can get.
17:05Gijsmrbkap: what's a good place for my wip patches?
17:05mrbkapGijs: Are there a lot of them?
17:06Gijsmrbkap: I also started by removing the keyword load flag from all the docshell calls, and realized too late that docshell uses it to also determine "can I do keyword lookups on this request after it initially fails"
17:06Gijsso I kind of need to revert bits of those
17:06mrbkapGijs: ah, ok.
17:06Gijsmrbkap: I guess right now I have 6
17:06mrbkapGijs: do you use git by any chance?
17:06Gijstried to keep individual consumers separate
17:07Gijssorry :)
17:07Gijsmrbkap: I can rebase and export them, or just push to mozreview with the reviewers removed
17:07Gijswhich is easier for you?
17:08mrbkapGijs: probably rebased+exported.
17:08Gijsmrbkap: so, I really need to go have dinner first, but I'll make sure to do that later tonight, if that's OK with you?
17:08dolskeHmm. Just tried to get a profile of my Nightly slowness with the profiler, but Cleopatra is complaining with "Profiler Exception: TypeError: this._profiler.getSharedLibraryInformation is not a function"
17:09mstangedolske: what version of the add-on are you using?
17:10mstangedolske: please install a recent version from if you're using an old version
17:10mstangedolske: time to update :)
17:11mstange... to 0.17...
17:11dolske"Last Updated: November 22 2016" wait, what? Does it not update itself?!
17:11mrbkapGijs: yep, no worries.
17:11* mrbkap has plenty of other stuff to keep him occupied.
17:11mstangedolske: not sure what happened there, I thought it did...
17:11mstangedolske: oops, we never updated to point to the new add-on
17:12mstangedolske: sorry
17:12dolskeah, seems the new thing is a WebExt, so presumably something didn't upgrade across
17:13mstangethere were two major switches; the first one was to 2.0, and the second was to the webext, and the migration path for the second one works
17:13mstangebut we never had one for the first one
17:20dolskemstange: I just saw a Slack thread where other people were also saying Nightly has become slugging -- is this perhaps already a known issue?
17:20dolskes/slugging/sluggish on macOS/
17:22mstangedolske: I don't know about it
17:41ntimjaws: ping
17:41jawsntim: in a meeting right now
17:42ntimok, let me know when you're free
17:48jawsntim: i'm free now
17:56zombiedolske: hey, if you have a few minutes today, i'd like to chat about the plans for the tab+tabstrip hiding issues
17:59dolskezombie: what did you want to ask? I don't have anything to add to what's in the bug...
17:59zombiehm, apparently i need to read the bug first ;)
18:00jawsmconley: yo, which room should i join?
18:00mconleyjaws: I'm in yours
18:03zombiedolske: i'm probably failing to find the right bug.. i wonted to get up to speed what you, bwinton and dietrich discussed in SF, i think andym|out wanted me to own those issues from the webextension side
18:05bwintonI don't think there was a bug. Just a lot of in-person conversations
18:06zombiestupid in person conversations, can't CC people ;)
18:08bwintonWell, you can, but it's a very manual process. :)
18:13mixedpuppydolske: who would be good to ni a bug about an autoscroll/middle click issue affecting extensions?
18:14mconleywhat the
18:14jawslol well then
18:15firebotBug 1384145 ASSIGNED, Permanent failures on Cedar: browser_roundedWindow_open_min.js, browser_roundedWindow_windowSetting_
18:16mixedpuppyor anyone ^
18:22firebotBug 1332447 NEW, WebExtension API to hide the tabstrip
18:22dolskemixedpuppy: hard to say, bug #?
18:23mixedpuppydolske: bug 1374188
18:23firebot UNCONFIRMED, Add WebExtensions API to enable/disable autoscrolling
18:23* mixedpuppy thought we werent going to hide tabstrip, but I know some people would love it
18:24bwintondolske: (The one you and I talked about was hiding sets of tabs, for TabGroups)
18:24bwinton(But I would also like to hide the tabstrip, for Tab Center Redux. :)
18:26zombiedolske: ok, and for the tab hiding, i hear you plan to commit to something like the existing showOnlyTheseTabs(), but probably after 57?
18:26dolskeI've committed to nothing!
18:28dolskethat sounds more like what bwinton was talking about -- there are ideas about APIs and approaches, but it's a very long way from anything concrete or resourced.
18:28bwintondolske: Will you commit to _not_ supporting tab.hide? ;)
18:28zombiei can probably do some of the lifting for that, as long as it's aligned with your plans and i get some guidance/reviews
18:39rhelmerhm. latest m-c build is getting content process crashes immediately on startup - happening for everyone or just me? :) (artifact build on macOS)
19:29mconleyMattN: ping
19:48makmconley: with you experience on e10s, do you know by chance why passing an array or nsIURI to a dialog window, on the other side I get things that are not instanceof Ci.nsIURI?
19:49makarray *of* nsIURI
19:49mconleymak: via openDialog?
19:49makmconley: yes!
19:50mconleymak: from the parent process, I assume
19:50makmconley: yep
19:50makit always worked
19:50mconleymak: can I see how you're calling it?
19:50maknow I'm debuggin a regression and it doesn't work anymore
19:50makmconley: this is the call
19:51mak(still investigating, I don't have the full story 100%)
19:52mconleymak: huh... can't see how that would change with e10s if this is all happening in the parent. What's coming out the other side?
19:52mconleyyou said they're not instanceof Ci.nsIURI... are you able to introspect? What are they?
19:53makin window.arguments[0] I get back the info object, but the array of nsIURI becomes an array of objects that look like nsIURI (same properties) but fail instanceof
19:53makI was just wondering if there was any known change across windows boundaries... I'll keep investigating and see what I can find...
19:54jawsmconley: look good?
19:54firebotBug 1384272 NEW, Add a talos test that tracks the performance of opening the preferences
19:55makmconley: fwiw, the array is filled up with browser.currentURI values
19:55mconleymak: nothing tickles my memory about this. :/ Let me direct you to someone who might have some more info in #developers
19:58makthanks, let's see!
20:11MattNmconley: pong
20:13jawsMattN: hey! mconley and i were talking about mozscreenshots and curious if you had thought about how hard it would be to store if a screenshot was "good" or "bad", and thus cause future screenshots to fail if they don't match the "good" one
20:13jawssimilar to talos builds
20:13jawsthis would mean we would need to store all screenshots
20:19MattNjaws: I think it wouldn't be great having all those images in the tree but there are ways we could make it a job fail on TreeHerder without doing that
20:19Mardakanyone know what's the current state of "A new feature will merge from Nightly to Beta only when its deemed ready, based on pre-established criteria determined by engineering, product, and product integrity." (from
20:20Mardakdmose and i are confused about that. does it mean some code doesn't automatically move from m-c to m-beta?
20:20jawsMattN: just store the hash of the last 'good' build?
20:20Mardakor that code is by default ifdef NIGHTLY_BUILD and needs some other approval to have that check removed?
20:21MattNjaws: hashes would work for identical images but I currently use a fuzz factor to avoid intermittents
20:21MattNIf we fixed all discrepancies (I think requiring gfx help) then hashes would work great
20:22zombieMattN: how big of a fuzz factor? if it&#39;s <3 pixels, maybe hash of a downscaled screenshots?
20:23MattNHaving a known good image is on the TODO list but I think we don&#39;t need to block on having it on TH. Just having reliable alerts working would be a great start
20:23MattNzombie: 3%
20:28zombie hm, 4x downscale with nearest-neighbor might work
20:28mconleyMattN / jaws: I wonder if could help
20:29mconleybecause the hashes it calculates can also be diffed to get a sense of how similar two images are
20:29mconleyI believe.
20:29mconleyalthough this sort of thing is definitely outside of my area of expertise
20:29mconleyno guarantees on how accurate this thing is.
20:30MattNWe want it to be sensitive enough to detect changes live an image switching from 1x to 2x though
20:30MattNIMO we don&#39;t need to change the comparisons in order to have known-good images
20:31mconleyright - note that I&#39;m not talking about changing the comparisons or how that works
20:31mconleyI was mostly talking about how we could store a hash for the &quot;known good&quot;, and that a hash like ImageHash generates could be useful
20:32jawsMattN: why would it not be great to store all the images in the tree? we store all the builds and those are much larger, right?
20:32jawsplus storage is cheap, right?
20:32MattNI could see it useful for speeding up comparisons but I also vaguely recall that the `compare` command I use does some kind of fast hash check
20:33MattNjaws: we don&#39;t store all the builds in m-c
20:33MattNI&#39;m talking about m-c
20:33jawsoh, i wasn&#39;t talkign about m-c
20:33jawswhat would we gain about storing things in mc?
20:33MattNwell I already store all the images so that&#39;s not a problem
20:33MattNhaving the reference images change at the same time as the code
20:33MattNthey need to be in-sync
20:34MattNbut hashes or references to other repos are also fine
20:35MattNI guess to answer your original question, this isn&#39;t hard to do, it just requires attention to mozscreenshots for some time
20:36MattNand making it reliable would be part of that
20:37jawsMattN: would this be something you think a student team could work on?
20:37GijsMardak: I expect it&#39;s just about features that are behind either compile-time flags or prefs (prefs with per-channel defaults)
20:37jawsMattN: is it easily paralellizable?
20:37GijsMardak: what are you actually asking, though? :)
20:38MardakGijs: specifically, activity-stream is currently only built on nightly as well as only being preffed on on nightly. the plan is to have beta 56 allow shield studies to turn on the pref, so for that to work, it needs to be built outside of nightly
20:38MattNjaws: hmm the in-tree runner and snapshot configuration stuff has a good foundation for students to build on and improve
20:39MattNjaws: but I think we would need more coordination with a-team on the comparison and web front-end so that we do things in a way that a-team can maybe take over
20:40GijsMardak: sounds like you would be good if you have agreement from &quot;engineering, product, and product integrity&quot; that that&#39;s a good idea?
20:40jawshmm... yeah
20:40MattNjaws: well there are many a-team/releng bugs that could be mentored
20:40MattNbut that requires a mentor who knows that stuff
20:41MattNjaws: see the RelEng/A-team heading at
20:41MardakGijs: i suppose an email to PI would clear things up :p thanks
20:41MattNthose are almost all actionable and discrete
20:42MattNthe &quot;Mach mozscreenshots (Bugzilla)&quot; may have enough stuff too and we could mentor but it&#39;s not solving the original problem you mentioned directly
20:43jawsright, i think it would be nice to have pushes fail when they break UI
25 Jul 2017
Last message: 5 minutes and 46 seconds ago