I updated cbindgen (with `cargo install cbindgen --force`), and it is crashing when I build m-c. How do I downgrade it?

oh right, mach wants to only use the version that doesn't work.

Thanksfully, it looks like the cbindgen installed by `mach boostrap` is not crashing.

glandium: would you have a doc somewhere explaining on how to use mozilla-unified ?

typically, I would create a branch with git checkout -b central central/default/tip

by how to use, I mean with bookmarks

glandium: never mind, I see in your workflow for gecko development, there's reference to it

hello

how to send pull request after commit?

mib_6cqfal, hi, on what project? mozilla-central ?

yes

mib_6cqfal, what repo have you cloned to write your patch ?

(we have a lot of repos)

I cloned the mozilla-central

using mercurial ?

yes

cool

so you need to send the patch to phabricator and ask for a review

instructions are there: moz-conduit.readthedocs.io/en/latest/phabricator-user.html, it's a bit long I'm sorry

Do I need to make separate branch like git to make changes or I can just make changes on master branch?

padenot: thanks

mib_6cqfal, as you wish, you don't need to do anything in mercurial, you can commit on top of tip (=master)

so i just cd into mozilla-central make changes and commit?

yes

ok Thanks

and then you push this commit to phabricator

thank you my friend

nooo

heh

sorry padenot :)

don't be, this patch is 100% wrong

padenot: way to start the week! :)

is it still true I need java 1.8 to build Gecko for Android ?

I'm getting this error on run mach build - cbindgen version 0.6.7 is too old. At least version 0.7.1 is required

how do i solve this. I'm running windows

mib_6cqfal: easiest way normally is to run ./mach bootstrap

nical: there is a wrench bustage on your push from inbound: treeherder.mozilla.org/#/jobs?repo=…=unclassified&selectedJob=227587400

Standard8: It didn't work

stefan_hindli|sheriffduty It appears that I failed to update one of the reference images. It is a simple thing to do, though, so I can land the fix in a few minutes, unless you prefer to back the patch series out and re-land with the fix (all 4 patches need to be landed/backed out together)

I can do whatever you want

If it doesn't take too long go ahead and land a fix

but take into consideration that the tree will be closed when you'll land the fix

stefan_hindli|sheriffduty: thanks, I will just land the fix then

mib_6cqfal: can you pastebin the output of that command you just ran?

mib_6cqfal: try `cargo uninstall cbindgen` then `./mach bootstrap` again

Standard8: Where do i get cargo from? its not in my env variables

mib_6cqfal: try looking in ~/.cargo/bin

mib_6cqfal: so probably c:/Users/Utkarsh/.cargo/bin/cargo.exe

stefan_hindli|sheriffduty: looks like wrench QuantumRender opt passes now. Apologies for the bustage

nical: no problem, glad you fixed it so quickly!

tnguyen: want to explain a bit more, why classinfo is needed?

why some QI call somewhere isn't enough?

if there's no classinfo, serialization will stop at that line

smaug:

tnguyen: ok, thanks. The patch just doesn't show where serializable stuff is used, so it was hard guess :)

thanks, the usage of that is in js, sendasyncmessage.

ok, so in some other patch

glandium: how do we go from RustLibrary('gkrust', gkrust_features) to the cargo invocation?

glandium: here, I gather searchfox.org/mozilla-central/source/config/makefiles/rust.mk#214 (somehow)

gl: looks like there's an ES lint failure on your push from inbound - treeherder.mozilla.org/#/jobs?repo=…=unclassified&selectedJob=227618186

tnguyen: sorry, slow here... is the idea that also nsI[Http]Channel will eventually have referrerinfo?

...isnt referrerinfo in loadinfo?

nvm

masayuki: ping?

jmaher: in latest m-c i still see some WPT meta files with version == "10.0.15063", is that expected?

39 instances of it

kats: then those are tests that don't fail/pass anymore specifically for win10 and we can remove those intances

instances

jmaher: ok

mconley, ping

jmaher: also a file like searchfox.org/mozilla-central/sourc…ender/reftests/blending-001.svg.ini - the default is PASS, right? so this file is entirely redundant?

kats: I assume so

ok

smaug: yes, that's the idea, but nsIhttpChannel should be one of the last step.

bz: just sent phabricator.services.mozilla.com/D19362 your way. It affects outlook.com, and I think is relatively low-risk, so if you have a minute it'd be great

bz: gotta go afk in a second, but will be back in a bit

padenot: pong

emilio: done

mconley, so, I tested on a "normal" machine, it's very easy to make audio glitch badly

padenot: with the latest Nightly?

mconley, and we've had enough bug opened to disable this

mconley, yes, with the latest version

mconley, to make things a bit clearer, any audio playback _relies_ on having a thread being the highest priority on the system, being less than is not going to work

padenot: I guess BELOW_NORMAL_PRIORITY_CLASS isn't high enough. :/

it's less than normal priority

if what you said in the bug is really correct (wouldn't be the first time microsoft documentation is unclear or wrong)

padenot: I guess for now, the simplest solution might be to put audio wake-lock'd background tabs back in the NORMAL_PRIORITY_CLASS bucket

mconley, what's really important (and I don't know if what you propose is doing that), is to not touch the priority of the thread servicing the audio callback, that is using MMCSS

padenot: so, to be clear, we're not touching any of the thread priorities directly. We're _only_ touching process priority.

which, as we've discovered, also influences thread priority

unsuprisingly. :/

using my tool, I've observed that this callback is not being scheduled on time rather frequently when the tab is in the background, and this causes the glitch (versus, for example, under-run because we're out of audio data to feed to the audio stream because of a network issue, or a decoding delay, etc.)

padenot: right, presumably because of the lower priority.

mconley, yes, you mentioned some sort of priority inheritance for all the threads of a process, in the bug

padenot: do you concur that leaving background tabs playing audio at the normal priority is a sufficient solution?

mconley, we've also observed this issue when remoting the audio stream to the parent process, the thread in the child process is at a normal priority, glitch happens frequently

mconley, I don't know, I guess we need to check whether or not setting a process at normal priority resets what we do using MMCSS

of course my "normal" windows machine is not booting now, it just blinks an orange led...

padenot: if I put together a patch for you, would you be willing to test that hypothesis?

probably if I can get it to boot

or breaking in a debugger and observing the respective thread priorities might work

we still haven't received the reference laptop in paris somehow

did anyone happen to see the Hamburger Menu failure mentioned in Bug 1526661 in the nightly build?

Just trying to get info on how to reproduce it

cmartin: I had that bug, it was that the hamburger menu (and some of the others, like the downloads panel) just would not open when you clicked on them.

but if you had the bookmarks menu on your toolbar, it worked fine

when did we drop the field0-0-0 syntax for bugzilla queries? seems to work on the rest api, but not in links to buglist.cgi?

mhowell: Thanks! I am wondering how it made it through my Mochitest Try Build without anything failing :S

cmartin: yeah, I immediately thought that seemed odd, too; good luck!

mhowell: I will try to repro, but I may ping you later for info about your gfx card/drivers/etc if that's alright with you :)

cmartin: cool, that machine is a desktop with an Nvidia 1080 in it, I can get you whatever else you need

kats: thanks for pointing me to that apz test-case! Would've taken me quite a bit to write a regression test otherwise

emilio: np, glad it was helpful :)

Anyone else getting a "Network Protocol Error" accessing searchfox?

Mossop: nope

Oh, seems to be just this one url .... which is 25k characters long

Mossop: Webtools :: Searchfox on Bugzilla.

code lands on github, but we track issues on Bugzilla

Thanks

Mossop: oh, is this bugzilla.mozilla.org/1511025?

emilio: Looks similar. I filed bugzilla.mozilla.org/1527008, feel free to dupe

What is the difference between NS_INTERFACE_MAP_ENTRY and NS_INTERFACE_MAP_ENTRY_AMBIGUOUS ? When should I use one over the other?

brennie: When your class inherits from two different things that implement an interface NS_INTERFACE_MAP_ENTRY_AMBIGUOUS allows you to say which classes implementation of an interface gets used

ohh

Mossop: ok thanks, that makes sense

eeeeeeugh

does supremecourt.gov/DocketPDF/18/18-52…1018143448223_Fort%20Bend%20Pet.pdf display horrifically in Firefox for anyone else?

evince locally seems to do fine with it

I see overlapping codeunit-like things

Looks like, the font selection doesn't work properly when drawing on canvas?

arai: yeah, I dunno

arai: it's a SCOTUS cert petition, never had problems viewing them before, and they all tend to have similar fonts'n'such, wonder what's up with this one

I see `Warning: Failed to load font "g_d0_f1": SyntaxError: An invalid or illegal string was specified` in browser console

it might be related

seems plausible

do we have pgo builds for arm64?

smaug: i don't think so, no

smaug: We don't have automated testing setup for them (there is in progress work for that)

tomprince: so nightlies aren't pgo either?

smaug: No.

We don't have pgo for any cross-compiled builds. m.shal and ch.manchester are working on that for macos and android.

Once we have machines setup to test win64-aarch64, we can probably re-use that for getting pgo there.

arai: you have a Github username?

Waldo: arai-a

hum, I tried to autocomplete that and it did not

(Thought it will require more hardware, probably, as we need a separate pool of workers for pgo)

or at least the account wasn't obviously you

tomprince: ok. another question, should target.zip for win64-aarch64 work on an arm64 laptop?

smaug: Should, yes.

tomprince: I get windows popup telling "you ned to install an app for this task"

smaug: I don't actually have hardware, so I'm not familiar with the experience.

aha, target.installer gives the same thing

filed mozilla/pdf.js #10544 for anyone in the peanut gallery who wants to follow along

the file is rendered correctly on macOS QuickLook, Preview, and Acrobat DC

Waldo: hmm, is that the issue bdahl was just fixing

my fix was just for printing, but we have a fix for the above issue

kats: question about a test and apz on esr 60: treeherder.mozilla.org/#/jobs?repo=…Str=windows10-64%2Cmochitest-e10s-4

kats: we seem to be timing out after 300seconds when the test normally takes <2 seconds; the last line in the log is |Flushed APZ repaints, waiting for callback...|, and for the failures, it seems that we are looping through some mousemoves and fail in a few spots but always have the same last message

and this is only showing up after upgrading to windows v1803; possibly there is something we can do to figure this out or maybe uplift?

jmaher: uplifting bug 1368603 might fix it

jmaher: if that doesn't do it i can't think of anything else offhand that landed since 60 that might affect this

kats: thanks; we can start there

bz: (following up from the question I left in the channel last week): would it be possible to make the performance object available to a non-window context? maybe through Cu.importGlobalProperties?

felipe: We could try. The spec for performance is very window-bound

felipe: we would need to go through all the things hanging off it one by one and decide how they should behave in a chrome context.

felipe: what are you actually trying to do?

felipe: Do you just need performance.now() ?

bz: no, there's some activity stream code that uses .now, .timeOrigin, .mark, .getEntriesByName

bz: and it uses the one from the hidden window

bz: and i'm trying to reduce the usage of the hidden window

mmhm

I guess I could use the one from the top-level window, but that's code in a .jsm and if the window is gone it would lose the existings markers..

So yeah, a bunch of that code is really tied to a Window

And the .timing and .navigation bits

In particular, the API design makes it so getEntriesByName might return all sorts of things, including the entries from timing/navigation

It's possible we could init some sort of performance object with all the timing/navigation stuff stubbed out

I'd have to go look at the code pretty carefully to tell

Anyway, we'd need to audit and possibly modify all of searchfox.org/mozilla-central/searc…h=dom%2Fperformance%2FPerformance.h

bz: ok, given the complexity I think I should see first how feasible is to use one of the existing windows

It's doable, but it's work and it's work that has to be done pretty carefully...

I have code that is crashing in `CallMethodHelper::Call` calling a method on a native xpcom object from js. How can I figure out what actually is being called from my stack frames? I'm using VS to debug

ok

I mean, it's not great that this is the situation

And adding marks to a random window is not great either...

it's some weird code that is used for both privileged and non-privileged content

*contexts

it should probably be split up and use something else from chrome (maybe TelemetryStopwatch)

brennie: See the debugging JavaScript section at developer.mozilla.org/en-US/docs/Mo…illa/Debugging/Debugging_on_Windows

Mossop: thanks

RyanVM|bbl: when do we turn off early_beta? I'm looking at calendar.google.com/calendar/embed?…AZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbQ and it doesn't seem listed?

gijs: tomorrow

Aryx: ta

The list of good first bugs keeps growing: searchfox.org/mozilla-central/search?q=shorcut

chutten, searchfox.org/mozilla-central/searc…=oh+well&case=false®exp=false&path=

bbouvier++

searchfox.org/mozilla-central/searc…shack%5Cs&case=false®exp=true&path= is still below 1k

Mossop: I'll look at those patches tomorrow

froydnj: No problem. It's pretty well split up now so should be fairly straightforward

Mossop: thank you for splitting it up!

yzen: ping

glandium: ping

nataliaCs|sheriffduty: pong

glandium: any chance you could check treeherder.mozilla.org/#/jobs?repo=…37678e2593c9a&selectedJob=227757046 ? it's tier 2 but still :D

wtf

aklotz: I've found github.com/MahdiSafsafi/opcodesDB/blob/master/AArch64.json#L273 really helpful for seeing all the encodings in one place. The manuals make my head hurt :)

dmajor: thanks!

kats: would it be silly to make searchfox.org/mozilla-central/source/gfx/layers/ipc/PAPZ.ipdl#54 a tad higher priority? Perhaps even high prio

birtles++

this information about transitioncancel is super useful! thanks a lot

felipe: no problem -- let me know if there are any specific things I can help with

birtles: sure thing. I haven't investigated yet the problem but from the description it looks exactly like it, since there are some transition-property changes involved

also, I didn't know there was a getTransitions() that can return a promise

I was considering manually implementing a promise for that

I mean, getAnimations?

felipe: Yeah, there are a few examples of how to use it at: drafts.csswg.org/web-animations-1/#use-cases

nataliaCs|sheriffduty: this should fix it: try/611ee40e304b7ba2500b70232ab07941b44d362e

nataliaCs|sheriffduty: corresponding try: treeherder.mozilla.org/#/jobs?repo=…04ee0791785523278dec2fee645c4d78446

felipe: You'll want to check `anim instanceof window.CSSTransition` though to get only transitions

do we have a DOM peer group on Phabricator?

birtles: where can I see a description or idl of this animation object?

felipe: drafts.csswg.org/web-animations-1/#the-animatable-interface-mixin and drafts.csswg.org/css-transitions-2/#the-CSSTransition-interface

glandium: thank you

felipe: bear in mind that getAnimations is currently only enabled to chrome callers on release channels -- but that shouldn't be a problem since we already use it to synchronize the tab throbber animations

felipe: this might be the best reference actually: searchfox.org/mozilla-central/sourc…/base/content/tabbrowser.js#647-659

birtles: amazing

birtles: I might ping you again with some more questions about transitions and animations

felipe: sure, any time

there were a few paths that I tried to go when implementing these new tab animations, but didn't work out well because it felt some features were lacking.. but maybe they existed and I didn't know about them

hiro: you can probably know by typing "#" in one of the reviewer fields on Phabricator and seeing what autocomplete options come up

birtles: I'll probably e-mail you this week with some questions!

felipe: sounds good!

gah! idiottttttt!

hiro: FWIW, we have discussed having a DOM peer group on Phabricator, but we have not actually done anything about it.

mccr8: it would be really a nice group for all of us. :)