mozilla :: #media

19 Apr 2017
00:12debianuserAh, yeah, I mean write access. Alsa also reads configs in /etc/asound.conf and ~/.asoundrc. And every library would need access to /usr/lib* (depending on the distro) and probably /usr/share. Does sandboxing restrict reads too?
00:15rilliankamidphish: pong
00:15kamidphishrillian: I was looking for the version of rust that gecko compiles with
00:16rilliankamidphish: 1.15.1 is the current minimum
00:17kamidphishrillian: Thanks
00:17rillianofficial builds use 1.16.0
00:17kamidphishrillian: I'm trying to setup Travis CI to compile all the versions for me
00:17rillianah, cool.
00:18rillianI think travis uses rustup, so you might be able to pass a version number instead of a release channel.
00:18kamidphishrillian: Is this visible?
00:18rillianThe minimum for gecko is at
00:18rilliankamidphish: looks like it works!
00:19kamidphishrillian: So try showed me that 32 bit didn't compile, so I didn't want a repeat of that!
00:20rilliangood plan
00:23rilliannow you just need a test that verifies your travis minimum version against gecko's :)
00:29kamidphishrillian: Yeah, *sigh*
00:43rillianhmm, that rust-matrix script doesn't use rustup or the install scripts.
00:44rillianI wonder if you could replace with `before_install: rustup target add $ARCH`
00:45rillianor `rustup target add $ARCH-unknown-linux-gnu` if you don't want to put in a full triplet
00:46kamidphishrillian: I just googled and took the top answer. :-)
00:47kamidphishrillian: Patches accepted
01:16TD-Linuxdebianuser, yes, it restricts reads. otherwise you could read ~/.ssh etc
04:11kentuckyfriedtakahegerald: does bug 1166166 interest you?
04:11firebot NEW, use mAsyncExecutionThreadIsAlive to shrink memory on the async thread when possible.
04:11geraldI like the number
04:13geraldkentuckyfriedtakahe: I don't understand it (yet), plus I'm still fighting with the other two (gfxVars stuff was backed-out, ContentDeviceData is causing Talos regression), so if you could find someone else that'd be great. Otherwise I'll add it to my queue...
04:14kentuckyfriedtakahegerald: don't worry about it then
04:14rilliankamidphish: it's complicated by the fact that rustup treats trying to install the host as a target as an error.
04:15kamidphishrillian: I'm confused.
04:20rilliankamidphish: was working on a patch to replace the rust-matrix script on travis.
04:27kamidphishrillian: OK. Cool!
04:28kamidphishrillian: I admit to just cargo-culting bits and pieces to get something to work.
04:33kamidphishrillian: So I was just looking for a "zeroconf" solution. The carllerche github example looked to do what I needed.
04:52rillianit does, I just worried it would be brittle.
04:59kamidphishrillian: : Re: Gonk. Well it turned out the Cargo.toml changes to enable static was more than I expected. Are you saying just forget about Gonk working?
04:59kamidphishGonk isn't the same as Android, is it?
05:02rillianyou should probably check with one of the taiwan devs. It was originally just firefox os, but I suppose it's possible part of it got borrowed for Android.
05:02rillianjw_wang ^^
05:56rilliankamidphish: my clever one liner doesn't work on travis. Enough fiddling for today.
06:15bwukamidphish: what is your question?
06:15kamidphishbwu: Is Gonk just for Firefox OS?
06:16kamidphishbwu: Meaning that we don't need to support it.
06:17bwuYes again. :)
06:20bwuBasically, many gonk codes have been removed.
06:21bwuSo if you still see some, please fire bugs. Thanks.
06:23* bwu still has some feelings about Gonk...
09:49chunminpadenot: Do you know how to hook Window that we have a multi-track recording device(more than 2 track)? I need to hook Windows that I have a connected N-track device, where N > 2, to hit the assertion(
09:49chunminI tried using VB-Audio Virtual Cable but I get only two track for recording. Do you have any workaround?
13:07mreavyachronop: ping
13:07achronopmreavy: pong
13:08mreavyhi Alex, is now a good time to talk?
13:09mreavyachronop: I'm experiencing a lot of drop outs -- audio and video
13:12achronopmreavy: you are gone ...
13:13mreavyachronop: I had muted my video to see if it helped
13:31mreavypadenot: standup?
13:32mreavyjesup: standup?
13:32padenotmreavy, yes
13:48jesupVideo quality in moz-webrtc (SFU) was much improved!
13:48jesupfippo: ^
13:51fippojesup: \o/
13:52fippocompared to when? (i'd like to figure out what changed... :-)
13:53achronopone huge improvement is that there were no video freeze
13:54jesup2 weeks ago (no standup last week)
13:54jesupyes, the video freezes was the difference - your NACK fix I presume
13:57fippoyeah, that was deployed 12 days ago. \o/
13:59fippojib: can you take a look at please? will create a conflict in your addTrack shim I think.
14:00fippooh wait. I meant 507. but you can take a look at 509 too (-:
14:02achronoppadenot: can you review the cubeb import ?
14:05padenotachronop, yep, will do in a sec
14:05achronopit's not up yet
14:06padenotin any case, I can have a look, yeah
14:06padenotyeah, video quality this week was great
14:07padenotalso, I wasn't using a headset and had zero echo, be it in the standup or just talking 1:1 with mreavy
14:09achronoppadenot: it's up now
14:10padenotachronop, having a look
14:13mreavypadenot: very good
14:16robswain-Mhas anyone ever seen chrome not gather ICE candidates for, say, video, but gather them for audio?
14:24jibfippo: on PTO the rest of this week. Will try to look later today
14:24fippoit has time until next week!
15:26jesupjya: ping
15:26jesupjya: thoughts on bug 1357640?
15:26firebot NEW, Permaorange in test_mediarecorder_record_changing_video_resolution.html | uncaught exception - TypeE
15:29jesupjya: does moving to beta make the pref no longer work?
15:29jesupachronop: ^
15:40padenotthe pushprevenv code is weird
15:40padenotthat's probably the reason
15:41padenotachronop, jesup, if you write an js object litteral with multiple times the same key, only the last one is in the final object
15:41padenotlike, var a = { "key":1, "key":2 }
15:41padenota["key"] is 2
15:42achronopI see
15:43achronophow can I set 2 pref in that case ?
15:44achronopprobably in one array ... like {"set", [[pref1, true], [pref2, true]]}
15:44padenotachronop, I'll upload a patch in a sec, it's easier to write than to explain, and it's not very interesting
15:47jyajesup: I'm not familiar with that pref
15:48padenotachronop, I'll let you review
15:49padenotachronop, see the example here:
15:49achronoppadenot: thanks!!
15:50jesupjya: I think padenot figured it out, thanks
15:51achronoppadenot: that makes sense, that's why it's a 2d array ...
15:52padenotI mean, I did the same kind of mistake before, but the pref I was tweaking was false by default, so my test simply wasn't working
15:53achronoppadenot: landed
15:54padenotachronop, cool, thanks
15:54achronoppadenot: thank you
15:54padenotthis autoland thing really is good for very fast turnaround
15:54* padenot goes back to reviewing the cubeb uplift
16:31padenotachronop, is what I had in mind, what do you think ?
16:33achronoppadenot: why 8?
16:33padenot7.1 input I suppose
16:33padenotit does not matter it's just an addition on the stack pointer
16:33padenotwhat we really want though is to never allocate
16:34achronoppadenot: after DeinterleaveAndConvertBuffer, don't you need to copy write_channels to channels?
16:35padenotin the future, we'll have to remove allocations and locks from the MediaStreamGraph to reduce the standard-deviation on the time it takes to do a callback, so we should be trying to not add allocations and locks
16:35padenotachronop, no, it's pointing to the same data
16:35padenotthe arrays just hold pointers to the actuall channels, one array is read only, one array is read-write
16:36padenotit's cool to enfore their read-onliness, because in Web Audio, we can do optimizations because we know those arrays are written to in very specific locations in the code
16:36padenoton read-only most of the time
16:39achronoppadenot: I see ... other than that the assert expect mono or stereo, do you suggest to remove that and expect up to 8 channels?
16:39padenotachronop, I don't know, but what I know is that we have had 80 crashes because WASAPI asserts when a sound card has more than 2 channels of input
16:40padenotand that the cost to support more than two when you support two is not very high
16:40padenotit's more like mono -> stereo that is complicated
16:41padenotalso, when we support stereo, we'll have crazy people asking for more
16:41padenotand then it won't be enough
16:42achronoppadenot: also do we need to set channels.SetLength(aChannels) for mono? now that we create space for 8 channels in advance
16:42padenotyeah to 1
16:42achronopanything more than stereo will go untested
16:42padenotachronop, it's unit-tested in C++
16:43padenotat least the deinterleave* methods
16:43padenotthe rest, well, we don't test for anything but stereo input // mono input for now
16:43padenotbut at some point, we'll change that
16:43achronopcool, I will update the code and repush ...
17:40padenotachronop, r=
17:40padenotwhat would r= mean
17:45achronoppadenot: ok thanks
20:35ngFx has gone completely non-responsive. back in a minute
20:37padenotng, take a profile and open a bug !
20:58mreavyng: can you add your update to the shared doc?
20:58mreavyng: ^^
20:59ngmreavy: absolutely
21:10ngmreavy: notes are in, I'll fill in the details when I get a working browser
21:11ngall my CPU cores are pegged and the CPU die is sitting at 65 (abnormally low for full load), laptop fans aren't running ...
21:23ngIf anyone else wants to know where Apple has hidden instruments now, it is inside of the bundle itself.
21:27drnong: are you saying you can only launch Instruments now through XCode?
21:28ngdrno: no, thankfully
21:29drnong: so you are just unhappy that you have to download the full xcode to get to Instruments :)
21:29ngTo get to the app you have to right-click on and "Show Contents" then navigate in side the bundle to Applications
21:30ngNo, I am unhappy that I can't seem to launch it with spotlight
21:32ngdrno: when my machine is sitting at almost 600% CPU (on a 4 core?) it is surprisingly hard to navigate in the finder.
21:41drnong: fair enough. when my machine is loaded Alfred also becomes hard to start via my normal keyboard shortcut
21:44nganother datapoint that suggests it may be my machine, an icecc build I did right before the call took 4x as long as it normally does. icemon showed it to be distributing correctly.
21:50drnong: interesting. on my machine watchman is just going crazy
21:52ngdrno: as in the service that watcches for FS changes in hg repos?
21:53drnobut since that is python it only causes nice steady load on 4 core at 25% each
21:53drnobut that was enough to get the temp to 65C as well
21:55ng65C is low for full CPU load on this laptop. It will mostly stay in the 75-99 range during regular builds.
21:57ngAfter a reboot, a build took 5 min 24 sec. Very odd ...
22:00jesupng: clobber??? Even my dual-xeon isn't that fast...
22:00jesupah, icecc
22:00ngicecc is magic
22:00ngWell, magic that makes your office nice and toasty.
22:02ngAFIAK there is no icecc for windows, perhaps there is work underway. ted would probably know
22:13kamidphishrillian: Just need to handle the first, failing case
22:29rilliankamidphish: what does the *apt and &apt bit do?
22:30kamidphishrillian: That's a YAML feature for creating and derefing references.
22:30kamidphishrillian: So I don't have to cut'n'paste the list of packages for each addon: section
22:31rillianwow. I tried reading the yaml spec, and there was a bunch of stuff I'd never seen in a travis config, but it was really hard to read, so I didn't get much out of it.
23:04rilliankamidphish: there's a validator for travis.yml but it looks like it's not up to date
23:04rillianI tried adding a matrix.exclude line, but it didn't get rid of the default job.
23:05rillianAt this point people usually give up and write a script. :/
23:21Caspy7I have a user in #firefox reporting that "GPUProcess" does not occur at all in about:support on Firefox 53 (Windows) though they do report a 3rd process
23:21kamidphishrillian: Maybe writing a script is the answer.
23:22kamidphishCaspy7: Have you asked dvander in #gfx?
23:22dvander Caspy7 do they have any notes in the graphics section of about:support ?
23:23Caspy7dvander: this is theirs
23:23Caspy7Windows 7 SP1, they say
23:26dvanderCaspy7: mysterious. are they still on beta or is it 53 release?
23:26dvanderwhat is the value of layers.gpu-process.enabled in about:config?
23:28Caspy7dvander: true
23:28kamidphishrillian: I tried an exclude line and it didn't work, either.
23:28Caspy7they also said "I'm on EME free and 64bit version btw"
23:31dvanderCaspy7: oh, right. we don't show it unless you're on nightly/aurora.
23:31dvanderwe could stick the pid in there for release builds I guess.
23:33Caspy7dvander: is that to prevent simpleton users from thinking they've found another hack that makes things better (by repeatedly pushing the button)? :)
23:34Caspy7dvander: so it's not in Beta either?
23:36dvanderCaspy7: right, exactly. no reason to entice users into blowing things up :)
23:36dvanderCaspy7: ideally the gpu process deal is, if it's working you don't see anything
23:36Caspy7sounds good
23:38Caspy7perhaps a more technical user, but they agree about the PID thing as they're seeing the 3rd process the process explorer just shows them all as "Firefox"
23:39Caspy7most users don't need to know
23:58kamidphishrillian: Finally travis-ci success!!!!
20 Apr 2017
No messages
Last message: 9 days and 2 hours ago