mozilla :: #media

16 Mar 2017
02:01rilliankinetik, alfredo: meeting?
02:02alfredorillian: yeah.
02:28cpearcekentuckyfriedtakahe: and
02:28kentuckyfriedtakahecpearce: sigh
05:16geraldkinetik: My &#39;r?&#39;s bothering you? I was -->this<-- close to push the updated review! :-P
05:17kinetikgerald: i was worried i wouldn&#39;t get a notification when you were ready for review
05:17geraldkinetik: Fair enough, safer this way.
09:07bwujw_wang: alwu:FYI. A very good comparison method.
09:07firebotBug 1273902 DUPLICATE, Add VP9 hardware decoding support for Nvidia Pascal GPU family & Maxwell GM206 GPU
09:10jw_wangVPU=0% for FF
09:10jw_wanglooks like hw acceleration is off.
13:09jesuppadenot: can we either spin an aurora Try build with your and my logging patches, or a Nightly with aggregate devices disabled? No drifts yesterday
13:09padenotachronop, how would we disable aggregated devices on osx ?
13:09padenotjesup, also yeah
13:10padenotjesup, I can write a patch right now, but that means dropping other stuff
13:10padenotjesup, or we can make an nightly build with aggregated device disabled
13:15jesupI suggest disabling aggregate devices, since we&#39;d been seeing it in Nightly
13:15padenotwe can do a pref
13:16jesupthat might be smart anyways
13:16jesupwe could even land a pref on inbound
13:17padenotthe issue we have here is how to plumb the info accross the API boundary
13:17padenotcubeb&#39;s api boundary, that is
13:17padenotachronop or myself can find a way
13:32padenotI haven&#39;t found a way
13:35jesupa flag on open won&#39;t do it?
13:36padenotwe don&#39;t have a flag mechanism
13:37padenotit&#39;s supposed to be a cross-platform library that abstracts over the OS
13:38padenotwe could do a nasty hack, though
13:53jesupEnvironment var ... ;-) avoids touching the API
14:06bencanyone tried connecting Epiphan HDMI2USB3.0 adapter to Firefox?
14:06bencthere are &quot;clicks&quot; in the audio
14:07bencany other adapter that is supposed to work?
14:09dminorjesup: ping, I think my branch 57 updates are ready for you to look at. I was going to push them (unapplied) to the update repo if that is ok. Also, please let me know if there is something else you would like me to look at.
14:09padenotbenc, haven&#39;t tried
14:09padenotbenc, should work, what platform ?
14:09padenotbenc, also, what firefox version
14:09bencpadenot: Windows 7
14:09bencFirefox 52
14:09benclocal recording with
14:10bencreproduce the clicking sound
14:10bencthe audio is cut every few seconds
14:10bencsame issue in Chrome
14:10padenotbenc, is that using gUM ?
14:10bencyes. that&#39;s the official demo
14:11padenotbenc, interesting, this adapter does not touch input, does it ?
14:11padenotbenc, maybe its clock is wrong (firefox and chrome clock everything using the output device clock)
14:12bencpadenot: how can I test this?
14:12bencto see if the issue is in the adapter or in the browser
14:13padenotbenc, get a nightly build, activate verbose logging for the &quot;cubeb&quot; module, re-do what you do
14:13padenotthis will use a special logger that uses a lock-free queue, to be able to get info without perturbing too much what we&#39;re trying to observe
14:14bencpadenot: get it from here?
14:14padenotbenc, actually, scratch that, it&#39;s going to be in tomorrow&#39;s nightly
14:14benchow do I active logging for the &quot;cubeb&quot; module?
14:14padenotone sec, I&#39;m getting you links + instructions
14:15jesupdminor: great!
14:15jesupdminor: go ahead and push.
14:18bencpadenot: does the team have experience with sdi or hdmi with other adapters like blackmagic or magewell?
14:18padenotbenc, unzip this (no need to install), run firefox.exe like normal, about:config, `logging.config.clear_on_startup`, set to false, `media.cubeb.log_level` to &quot;verbose&quot;, `security.sandbox.content.level` to 0, `logging.config.LOG_FILE` to some paths
14:18padenotbenc, yeah
14:18padenotbenc, at least me
14:19padenotbenc, a link to a package that works
14:19padenotand once those prefs are set, restart Firefox and do you test
14:19padenotsorry for the mess, we&#39;re working to make this better for users
14:20bencpadenot: thanks. I&#39;ll test it locally and than in the studio
14:20bencwhat adpater did you test? did you get good results?
14:21padenotbenc, haven&#39;t tested with Firefox (used them for personal projects), but I can debug and make it working if needed
14:21bencpadenot: will I be able to do this logging with FF 55 when it is released or will it always require nightly?
14:21bencpadenot: what adapter do you have?
14:21dminorjesup: &quot;abort: authorization failed&quot; when I try to push :)
14:22padenotbenc, it was a high-end blackmagic one, can&#39;t remember, it was not mine
14:22padenotalso it was some time ago
14:22bencwe got poor results with blackmagic
14:22bencepiphan has this audio issue
14:22padenotbenc, audio, video, boths ?
14:23bencmagewell works well in firefox but freeze every few seconds in chrome
14:23jesupdminor: have you pushed to that repo before?
14:23bencpadenot: if I remember correctly, blackmagic had bad video quality and audio/video sync issues
14:23dminorjesup: nope
14:23padenotbenc, if you tell me what&#39;s up, we can get fixes in
14:23padenotbenc, we&#39;ve fixed a number of weird USB headsets over the years, for example
14:23jesupand you have push using ssh: right? (in .hg/patches-foo/.hg/hgrc)
14:24dminorjesup: yeah, that was the problem. sorry to bug you!
14:24bencpadenot: appreciate it. I&#39;ll try to get the logging you requested
14:24bencpadenot: what is
14:25padenotbenc, it&#39;s our new and shiny CI infrastructure
14:26padenotbenc, very very nice, you can get builds from amazon machines, and also ssh into a machine to do tests, and stuff
14:27bencnot sure I can ask the studio to use exe from a random url :)
14:27benctommorow it&#39;ll be possible to download nightly?
14:27padenotbenc, yes
14:28padenotthis URL is from here:
14:28padenotbut tomorrow&#39;s nightly will work fine, and probably even the one that will pop up in a couple hours or so
14:28bencpadenot: is it possible to create fake audio/video devices in firefox like in chrome?
14:28padenotbenc, fake as in ?
14:29padenotbenc, do you have a link ?
14:29bencyou can start chrome with &quot;--use-fake-device-for-media-stream&quot;
14:30bencit simulates mic and cam devices
14:30padenothave have some stuff, but not on all platforms
14:30padenotjesup, do you know where we stand on fake devices, ^ ?
14:37bencpadenot: downloaded firefox from your link
14:37benccan&#39;t find the logging.config.LOG_FILE config
14:37padenotright click, create
14:38padenotsorry I was unclear
14:38padenotyou want a &quot;string&quot;
14:38padenotsince you&#39;re specifiying a path
14:41padenotjesup, mreavy, we need this for webrtc and related modules:
14:41firebotBug 1303762 FIXED, Add button to enable logging to about:networking
14:42bencpadenot: I&#39;m getting three files. main.2260, child.1936, child.2744
14:42bencchildren are empty
14:43padenotbenc, do you see the prefs you changed in about:config ?
14:44bencI&#39;m trying to trace this
14:45bencjust to test if logging works
14:45bencthis is the only thing I&#39;m getting in main
14:45benc[Unnamed thread 00D1CB40]: E/cubeb z:/build/build/src/media/libcubeb/src/cubeb_log.cpp:80: Starting cubeb log
14:46jesupbenc: jsut a guess, but a datachannel test wouldn&#39;t use audio....
14:46bencok :)
14:46* jesup hasn&#39;t looked at that test
14:46bencI thought that it will log everything
14:46benctrying audio
14:47padenotbenc, yeah it&#39;s just audio for now
14:49jesupwe have hundreds of different log &#39;modules&#39;.
14:49padenotjesup, what do you think about doing like what the necko people do for logging ?
14:54jesupyou mean like
14:55padenotjesup, yes, and also the bug I linked just a bit earlier
14:55padenotbug 1303762
14:55firebot FIXED, Add button to enable logging to about:networking
14:56jesuppadenot: we already have something like this in about:webrtc (without the &quot;set the file&quot; and module-list options)
14:56jesupwe really need about:logging 1/2 :-)
14:57padenotjesup, yes but it&#39;s not good enough I think
14:57padenotalso yeah you&#39;re right
14:58jesupwhat they implemented there I think *is* about:logging, just hidden inside about:networking
14:59jesupyou can set any log from there it appears
14:59jesupso we don&#39;t need to add it to about:webrtc ;-)
15:00jesupjust file a bug to move that page to about:logging
15:01jesupthe only advantage that has over a generic about:logging is that it pre-fills the modules list with networking ones. a real about:logging page could have a dropdown or radio buttons or checkboxes for adding different presets (networking, DOM, Webrtc, media playback, etc) to the module log field
15:02* jesup will go file that
15:02jesupfor now, tell people to go there, and replace the list of modules with whatever
15:02jesuppadenot: ^
15:02bencgetUserMedia with fake:true should work in windows 7
15:03jesupfake:true should work everywhere -- but it&#39;s not real audio input!
15:03padenotjesup, that&#39;s good
15:03jesupdoesn&#39;t go through cubeb
15:04bencjesup: ok. just wanted to test logging because I don&#39;t have the device here
15:04bencI&#39;ll try with a real mic
15:26* sheppy learns not to search for JSFiddle-bas written by jib by searching on &quot;fiddle jib foo&quot; when all he gets are results with jigs played on the fiddle.
15:26jesuppadenot: bug 1347955
15:26firebot NEW, Move the logging sub-page of about:networking to about:logging
15:27jesupsheppy: probably &quot;Searching for jiddle jig... search for &quot;fiddle jib&quot; instead&quot;
15:27jibif it&#39;s google try &quot;verbatim&quot; mode
15:29jibthat&#39;s all of &#39;em
15:30jibunfortunately jsfiddle I haven&#39;t found any search feature for searching my own fiddle
15:30jibso I rely on awesome bar
15:30* jib will bbl. parent-teacher meeting
15:30sheppyThat&#39;d be nice; a way to search fiddles by username and then keywords.
15:31sheppyjib: good luck :D
15:31jibthat would be awesome
15:31jibit&#39;s not for me :)
15:31* jib hasn&#39;t worried about bad grades for a while
15:32sheppyI just meant, &quot;Good luck; I hope your kid isn&#39;t in trouble.&quot; :)
15:50* sheppy files :)
16:50benchow do I set media.navigator.streams.fake ?
16:50bencI don&#39;t see it in about:config
16:52padenotbenc, create it (right click)
16:52padenotthere is a number of prefs we don&#39;t list, because it&#39;s kind of for internal use
16:52padenotfor unit tests and stuff
16:53bencpadenot: thanks
16:54bencYour browser can not play
16:54benc media clip. event: {&quot;isTrusted&quot;:true}
16:58bencpadenot: this demo cause a bug in nightly with the fake stream
16:58bencin case you are interested
16:58bencjust played with it. not really important to my use case
17:18fippopehrsons: ^^
18:33fippojib: ping. any ideas why GUM in firefox 52 would fail on travis-ci? I can&#39;t reproduce locally unfortunately
18:34jibreal or fake camera?
18:53fippo -- fake device
18:58bencfake device is made in china?
19:42jibjesup: ping
20:01jibfippo: not sure why it would fail. stressing 52 with that test works fine for me
20:11mreavydminor: ping
20:12dminormreavy: pong
20:14mreavydminor: Hey Dan, bsmedberg is organizing an introductory meeting between some members of the new platform QA group (which I believe bsmedberg now manages) and the &quot;webrtc team&quot;. He&#39;d like to start talking about automating some of the webrtc compatibility tests (which are now being done manually).
20:14mreavydminor: he is asking if 10am tomorrow works
20:15dminormreavy: sure, that is fine
20:15mreavyI was thinking that it&#39;d be good for you and jesup to be there. It&#39;s 7am on the west coast; so I figure drno|irccloud can meet them later.
20:15abrdminor is on the east coast? Or is this a different meeting?
20:15abrAh, okay. :)
20:15dminorabr: I&#39;m in Ottawa
20:15mreavydminor: cool, I&#39;ll add you to the invite
20:16drno|irccloudmreavy: I can probably do 7am
20:24mreavydrno|irccloud: I&#39;ll add you as optional -- it&#39;s just a meet and greet. If you wind up staying up late tonight or if getting on the call tomorrow is challenging, don&#39;t sweat it.
20:44mjfbwc: ping
20:53bwc1mjf: pong
20:54mjfI saw your comment on nr_ice_get_local_addresses name change.
20:56mjfA lot more was going on in that function, which is why I didnt change the name. If you think the new name still makes sense given the other functionality happening there, Ill start making those changes.
20:57mjfbwc1: ^
20:59bwc1Yeah, that function should be thought of as taking the array of addrs, applying policy/filtering to it, and then setting it on the ctx.
20:59mjfAlrighty. Fair enough. Ill jump on it.
21:00bwc1The naming should be consistent with that.
22:14mreavykentuckyfriedtakahe: ping
22:15kentuckyfriedtakahemreavy: kia ora
22:17jyamattwoodrow: so i now set the MF_MT_FRAME_SIZE attribute during initialisation, i set get the MF_E_TRANSFORM_STREAM_CHANGE code
22:17jyaany ideas on what else should be set?
22:19mattwoodrowjya: picture region?
22:20mattwoodrowjya: iirc visual studio debugger is smart enough to display all set attributes on the IMF types
22:20mattwoodrowso you should be able to inspect the before/after output types if you break in the right spots
22:22jyaoh good idea...
22:22jyafound that new laptop rather slow to compile, 40 minutes... 48MB/s write speed with the ssd.. no wonder
22:27jyamattwoodrow: oh, it&#39;s on the output type that you need to set MF_MT_FRAME_SIZE right? not just the input one?
22:36mattwoodrowjya: Indeed
22:39jyadoesn&#39;t make a difference. i set it up on both the input and output
22:39jyaVS doesn&#39;t show me the content unfortunately, just load symbols for additional information.
22:39jyabut which symbols
23:15mattwoodrowjya: Can you load the symbols for the mft dll?
23:15mattwoodrowI forget what its called
23:30jyamattwoodrow: the symbols missing are from a mfplat.dll
23:31jyawhen i attempt to load those, i get a form to load the matching pdb
23:31mattwoodrowjya: Do you have the windows symbol server setup?
23:34mattwoodrowHmm, Bas just complained about missing symbols in #developers
23:34mattwoodrowI wonder if thats related
23:34mattwoodrowIm 90% sure mfplat used to resolve
23:35mattwoodrowYeah, Ive got some mfplat.pdb versions in my local symbols cache
23:36jyai guess i&#39;ll have to enumerate all the attributes present, assuming i find a way to do that
23:38mattwoodrowjya: IMFMediaType inherits from IMFAttributes, which has GetCount(), GetItemByIndex(), GetItemType()
23:46jyayes, just finished implementing that, at least i can see the key name
23:59jyamattwoodrow: exact same keys before and after
17 Mar 2017
No messages
Last message: 7 days and 6 hours ago