mozilla :: #media

19 Mar 2017
01:41maelcumhey guys. i'm interested in fixing the alsa deficiency of the new sandboxed (for apps, i guess) audio functionality. can't find the code in hg default, though. is it in some branch?
01:44maelcumany mailing list i should look at?
02:00Caspy7maelcum: here's this https://groups.google.com/forum/#!searchin/mozilla.dev.platform/alsa%7Csort:relevance/mozilla.dev.platform/jRAqSTri66I/2Lu7BX4SBgAJ
02:03maelcumyeah, well. i'm *not* interested in bringing alsa up to par if it stays disabled by default.
02:06Caspy7maelcum: maybe we didn't read the same post, but that post did not preclude reenabling it. If it were up to par with PA, then presumably it would be reenabled
02:06maelcumi have a "few" (well, the whole plasma desktop and the graphics drivers...) things i like to compile myself, but not firefox. obviously i'd need to keep it around to take care of bugs, but i don't want to compile / run the dev version all the time.
02:08maelcumwell, sometimes it goes like... you want to sell a decision, you pick the most palatable rationale. if that one goes away, you find another one.
02:09maelcumi don't want that to happen. i want someone to say "if you implement this, alsa will be on by default again".
02:21Caspy7maelcum: then you can talk to the devs about it and pose that question
02:21maelcumright
22:36SingingTreeDo we have any functionality for handling timestamps in our logging that I should use instead of %llu?
22:51benc1what is the nack buffer size in firefox?
22:51benc1how many packets does firefox save to respond to nack requests?
23:26benclost packet is ignored only if it's age is 10,000
23:26benchttps://chromium.googlesource.com/external/webrtc/+/master/webrtc/modules/video_coding/nack_module.cc#23
23:26bencisn't it too big? 10,000 age is around 200 seconds
23:39jyamattwoodrow: found why we get called with stream_chamge
23:39mattwoodrowjya: tell me more
23:40jyaanyway.. like for a 400*300 video MFzt_FRAME_SIZZE
23:40jyacomes out as 400*304
23:40jyaAnd then it has a minimal aperture window set to 0,0,400,300
23:41jyaSeems that for nv12/hw makes all a multiple of 16
23:42jyaThat causes some decode failure, as when it loops through all the output type, it finds none that matches
23:43jyaSo maybe the workaround the issue on win8 where the stream change isn't called, is to, when the first frame is output, if we didn't get a stream change, we just simulate it..
23:44jyaI have no better ideas for a fix, that won't cause unknown regressions
23:44mattwoodrowjya: Cant we figure out that it will get adjusted to 304 before we Init?
23:45jyaThat's one particular example, I'm rather scared of assuming how it will round things under different circumstances.
23:48mattwoodrowThats fair
23:50jyamattwoodrow : Any other suggestions I could try?
23:51mattwoodrowjya: Can you reproduce the win8 issue?
23:52mattwoodrowIts really weird stream_change not being called, since there shouldnt be an output size set at all
23:52jyaYes. On the loaner
23:52mattwoodrowI guess we can simulate the event, and assert that a size got set?
23:52mattwoodrowAssuming that theyre setting the size, just not notifying us
23:52jyaWhere would you check that a size is set?
23:53jyaSeeing that we only set the size of stream change is returned?
23:53jyas/of/if
23:53mattwoodrowjya: We read the size in ConfigureVideoFrameGeometry
23:53jyaSorry on iPad.. keeps correcting stuff.
23:54mattwoodrowSo if we simulate the stream_change, and call ConfigureVideoFrameGeometry ourselves, the size we read better not be 0,0
23:54jyaah that way... I see.
23:54jyaI can try I guess. Easy things to do.
20 Mar 2017
No messages
   
Last message: 101 days and 13 hours ago