mozilla :: #media

14 Jul 2017
02:41cpearcekentuckyfriedtakahe: bug 1279171 is clearly not fixed. we still have crash reports for this. maybe dmajor can look into that? I never was able to make any progress on that one.
02:41firebot REOPENED, Crash in CMFPropertyStore::CreatePropertyStore
02:45cpearcekentuckyfriedtakahe: bug 1292287 is still happeneing.
02:45firebot NEW, [DXVA] crash in igdumdim32.dll | CreateDecodeDeviceLH
02:46cpearcekentuckyfriedtakahe: bug 1292262 is still happeneing
02:46firebot NEW, Crash in igdumd32.dll | nvumdshim.dll | DdDestroySurfaceLH
02:47cpearcekentuckyfriedtakahe: bug 1285333 is still happening
02:47firebot NEW, Crash in atidxx32.dll | atiuxpag.dll | CContext::EmptyOutAllDDIBindPoints
02:49kentuckyfriedtakahecpearce: how do we decide to stop caring about gpu process crashes?
02:53cpearcekentuckyfriedtakahe: those crashes are in the content process, I guess they're people who don't get the gpu process?
02:53cpearcekentuckyfriedtakahe: bug 1279171 is a long standing crash inside WMF. we initialize WMF to see if we have an MP3 decoder, and it explodes.
02:53firebot REOPENED, Crash in CMFPropertyStore::CreatePropertyStore
03:04kentuckyfriedtakahemattwoodrow: how do you decode video in the content process?
03:06mattwoodrowkentuckyfriedtakahe: media.gpu-process-decoder=false ?
03:10kentuckyfriedtakahemattwoodrow: I'm not really asking how do I do it, I'm more wondering how people out there in TV land get to that code path
03:10mattwoodrowkentuckyfriedtakahe: If they dont have a gpu process, then theyll decode in the content process?
04:02kentuckyfriedtakahemattwoodrow: how do you end up without a GPU process?
04:02kentuckyfriedtakaheIIUC non-e10s is a "" process
04:15mattwoodrowkentuckyfriedtakahe: That would do it, as would turning off the pref
04:16mattwoodrowor just crashing your gpu process enough so that it doesnt restart
04:16mattwoodrowbut no accel decoding then
04:24kentuckyfriedtakahemattwoodrow: is the GPU process only used when acceleration is enabled?
05:00achronopdjg|kamidphish: did you have a chance to look at the review?
05:03djg|kamidphishachronop: Thanks for the ping. I forgot to do that.
05:06achronopdjg|kamidphish: thanks, suppose to be a straight forward review unless I did something really wrong
05:07djg|kamidphishachronop: Run from cubeb-pulse-rs folder should be all that's required...
05:07achronopright, thanks for the review
05:08djg|kamidphishachronop: Did you investigate porting the changes to channel_layout::undefined made to cubeb_pulse.c to cubeb-pulse-rs?
05:09achronopdjg|kamidphish: yeah and I end up that I do not need to do anything for rust back end
05:10achronopthe corresponding patch for Linux only mute some assert, the actual functionality is there as soon as we do not assert
05:11achronopI end up to the conclusion that rust back end does not assert for that
05:51mattwoodrowkentuckyfriedtakahe: yes
15:27lgrahlHello :)
15:49jesupjya: sorry, didn't get to it last night; crashed. Looking now
16:06jesupjya: fyi, the OMX decoder (and encoder) did work back in B2G for Flame/etc, but probably not relevant anymore.
18:31ngbwc: ping
18:31bwcng: pong
18:32ngbwc: does following look familiar to you?
18:32ngError executing test: [Exception... "Unexpected error" nsresult: "0x8000ffff (NS_ERROR_UNEXPECTED)" location: "JS frame :: http://mochi.test:8888/tests/dom/media/tests/mochitest/pc.js :: PeerConnectionWrapper :: line 780" data: no]
18:32ngbwc: ^
18:33bwcIn specific, no, but this kind of opaque error is much too common.
18:34bwcI guess go looking at pc.js:780.
18:41ngbwc: thanks, **resumes chasing build geese**
18:41mjfbuild geese that are on fire
19:35ngmjf: it works with e10s disabled
19:36mjfng: confirmed.
19:37mjfOk - so environment related and tied to e10s.
19:41ngmjf: e10s debugging wiki page:
19:41ng"This page has been accessed 7,979 times."
20:10lgrahljesup: ping
20:11mjfng: lol
20:11mjfsuch a tease!
20:38jesuplgrahl: pong but have to go in a minute
20:39jesup7984 times :-)
20:49ngmjf: mochitests are working when I perform an in-tree build
20:50lgrahljesup: okay, I'll just send you a bunch of buffered questions and you can answer when you have time.
20:50* jesup afk
21:47mjfFor anyone playing along at home:
21:47firebotBug 1380132 NEW, SSL info in url bar broken when launching from symlinked path or when objdir outside of repo
21:47mjfIf you build objdir outside of source, that bug causes (at least) peerconnection mochitests to fail on OSX. This was a 2 day lesson learned.
15 Jul 2017
No messages
Last message: 6 days and 13 hours ago