[Ffmpeg-devel-irc] ffmpeg-devel.log.20150404

[burek]

[00:00:50 CEST] <kierank> yes
[00:01:00 CEST] <jamrial> anyway, yes, it should be fixed, but your comment above sounds like "we care about pointless features while h264 is thorougly broken" when it's not the case strictly speaking
[00:01:01 CEST] <kierank> or people just find a version of ffmpeg that works
[00:01:20 CEST] <kierank> from what I understand the crashes are ER related
[00:01:23 CEST] <kierank> and libav disabled ER 
[00:01:25 CEST] <kierank> to fix them
[00:08:25 CEST] <kierank> on the bright side threads=1 seems ok
[00:14:40 CEST] <iive> ER as Error Recovery?
[00:15:02 CEST] <iive> or was is error resilience 
[00:19:21 CEST] <jamrial> kierank: ok, so it's not just you reporting. cehoyos seems to have similar issues (ticket 4416) and points to a recent commit being at fault, so if this is actually a recent regression then yeah, it's kinda high priority
[00:47:39 CEST] <cone-952> ffmpeg 03Anton Khirnov 07master:3892bdab9b65: avconv: do not overwrite the stream codec context for streamcopy
[00:47:40 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:4d02dfbde475: ffmpeg: Fix extradata allocation
[00:47:41 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:959894632ae6: Merge commit '3892bdab9b652eb003ab95e167f1765e0b0ea035'
[01:38:40 CEST] <nevcairiel> kierank: if ER is the only reason for those crashes, I believe you can disable it on the configure level
[01:38:55 CEST] <kierank> for some of them I think it's the reason for the crashes
[01:39:06 CEST] <kierank> and I got lots of errors in the build when i tried to disable it in ffmpeg last time
[01:39:21 CEST] <nevcairiel> the google people use that, i figure they keep fixing it when it breaks
[01:40:24 CEST] <kierank> I don't think google uses sliced threads
[01:40:32 CEST] <nevcairiel> i dont like libavs approach to the problem either btw,  just disabling something because you cant bother to fix it, and then calling it "fundamentally broken" is just such a cheap-out .. its always possible to detect the conditions when it would otherwise crash and then just dont do it
[01:41:13 CEST] <kierank> well it's the lack of coordination that is worse
[01:45:02 CEST] <cone-952> ffmpeg 03Anton Khirnov 07master:a4d34e218f54: h264: disable ER by default
[01:45:03 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:6da7625cc395: Merge commit 'a4d34e218f548d381e09c483e8dc6ad18a8d571c'
[01:45:04 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:4b7356ce8f2c: avcodec/h264: reenable ER except with slice threads by default
[01:45:37 CEST] <nevcairiel> wasnt there a comment in there before that it doesnt really work with slice threading anyway
[01:50:20 CEST] <kierank> dunno but there are still tons of bugs in slice threading
[01:50:46 CEST] <nevcairiel> probably because noone uses it :d
[01:51:01 CEST] <nevcairiel> but there you go, slice threading doesnt use ER
[01:58:24 CEST] <kierank> seems to be an improvement 
[01:58:29 CEST] <kierank> i.e no waterfall of segfaults
[01:59:21 CEST] <kierank> oh wait
[01:59:29 CEST] Action: kierank should test the right thing
[02:00:26 CEST] <kierank> way more crashes
[02:00:46 CEST] <kierank> zzuf just causes a waterfall
[02:02:13 CEST] <cone-952> ffmpeg 03Lou Logan 07master:893e61734f96: MAINTAINERS: add Reynaldo as a Twitter maintainer
[02:02:54 CEST] <nevcairiel> but that cant be ER anymore then, unless the disable doesnt work properly
[02:03:11 CEST] <kierank> dunno
[02:03:16 CEST] <kierank> probably more things broke in the merge
[02:03:34 CEST] <nevcairiel> didnt you say libav is also crashing like crazy
[02:03:44 CEST] <kierank> yes both are awful
[02:22:30 CEST] <cone-952> ffmpeg 03Carl Eugen Hoyos 07master:f8413f74a184: lavf/mpeg: Support more audio codecs in Hikvision CCTV recordings.
[02:22:31 CEST] <cone-952> ffmpeg 03Carl Eugen Hoyos 07master:036079ce3431: lavf/mpeg: Support alaw in Hikvision CCTV recordings.
[02:22:32 CEST] <cone-952> ffmpeg 03Carl Eugen Hoyos 07master:c4367f950d5b: Fix codec fps diplay for very small fps.
[02:22:33 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:333eb37a967a: avcodec/h264: silence ER warning when no slice threads are used
[02:22:34 CEST] <cone-952> ffmpeg 03Michael Niedermayer 07master:241913c0d041: Merge remote-tracking branch 'cehoyos/master'
[02:42:00 CEST] <BBB> so, I dont understad
[02:42:16 CEST] <BBB> all this stuff was zzufed quite well
[02:42:19 CEST] <BBB> what changed?
[02:42:27 CEST] <BBB> is this all because of libavs h264 rewrite?
[02:42:30 CEST] <BBB> or am I missing something?
[02:50:55 CEST] <kierank> 1: I think sliced threads were not zzuffed 2: I think because h264 changes, yes
[02:53:49 CEST] <kierank> michaelni: can you add me to whatever outreachy thing I needed to be added to
[02:56:44 CEST] <michaelni> kierank, i think you have to register as mentor or something for outreachy ffmpeg
[02:57:49 CEST] <michaelni> and if you want to mentor yayoi theres a "willing to mentor" button at https://outreachy.gnome.org/?q=view_projects&prg=4&p=612
[02:58:31 CEST] <kierank> I don't think we can allow yayoi
[02:58:55 CEST] <kierank> I think she has to commit to the 40 hours of work
[02:59:06 CEST] <kierank> ok applied to mentor
[02:59:09 CEST] <kierank> I never found that page...
[02:59:46 CEST] <michaelni> yes she should comit to 40h, wasnt aware that she didnt
[02:59:59 CEST] <kierank> I need to ask her
[03:01:57 CEST] <michaelni> also if you get a strange error when loging in in a few days, delete the cookies in your browser, thats a long standing bug in the opw/outreachy thing, i think it wasnt fixed yet
[03:02:28 CEST] <michaelni> the cookies for *gnome.org IIRC
[03:02:49 CEST] <kierank> ok
[03:03:15 CEST] <kierank> michaelni: do you need to approve me?
[03:03:18 CEST] <kierank> or does outreachy?
[03:03:40 CEST] <michaelni> i think outreachy needs to approve
[03:04:17 CEST] <michaelni> could take them till monday i suspect
[03:31:16 CEST] <cone-952> ffmpeg 03Andreas Cadhalpun 07master:f62880bfb8c7: mpeg4: use a self-explanatory #define MAX_NVOP_SIZE
[03:31:17 CEST] <cone-952> ffmpeg 03wm4 07master:511585ce7f72: matroskadec: export cover art correctly
[03:54:01 CEST] <BBB> kierank: I zzufed sliced h264 a few years ago
[03:54:25 CEST] <BBB> (I zzufed basically all threaded code paths, it was utterly painful)
[03:54:26 CEST] <kierank> I see
[03:54:43 CEST] <kierank> dunno if it makes a difference but I'm zzuffing 4:2:2
[03:54:44 CEST] <BBB> also it wasnt just me, there were various people working on it
[03:54:56 CEST] <BBB> I was mostly on the fixing/reporting end of things maybe
[03:55:07 CEST] <BBB> and michael fixed way more than me
[13:47:31 CEST] <cone-945> ffmpeg 03Rodger Combs 07master:f0c85d9b9966: lavf/assenc: style nits
[13:47:31 CEST] <cone-945> ffmpeg 03Himangi Saraogi 07master:4b030025278a: avformat/rtsp: Fix null pointer dereference
[14:08:44 CEST] <Compn> anyone feel like seeing if this in-flight app uses lavc ? someone sent it to mplayer because it is using the mplayer name... http://www.monarch.co.uk/offers/flights/m-player
[14:08:50 CEST] <Compn> carl maybe ?
[14:16:38 CEST] <kierank> you can just download the apk
[14:23:22 CEST] <Compn> that is how to check, yes
[14:32:59 CEST] <kierank> hmm looks like all the segfaults are in edge_emu now
[14:35:47 CEST] <kierank> oh some aren't
[14:46:08 CEST] <Compn> you submitted captures of h264 with signal interference so that these can be fixed in ER ?
[14:46:12 CEST] Action: Compn didnt check the bugs
[14:46:27 CEST] Action: Compn runs
[15:05:42 CEST] <kierank> no but I used zzuf
[15:05:55 CEST] <kierank> which exhibits what are almost certainly the same bugs (broken slices)
[17:54:37 CEST] <Daemon404> oooo slice threads fixes
[18:06:59 CEST] <kierank> Daemon404: lol no
[18:11:24 CEST] <Daemon404> kierank, reword: *some* fixes (maybe)
[18:18:13 CEST] <kierank> oh that slice thread ER fix
[18:18:14 CEST] <kierank> ok
[18:18:37 CEST] <Daemon404> yeah
[18:27:07 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:514d9bb55d4f: avcodec/h264: use goto end instead of return for hwaccel fails as well
[18:27:08 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:98c5db89bb7a: avcodec/h264_slice: make current_slice reset code more strict
[18:27:09 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:429de0432022: avcodec/h264_refs: Do not set reference to things which dont exist
[19:41:10 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:60e7d2d7b9eb: avcodec/h264_mb: assert that frame waiting code actually references 2 fields/1 frame
[19:41:11 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:fc58d5c43b4c: avcodec/h264: Fail for invalid mixed IDR / non IDR frames in slice threading mode
[19:41:12 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:aa6f58dd96f3: avcodec/h264_refs: More completely clear slice contexts in ff_h264_remove_all_refs()
[20:58:16 CEST] <cone-586> ffmpeg 03Carl Eugen Hoyos 07master:072198166e9a: lavf/mov: Do not overread iTunes metadata.
[20:58:17 CEST] <cone-586> ffmpeg 03Michael Niedermayer 07master:afb327caf3d1: Merge remote-tracking branch 'cehoyos/master'
[21:33:23 CEST] <Prelude2004c> hey everyone, ... http://pastebin.com/iaDEFsv8 < - is this a bug ? in a loop it messes up the audio bit rates or is something wrong with my code?
[21:43:08 CEST] <kierank> michaelni: thanks, will continue to fuzz
[22:11:27 CEST] <kierank> anyone remember that thing where google said they'd give free CPU to OSS projects?
[22:12:29 CEST] <nevcairiel> somehow your copy paste tries to copy the terminal colors codes and leaves funky markings in your pastebins
[22:12:56 CEST] <kierank> it's not a copy and paste
[22:12:58 CEST] <kierank> it's typescript
[22:14:08 CEST] <kierank> I'm forced to do that because copy and pasting 18k lines is not easy
[22:15:45 CEST] <nevcairiel> and i thought ffmpeg was smart and disables colors when you pipe or redirect its output
[22:16:51 CEST] <kierank> it's running in gdb though
[22:17:28 CEST] <nevcairiel> guess you successfully managed to confuse it :D
[23:03:54 CEST] <Compn> kierank : sounds familiar but i dont remember details. maybe it was audio only in person meeting ?
[23:04:05 CEST] <kierank> there was definitely a blog post iirc
[23:10:38 CEST] <Compn> kierank : any idea what year they announced it ?
[23:10:44 CEST] <Compn> i am finding nothing :\
[23:10:57 CEST] <kierank> 2013 possibly
[23:12:13 CEST] <Compn> i see intels open source resources, but dont see any cpu, https://software.intel.com/en-us/oss
[23:14:41 CEST] <Compn> ehe amd lists ffmpeg on their website , http://developer.amd.com/tools-and-sdks/open-source/
[23:18:26 CEST] <BtbN> A few free GPUs from nvidia would be nice.
[23:18:31 CEST] <BtbN> But it's nvidia...
[23:27:06 CEST] <Compn> what? no good for bitcoin mining? :P
[23:29:03 CEST] <BtbN> More thinking about actualy beeing able to test h265 encoding.
[23:29:18 CEST] <BtbN> And maybe a nvenc capable fate server.
[23:50:18 CEST] <Compn> ah
[00:00:00 CEST] --- Sun Apr  5 2015