[Ffmpeg-devel-irc] ffmpeg-devel.log.20150616

[burek]

[00:28:53 CEST] <philipl> what are the chances the new nvidia driver release fixes any of my problems? Low.
[03:53:55 CEST] <cone-906> ffmpeg 03Christian Suloway 07master:907ac20aa293: avformat/hlsenc: added HLS encryption
[07:32:59 CEST] <philipl> Hey everyone. I've been thinking about how we (mechanically) do code reviews. At my paid job, we use review board for reviews, and I find it a far more pleasant experience than doing reviews by email. It makes it much easier to do inline comments, to track whether fix requests are acknowledged, and whether a change is committed.
[07:33:31 CEST] <philipl> I think it's really great software for doing reviews, and we'd probably benefit a lot from using it, or a similar tool.
[07:34:00 CEST] <philipl> It can send review requests and replies to email so there would still be a record in the mailing list archives.
[07:34:05 CEST] <philipl> https://www.reviewboard.org/
[07:34:08 CEST] <philipl> Just a thought.
[07:35:19 CEST] <BBB> Im partially with that idea
[07:35:28 CEST] <BBB> (I personally like gerrit but I guess I dont care what we use)
[07:35:53 CEST] <BBB> having email notifications is obviously useful, but reviewboards add some extra stuff to reviews imo
[07:36:14 CEST] <BBB> like maintaining comments between iterations, easy ways to see diffs between diffs, etc.
[07:36:24 CEST] <philipl> Yeah. Lots of good stuff.
[07:36:56 CEST] <philipl> Gerrit has strong workflow features like pre-checkin staging and tests, but reviewboard is a far better tool for actually reviewing code.
[07:40:01 CEST] <BBB> sounds good to me then :)
[07:41:51 CEST] Action: BBB goes sleep
[08:39:08 CEST] <j-b> 'morning
[09:10:47 CEST] <durandal_1707> good morning
[12:23:09 CEST] <cone-928> ffmpeg 03James Zern 07master:202188a26c1a: doc/muxers.texi: properly insert {}s in example
[13:03:30 CEST] <cone-928> ffmpeg 03George Boyle 07master:2469ed32c81e: avcodec/flacenc: Fix Invalid Rice order
[13:29:21 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:cf11fd632eef: avformat/mxfdec: Change version byte for JPEG2000 to match mxf.c and RP224v12
[14:10:34 CEST] <ubitux>  int32_t fmt = 'avc1'
[14:10:38 CEST] <ubitux> damn, this is valid code?
[14:11:35 CEST] <nevcairiel> Surprisingly yes
[14:18:30 CEST] <__gb__> dates back from apple's fourcc codes for quicktime
[14:18:56 CEST] <__gb__> imho, not standard C anyway
[14:20:51 CEST] <__gb__> err, not specific to quicktime, but resource names -- pff very old things :)
[14:21:06 CEST] <__gb__> circa mac/68k
[16:02:12 CEST] <ubitux> is av_vda_default_init(fmt_ctx->streams[X]->codec) supposed to work out of the box or i missed a step?
[16:03:39 CEST] <ubitux> i don't see how tweaking the avctx is going to change anything, so i'm a bit confused at why it fails badly (the classic err -12473)
[16:13:24 CEST] <wm4> uh
[16:13:33 CEST] <wm4> don't use the demuxer codec context for decoding
[16:14:10 CEST] <wm4> but other than that, and implementing get_format, yes
[16:14:25 CEST] <nevcairiel> vda is like the easiest to get going
[16:14:29 CEST] <wm4> (you call av_vda_default_init() from get_format)
[16:14:40 CEST] <wm4> did I mention yet that the hwaccel API sucks? especially get_format
[16:15:01 CEST] <nevcairiel> get_format aint too bad, its just not very intuitive
[16:15:38 CEST] <wm4> it could be as simple as having to select a h264/vda-hwaccel specific decoder
[16:15:45 CEST] <nevcairiel> For a long time I didn't even init in get_format, i did that before the _open call, but I changed it some time ago because it works better with streams that change format
[16:16:13 CEST] <nevcairiel> well this allows you to bail out of hwaccel when format goes incompat
[16:16:28 CEST] <nevcairiel> although of course that could be implemented some other way
[16:16:48 CEST] <nevcairiel> i do not like falling back to software in the same context anyway
[16:16:54 CEST] <wm4> the decode call could just return an appropriate status code and return all unconsumed packets or something
[16:19:05 CEST] <ubitux> wm4: ah? i need to create another context?
[16:19:36 CEST] <wm4> ubitux: of course
[16:19:44 CEST] <nevcairiel> you can use the demuxing context, but its deprecated and not recommended
[16:19:53 CEST] <nevcairiel> so yeah, better start on the right path
[16:19:57 CEST] <ubitux> mmh; is there some doc about that?
[16:20:08 CEST] <ubitux> i'm still very much confused on how to use that api
[16:20:13 CEST] <wm4> haha
[16:20:16 CEST] <ubitux> :(
[16:20:22 CEST] <Daemon404> >own devs cant handle api
[16:20:26 CEST] <wm4> "ffmpeg dev attempts to use ffmpeg API"
[16:20:39 CEST] <wm4> makers great banter
[16:20:43 CEST] <wm4> *makes
[16:20:53 CEST] <ubitux> :D
[16:21:09 CEST] <wm4> ubitux: the current VDA API is approximately infinitely better than the old one
[16:21:28 CEST] <ubitux> no doubt about it, but i'm unable to make a distinction between the two anyway
[16:21:45 CEST] <ubitux> why do you need 2 decoding context btw?
[16:21:57 CEST] <wm4> because what happens with midstream parameter changes?
[16:22:05 CEST] <wm4> then you'd have 4 sets of parameters
[16:22:26 CEST] <nevcairiel> its not like its hard to setup the new one, there is an API to copy it
[16:22:36 CEST] <ubitux> nevcairiel: where?
[16:22:43 CEST] <wm4> the user-visible ones in AVCodecContext (1 for decoding, 1 for demuxing), and 2 internal ones (1 for the decoder thread, 1 for the state behind the packet queue)
[16:23:15 CEST] <kierank> 3:20 PM <"wm4> makers great banter
[16:23:19 CEST] <kierank> archbishop of banterbury
[16:23:23 CEST] <kierank> bantersaurus rex
[16:23:42 CEST] <nevcairiel> ubitux: pDecodingContext = avcodec_alloc_context3(NULL); avcodec_copy_context(pDecodingContext, fmt_ctx->streams[X]->codec);
[16:24:30 CEST] <ubitux> ok ok; but how is that necessary to make av_vda_default_init() not fail?
[16:24:41 CEST] <nevcairiel> dunno
[16:24:46 CEST] <nevcairiel> what does it say
[16:24:50 CEST] <nevcairiel> it should log an output when it fails
[16:25:42 CEST] <ubitux> it fails just like if i had no video output (same failure as when i'm in ssh iirc), while ffmpeg works (in the same environment)
[16:25:47 CEST] <ubitux> gonna paste the exact error, just a sec
[16:26:36 CEST] <ubitux> vtDecompressionDuctCreate signalled err=-8973 (err) (Could not select and open decoder instance) at /SourceCache/CoreMedia_frameworks/CoreMedia-1562.235/Sources/VideoToolbox/VTDecompressionSession.c line 1181
[16:26:38 CEST] <ubitux> [NULL @ 0x7fec2b006600] Cannot initialize VDA -12473
[16:27:22 CEST] <wm4> that happens to me too with some videos
[16:27:29 CEST] <wm4> or if another application uses VDA
[16:27:32 CEST] <wm4> (like Safari)
[16:27:44 CEST] <ubitux> the same videos work with ffmpeg
[16:28:19 CEST] <nevcairiel> are you sure ffmpeg uses vda and doesnt just swallow the error?
[16:29:08 CEST] <ubitux> pretty sure it does, but i can recheck
[16:31:01 CEST] <nevcairiel> maybe the app needs to do some magic to init something? I havent actually used it outside of ffmepg yet
[16:31:06 CEST] <nevcairiel> although i really should already
[16:31:10 CEST] <wm4> I had lots of trouble figuring out whether ffmpeg actually used hwaccel
[16:31:19 CEST] <wm4> actually, I think I could never confirm it
[16:31:39 CEST] <nevcairiel> you can really tell by the output format
[16:31:52 CEST] <nevcairiel> its usually nv12 for sane hwaccels or UYVY422 for vda
[16:32:38 CEST] <wm4> yeah, but where does it print this format?
[16:32:55 CEST] <nevcairiel> it usually mentions that when it auto-inserts the scaling filter to convert it
[16:34:24 CEST] <ubitux> arg, i'm an idiot, i wasn't using the same video input
[16:57:46 CEST] <Fyr> how long does it take for developers to confirm and fix a simple bug?
[16:57:54 CEST] <Fyr> *the simplest
[16:59:32 CEST] <wm4> 1 minute
[16:59:45 CEST] <ubitux> wm4 is accepting the challenge
[17:00:02 CEST] <ubitux> so mmh, vda related; cv_pix_fmt_type is not always honored?
[17:00:03 CEST] <wm4> a bug can be easily reclassified as "not simple"
[17:00:22 CEST] <wm4> ubitux: dunno, I suppose it depends on the hw support?
[17:03:39 CEST] <ubitux> :(
[17:05:12 CEST] <wm4> ubitux: what are you trying to achieve?
[17:05:49 CEST] <ubitux> i was just requesting rgb32
[17:06:02 CEST] <cone-928> ffmpeg 03Christian Suloway 07master:7e871d6999fd: avformat/hlsenc: removed empty/unused print_encryption_tag function
[17:06:03 CEST] <cone-928> ffmpeg 03Christian Suloway 07master:918abc19ff74: MAINTAINERS: add myself as hls encryption maintainer
[17:06:07 CEST] <ubitux> and expecting the same pix fmt in CVPixelBufferGetPixelFormatType(frame->data[3]) in the get_frame callback
[17:06:25 CEST] <ubitux> kCVPixelFormatType_32BGRA to be more specific
[17:09:53 CEST] <ubitux> well i guess i'm derping again
[17:12:49 CEST] <wm4> you expect it to decode to rgba?
[17:13:01 CEST] <wm4> I know packed yuv and (on newer systems) nv12 are possible
[17:13:26 CEST] <ubitux> i was assuming it was capable of converting
[17:14:24 CEST] <ubitux> but the problem is elsewhere anyway
[17:54:25 CEST] <Daemon404> innocent soul trying to suggest a code review tool on the ML
[17:54:59 CEST] <kierank> yeah
[17:56:13 CEST] <thardin> have they missed that that's what the ML is for?
[17:57:06 CEST] <Daemon404> just the usual "ML is crude, i prefer web based stuff" thing
[17:57:56 CEST] <nevcairiel> use a web mail client
[17:57:59 CEST] Action: nevcairiel ducks
[18:27:38 CEST] <philipl> innocent soul is me
[18:28:21 CEST] <philipl> I'm used to the cumudgeny response.
[18:50:12 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:36241c405541: avcodec/jpeg2000dec: Fix term_cnt check
[18:50:14 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:20657e05bc26: avcodec/jpeg2000dec: Improve end check
[20:49:51 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:8294ec67b5b1: avcodec/jpeg2000dwt: add special cases for handling length 1 in 5/3 decode
[20:49:52 CEST] <cone-928> ffmpeg 03Michael Niedermayer 07master:34121ca18734: avcodec/jpeg2000dwt: Fix 9/7 IDWT for small sizes
[21:19:06 CEST] <cone-928> ffmpeg 03Paul B Mahol 07master:e0dac4030692: avcodec/codec_desc: rename smv video codec name to smvjpeg to match decoder name
[21:59:39 CEST] <cone-928> ffmpeg 03Andreas Cadhalpun 07master:5b76c82fd7a5: matroskadec: check audio sample rate
[21:59:40 CEST] <cone-928> ffmpeg 03Andreas Cadhalpun 07master:92e79a2f7bf2: matroskadec: validate audio channels and bitdepth
[21:59:41 CEST] <cone-928> ffmpeg 03Andreas Cadhalpun 07master:72654526e448: configure: make makeinfo_html check more robust
[21:59:42 CEST] <cone-928> ffmpeg 03Andreas Cadhalpun 07master:28efeb6502df: doc: avoid incorrect phrase 'allows to'
[22:21:35 CEST] <wm4> michaelni: apparently I'm still getting "full chroma interpolation for destination format 'rgb48be' not yet implemented"
[22:21:42 CEST] <wm4> michaelni: what does it mean?
[22:21:46 CEST] <wm4> (ffmpeg 2.7 and git)
[23:01:07 CEST] <durandal_1707> this faasoft shift uses ffmpeg with its "own" 4gv decoder
[23:01:29 CEST] <durandal_1707> *shit 
[23:11:48 CEST] <michaelni> wm4, it means theres a code path missing ... wasnt aware of that being the case for rgb48be
[23:12:34 CEST] <wm4> (it's needed when writing 16 bit PNGs)
[23:37:04 CEST] <wm4> av_crc_get_table() isn't thread-safe under some configurations
[23:37:06 CEST] <wm4> hate
[00:00:00 CEST] --- Wed Jun 17 2015