[Ffmpeg-devel-irc] ffmpeg-devel.log.20171123

[burek]

[00:23:18 CET] <cone-484> ffmpeg 03Mark Thompson 07master:a050f56c0903: lavu: Add OpenCL hardware pixfmt
[00:23:18 CET] <cone-484> ffmpeg 03Mark Thompson 07master:b25d8ef0a7a8: lavu: OpenCL hwcontext implementation
[00:23:18 CET] <cone-484> ffmpeg 03Mark Thompson 07master:eb3e9f469a52: hwcontext_opencl: VAAPI to OpenCL mapping for Intel i965+beignet
[00:23:18 CET] <cone-484> ffmpeg 03Mark Thompson 07master:403cde8e8510: hwcontext_opencl: QSV to OpenCL mapping for Intel Media SDK
[00:23:18 CET] <cone-484> ffmpeg 03Mark Thompson 07master:9e7cda1b1e79: hwcontext_opencl: DXVA2 to OpenCL mapping
[00:23:19 CET] <cone-484> ffmpeg 03Mark Thompson 07master:f06754e4f3c2: hwcontext_opencl: D3D11 to OpenCL mapping
[00:23:19 CET] <cone-484> ffmpeg 03Mark Thompson 07master:7faae6e745a3: hwcontext_opencl: DRM to OpenCL mapping for ARM
[00:23:20 CET] <cone-484> ffmpeg 03Mark Thompson 07master:5c21c41b7da2: lavfi: Add some common code for OpenCL filtering
[00:23:20 CET] <cone-484> ffmpeg 03Mark Thompson 07master:59d6529333c4: lavfi: Add infrastructure for building OpenCL source into libavfilter
[00:23:21 CET] <cone-484> ffmpeg 03Mark Thompson 07master:9204b2deea85: lavfi: Add OpenCL overlay filter
[00:23:22 CET] <cone-484> ffmpeg 03Mark Thompson 07master:7faa8d8b0383: lavfi: Add OpenCL unsharp mask filter
[00:23:24 CET] <cone-484> ffmpeg 03Mark Thompson 07master:5514bab37a26: vf_hwmap: Pass mapping mode when deriving frames context on an existing device
[00:23:24 CET] <cone-484> ffmpeg 03Mark Thompson 07master:0f93cef2d640: hwcontext: Perform usual uninitialisation on derived frames contexts
[00:23:25 CET] <cone-484> ffmpeg 03Mark Thompson 07master:3650cb2dfa64: lavu,lavfi,ffmpeg: Remove experimental OpenCL API
[00:25:19 CET] <atomnuker> nice
[00:56:13 CET] <cone-484> ffmpeg 03Mark Thompson 07master:79a26ef802d6: tools/cl2c: Add a copyright header
[02:00:13 CET] <cone-484> ffmpeg 03Rostislav Pehlivanov 07master:6e0723470bea: opus_pvq: fix crashing on analysis of mono
[02:58:56 CET] <cone-484> ffmpeg 03Carl Eugen Hoyos 07master:8b62647a5fea: lavf/mov: Read aspect ratio from AVID jpeg2000 mov files.
[03:13:06 CET] <Emerica>  [ mixi         ] [ SortaCore      ] 
[03:13:08 CET] <Emerica> 19:12 [@TimothyGu    ] [ funman        ] [ MrZeus_      ] [ stevenliu      ] 
[03:13:10 CET] <Emerica> 19:12 [@ubitux       ] [ ggherdov      ] [ muken        ] [ stripes416     ] 
[03:13:13 CET] <Emerica> 19:12 [@wbs          ] [ gix           ] [ Myrsloik     ] [ tchebb         ] 
[03:13:21 CET] <Emerica> sorry
[03:14:20 CET] <Emerica>  having a hard time find a client that works on a mac, one quick swipe and it spewed that :/
[03:30:12 CET] <Emerica> https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/mpegtsenc.c#L1180   This flag for forcing the pat and pmt in mpegts files, am I understanding correctly that it's forcing a pat and pmt where there's a keyframe and the last payload wasn't a keyframe?   Is there a good reason to do this besides the first keyframe? Maybe HLS?  I'm curious as the pat and pmt distribution can be smoothed out quite a bit by removing it.
[14:50:22 CET] <jkqxz> Can we remove the gap in the middle of enum AVPixelFormat?
[14:50:33 CET] <jkqxz> (Unstable API/ABI, etc...)
[15:14:52 CET] <durandal_1707> jkqxz: yes, it was forgotten, please do
[15:21:25 CET] <cone-076> ffmpeg 03Derek Buitenhuis 07master:cce33ed10fdc: Revert "udp: fix compilation when HAVE_PTHREAD_CANCEL isnt defined"
[15:25:37 CET] <jamrial> rcombs: have any comments about my AVOutputFormat.deinit() -> close() patch?
[18:12:17 CET] <Tzimmo> I'm trying to debug why DVB-T subtitles in MPEG-TS mostly fail to get detected properly.
[18:12:49 CET] <Tzimmo>     Stream #0:3[0x403], 2, 1/90000: Unknown: none
[18:12:49 CET] <Tzimmo>     Stream #0:4[0x404], 821, 1/90000: Unknown: none
[18:13:21 CET] <Tzimmo> [mpegts @ 0x23df6c0] probing stream 3 pp:2499
[18:13:30 CET] <Tzimmo> 718                 int score = set_codec_from_probe_data(s, st, pd);
[18:13:51 CET] <Tzimmo> (gdb) print/x st->probe_data
[18:13:51 CET] <Tzimmo> $3 = {filename = 0x0, buf = 0x23df190, buf_size = 0xb, mime_type = 0x0}
[18:14:34 CET] <Tzimmo> There is only 11 bytes of data in the buffer... maybe it is not any valid start of a subtitle or something. I tried to read more data before calling this function by
[18:15:04 CET] <Tzimmo> tuning this line
[18:15:05 CET] <Tzimmo>         if (end || av_log2(pd->buf_size) != av_log2(pd->buf_size - pkt->size)) \
[18:16:15 CET] <Tzimmo> but regardless of whether I try to delay the call to set_codec_from_probe_data to get more data to it, it still fails to detect it... I'm puzzled how this is supposed to work.
[18:17:27 CET] <Tzimmo> Any idea what should I do to get more data for these subtitles before giving up? I have increased the probe sizes upto 4 gigabytes but it just doesn't seem to help here at all.
[18:21:31 CET] <nevcairiel> Any well formed ts stream should have code descriptors and not rely on probing
[18:25:22 CET] <Tzimmo> These are taken from DVB transmissions... and cut points are when the program changes. So it just starts from the middle of the stream.
[18:25:39 CET] <Tzimmo> Are those code descriptors the ones within PMT?
[18:26:04 CET] <Tzimmo> Or well, I know that they are there, but is that the place where ffmpeg reads them from?
[18:26:21 CET] <nevcairiel> Those should be repeated throughout the stream, so if you look close enough there should always be one coming up
[18:26:57 CET] <Tzimmo> Yes I know. I have written the whole capturing, parsing the data and splitting software myself.
[18:27:02 CET] <Tzimmo> I'll take a look.
[18:32:04 CET] <nevcairiel> And yeah the basic stream type should be in PMT afaik
[18:34:40 CET] <kierank> ffmpeg doesn't read PMT, it probes
[18:34:43 CET] <kierank> so that's the problem
[18:34:58 CET] <kierank> if you have n audio tracks it never orders them correctly
[18:35:59 CET] <nevcairiel> It does read PMT
[18:36:05 CET] <Tzimmo> yes, well, let's try to get this properly detected first then and worry about any ordering later
[18:36:41 CET] <nevcairiel> There is a long list of codecs it doesn't even have probing for
[18:37:26 CET] <Tzimmo> I have some files that contain PMT and some that don't because I've tried with both options.
[18:39:25 CET] <Tzimmo> Is there an easy way to list PIDs present in the TS file? PMT for this should be 0x0100
[18:43:23 CET] <kierank> nevcairiel: then why does it randomly shuffle the audio
[18:43:36 CET] <kierank> do ./ffmpeg -i udp://@:1234 with a ts and you get audio jumbled
[18:43:45 CET] <kierank> and often misdetected probing for teletext
[18:46:43 CET] <Tzimmo> I may be missing PMT from the recent files for some reason. ID'll investigate and return back later.
[18:47:18 CET] <nevcairiel> Because it does both, if it finds an unknown stream before the PMT is seen, it doesn't just throw it away but creates a stream for it, and stuff gets out of order
[18:48:47 CET] <kierank> ...
[19:03:38 CET] <SortaCore> Direct3D11 is in qsv.c, but not in hwcontext_qsv.c
[19:04:48 CET] <Tzimmo> [mpegts @ 0x2dd26c0] probing stream 4 pp:2500
[19:04:48 CET] <Tzimmo> [mpegts @ 0x2dd26c0] Probe with size=10225, packets=1 detected dvbsub with score=50
[19:05:15 CET] <Tzimmo> now it managed to find the subtitle... did not say anything about PMT at least from what I saw (debug logging enabled)
[19:05:38 CET] <Tzimmo> I'll try again to see if this was just luck or if it really works now
[19:06:45 CET] <Tzimmo> maybe it could be improved by searching for PMT first (with some limit like probe limits currently) and then start from the beginning again?
[19:06:57 CET] <Tzimmo> To keep the stream ordering consistent
[19:21:20 CET] <Tzimmo> Ok, now I got it, I have skipped storing PMT for years.
[19:22:05 CET] <Tzimmo> Is there any way to improve or force detection of some PID to dvb_subtitle?
[19:22:37 CET] <Tzimmo> I can fix this for the future but I have tens of terabytes of stuff without PMT, unfortunately.
[20:01:20 CET] <cone-076> ffmpeg 03Martin Vignali 07master:515555af6c17: avcodec/x86/exrdsp : use ymm constant for pb_80
[20:09:19 CET] <Tzimmo> When detection fails, I added code there to force it to dvb subtitle. Let's see if it fixes this. I can use this modified hacked ffmpeg for these old streams if it works.
[20:25:39 CET] <BtbN> Why is "Mikulas Patocka <mikulas at twibright.com>" sending a bunch of random asm patches to me? And why are they not on the ML, even though it's also a recipient?
[20:26:23 CET] <jamrial> he's probably not subscribed to the list
[20:26:48 CET] <jamrial> not sure who handles the ml queue. llogan i think
[21:03:42 CET] <durandal_1707> and Compn 
[21:45:23 CET] <Compn> sure
[21:45:29 CET] <Compn> i may have forgotten to check the queue
[21:45:34 CET] <Compn> it is holiday here right now
[21:45:36 CET] <Compn> and i must go 
[21:47:09 CET] <BtbN> That don't explains why he decided to explicitly add me as recipient. Or did he just add a bunch of people in BCC or something?
[21:47:14 CET] <cone-076> ffmpeg 03Werner Robitza 07master:5f278620b826: doc/filters.texi: explain infinite looping
[21:59:31 CET] <cone-076> ffmpeg 03Lou Logan 07master:a60b2425c3be: doc/filters: mention (a)loop defaults
[22:41:43 CET] <durandal_1707> now you can make audio filtering with ffmpeg like a pro
[22:45:14 CET] <atomnuker> what's lv2?
[22:46:01 CET] <durandal_1707> atomnuker: i thought you are audio  nerd
[22:46:32 CET] <durandal_1707> lv2 is plugin api, succesor of ladspa
[22:47:26 CET] <durandal_1707> there are bunch of lv2 plugins,  free or paid
[22:49:31 CET] <atomnuker> didn't know there was a successof al ladspa
[22:49:37 CET] <atomnuker> never actually used ladspa
[23:14:39 CET] <cone-076> ffmpeg 03James Almer 07master:ef8979f1992f: avfilter/Makefile: skip building opencl.h when opencl is disabled
[23:15:38 CET] <c3-Win> One can be an audio nerd without being a "linux" audio nerd. ;-)
[23:41:36 CET] <cone-076> ffmpeg 03Karthick J 07master:f335efc0add8: avformat/hlsenc: Minor fix in setting http options for master playlist
[23:42:30 CET] <cone-076> ffmpeg 03Karthick J 07master:241c1192d6b8: avformat/hlsenc: Refactor an inconsistent variable name
[23:58:06 CET] <SortaCore> c3-Win that sounds like orc mischief to me
[23:58:37 CET] <c3-Win> ROTFLOL
[23:59:20 CET] <cone-076> ffmpeg 03Karthick J 07master:efb51c8ebbe2: avformat/hlsenc: Added option to add EXT-X-INDEPENDENT-SEGMENTS tag
[00:00:00 CET] --- Fri Nov 24 2017