[Ffmpeg-devel-irc] ffmpeg-devel.log.20160201

[burek]

[00:00:24 CET] <Timothy_Gu> do you know of a specific reason for it to do that? performance or compliance?
[00:19:47 CET] <kierank> Need to meet more FFmpeg people now that I met J_Darnley 
[00:22:59 CET] <jamrial> Timothy_Gu: performance, maybe instruction size, can't really say
[00:26:42 CET] <Timothy_Gu> anyway agner's tables don't show any performance deficit when using 64-bit lea (or any other instructions) so changed locally
[00:33:53 CET] <J_Darnley> They're usually the same speed, just the size changes
[00:34:14 CET] <cone-985> ffmpeg 03Timothy Gu 07master:014e4e4499fd: vf_phase: Reduce the scope of several variables
[00:41:35 CET] <cone-985> ffmpeg 03popcornmix 07master:def56677e58d: wtv: Speed up wtv index creation
[00:44:33 CET] <cone-985> ffmpeg 03Timothy Gu 07master:180f9a09588d: all: Make header guard names consistent
[00:52:54 CET] <cone-985> ffmpeg 03Timothy Gu 07master:e5a6dcac47c2: wtvdec: Removed unused variable
[00:52:55 CET] <cone-985> ffmpeg 03Timothy Gu 07master:80075b014904: Changelog: add entry on wtvdec performance improvement
[00:55:43 CET] <Timothy_Gu> http://i.imgur.com/jWtUoOM.png 
[00:56:00 CET] <J_Darnley> oooh traffic stats
[00:56:09 CET] <Timothy_Gu> "developers.google.com"
[00:56:34 CET] <Timothy_Gu> hmm apparently it's this ’ https://developers.google.com/web/updates/2013/07/Alpha-transparency-in-Chrome-video
[00:58:30 CET] <kierank> I could probably make that data for trac
[01:31:55 CET] <kierank> BBB: so how do I add a GBRAP12 pixel format?
[01:32:51 CET] <kierank> I guess add it and flag it in swscale
[01:34:05 CET] <wm4> and then see if it explodes
[01:38:14 CET] <kierank> I'm not really sure how to test alpha looks right
[01:38:15 CET] <cone-985> ffmpeg 03Stephen Hutchinson 07master:70742e599b66: libx265: Enable 12-bit encoding
[01:58:27 CET] <BBB> kierank: yeah, that basically
[01:58:44 CET] <BBB> kierank: I believe all planar formats are automatically handled correctly in swscale so you shouldny need any special handling
[01:58:46 CET] <kierank> how do I check alpha looks right?
[01:58:51 CET] <BBB> except marking them as input supported"
[01:58:54 CET] <BBB> check alpha...
[01:58:55 CET] <BBB> dunno
[01:58:58 CET] <BBB> :D
[01:59:07 CET] <BBB> wait for cehoyos to file bugs?
[01:59:31 CET] <J_Darnley> kierank: there's a seprate planes filter
[01:59:48 CET] <kierank> yeah but still unclear whether the alpha plane is correct or not
[01:59:52 CET] <kierank> in cfhd, that is
[02:00:07 CET] <BBB> swap alpha/luma
[02:00:14 CET] <BBB> is best I can think of
[02:03:17 CET] <Compn> we have no alpha fate tests ?
[02:03:19 CET] <cone-985> ffmpeg 03Stephen Hutchinson 07n2.8.6:HEAD: libx265: Enable 12-bit encoding
[02:09:39 CET] <wm4> does swscale really handle >8bit fprm,a
[02:09:41 CET] <wm4> oops
[02:09:46 CET] <wm4> does swscale really handle >8bit formats correctly?
[02:10:19 CET] <kierank> for yuv it seems to
[02:10:23 CET] <kierank> can't comment on rgb
[02:10:26 CET] <kierank> or anything else
[02:13:50 CET] <BBB> I think it does for rgb also
[02:17:22 CET] <wm4> I seem to recall problems with dithering to 8 bit making the picture too green
[02:35:27 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:a2966c7d1f14: swscale/swscale-test: Fix slice height in random reference data creation.
[02:35:28 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:fc0f08f9fb5e: avformat/mxfenc: Do not crash if there is no packet in the first stream
[02:35:29 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:8132ed4a4377: avfilter/vf_mpdecimate: Add missing emms_c()
[02:35:30 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:ffda227636a3: avcodec/h264_refs: Fix long_idx check
[02:35:31 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:41289bc85322: swscale/utils: Fix intermediate format for cascaded alpha downscaling
[02:35:32 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:0affd64b1c21: avcodec/put_bits: Always check buffer end before writing
[02:35:33 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:7ea0e525edc1: swscale/utils: Use normal bilinear scaler if fast cannot be used due to tiny dimensions
[02:35:34 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:8158fb129e15: avcodec/h264_slice: Fix integer overflow in implicit weight computation
[02:35:35 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:593dea80f28e: avcodec/motion_est: Fix mv_penalty table size
[02:35:36 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:5fe8dad4671e: avcodec/mpegvideo_enc: Clip bits_per_raw_sample within valid range
[02:35:37 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:dd285715308f: avformat: Add integer fps from 31 to 60 to get_std_framerate()
[02:35:38 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:f6a503c4438d: avcodec/mss2: Check for repeat overflow
[02:35:39 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:5c0d8a838777: avcodec/mjpegdec: Fix negative shift
[02:35:40 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:d8cb5887c1ca: avcodec/dvdec: Fix "left shift of negative value -254"
[02:35:41 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:60bc36193ee8: avcodec/wavpackenc: Headers are per channel
[02:35:42 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:250e5cb71df4: avcodec/wavpackenc: Check the number of channels
[02:35:43 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:78f9c7dd14be: avcodec/mpeg4video: Check time_incr
[02:35:44 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:937f3058fa23: avformat/asfenc: Check pts
[02:35:45 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:66aeb5467eee: avformat/aviobuf: Fix end check in put_str16()
[02:35:46 CET] <cone-985> ffmpeg 03Maxim Andreev 07release/2.4:70b35708b915: avformat/hls: forbid all protocols except http(s) & file
[02:35:47 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:ed44b57935f6: swscale/yuv2rgb: Factor YUVRGB_TABLE_LUMA_HEADROOM out
[02:35:48 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:38369313b959: swscale/yuv2rgb: Increase YUV2RGB table headroom
[02:35:49 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:990abbd1c612: avformat/hls: More strict url checks
[02:35:50 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:c0df58b0e5ec: avformat/hls: Even stricter URL checks
[02:35:51 CET] <cone-985> ffmpeg 03James Almer 07release/2.4:2a2205b05184: configure: bump copyright year to 2016
[02:35:52 CET] <cone-985> ffmpeg 03James Almer 07release/2.4:a6ef7205e9fb: avcodec/wavpackenc: print channel count in av_log call
[02:35:53 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:e4b2c75c2a69: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_rgb24_wrapper()
[02:35:54 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:af384c870354: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_yv12_wrapper()
[02:35:55 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:f8728dc83417: swscale/x86/rgb2rgb_template: Fix planar2x() for short width
[02:35:56 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:5d40272ba8fc: swscale/swscale: Add some sanity checks for srcSlice* parameters
[02:35:57 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:7142ddcf92c6: avcodec/tiff: Check subsample & rps values more completely
[02:35:58 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:c88fa43a3adf: avcodec/put_bits: Assert buf_ptr in flush_put_bits()
[02:35:59 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:49ae02d36f25: avcodec/gif: Fix lzw buffer size
[02:36:00 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:a9a6e4e9c1f6: avcodec/ass_split: Fix null pointer dereference in ff_ass_style_get()
[02:36:01 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:5af593290488: avformat/avio: Limit url option parsing to the documented cases
[02:36:02 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:0732e7b0eab4: avcodec/mpeg12enc: Move high resolution thread check to before initializing threads
[02:36:03 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:9e44ea7c0f39: avcodec/wmaenc: Check ff_wma_init() for failure
[02:36:04 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:76de78a9dbef: avformat/avformat: Replace some references to filenames by urls
[02:36:05 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:fa9873cce8c5: avcodec/mjpegdec: Check for end for both bytes in unescaping
[02:36:06 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:4f52c0a619ed: doc/demuxers: Document enable_drefs and use_absolute_path
[02:36:07 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:53f5efcae117: avformat/concat: Check protocol prefix
[02:36:08 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:106e0fff2e88: avformat: Document urls a bit
[02:36:09 CET] <cone-985> ffmpeg 03Paul B Mahol 07release/2.4:ac8a265be81a: avcodec/flacenc: fix calculation of bits required in case of custom sample rate
[02:36:10 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:9a1433683cf3: avutil/opt: check for and handle errors in av_opt_set_dict2()
[02:36:11 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:a944744f197a: avcodec/jpeg2000dec: More completely check cdef
[02:36:12 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:a49d870aac6c: MAINTAINERS: remove unmaintained releases
[02:36:13 CET] <cone-985> ffmpeg 03Derek Buitenhuis 07release/2.4:8380f62155d9: mov: Add an option to toggle dref opening
[02:36:14 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.4:3709c43887c8: Update for 2.4.13
[02:45:39 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.7:d2b2c66b17a2: MAINTAINERS: remove unmaintained releases
[02:53:32 CET] <cone-985> ffmpeg 03Stephen Hutchinson 07n2.7.6:HEAD: libx265: Enable 12-bit encoding
[03:06:52 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:0f091f808af7: swscale/swscale-test: Fix slice height in random reference data creation.
[03:06:53 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:9acbe5fa8489: avcodec/aacenc: Check both channels for finiteness
[03:06:54 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:78c9e1f00b4c: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_rgb24_wrapper()
[03:06:55 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:aea2f5a6eeb7: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_yv12_wrapper()
[03:06:56 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:c48296d3bf1c: swscale/x86/rgb2rgb_template: Fix planar2x() for short width
[03:06:57 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:372ea28f6832: swscale/swscale: Add some sanity checks for srcSlice* parameters
[03:06:58 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:524ee420502c: avcodec/tiff: Check subsample & rps values more completely
[03:06:59 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:8394fa26961d: avcodec/put_bits: Assert buf_ptr in flush_put_bits()
[03:07:00 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:2b7f125af78b: avcodec/gif: Fix lzw buffer size
[03:07:01 CET] <cone-985> ffmpeg 03Derek Buitenhuis 07release/2.6:1733981ec3ac: mov: Add an option to toggle dref opening
[03:07:02 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:0b70b546a226: avcodec/ass_split: Fix null pointer dereference in ff_ass_style_get()
[03:07:03 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:4de748119497: avformat/avio: Limit url option parsing to the documented cases
[03:07:04 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:ee5ba0a1ad8b: avcodec/mpeg12enc: Move high resolution thread check to before initializing threads
[03:07:05 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:0d312030d25b: avcodec/wmaenc: Check ff_wma_init() for failure
[03:07:06 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:4950c02d44f9: avformat/avformat: Replace some references to filenames by urls
[03:07:07 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:302a3269d623: avcodec/mpegvideo_enc: Check for integer overflow in ff_mpv_reallocate_putbitbuffer()
[03:07:08 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:517c856d7fea: avcodec/mjpegdec: Check for end for both bytes in unescaping
[03:07:09 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:50ca8b72d57f: doc/demuxers: Document enable_drefs and use_absolute_path
[03:07:10 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:a967e5515717: avformat/concat: Check protocol prefix
[03:07:11 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:3bb83fd03369: avformat/libquvi: Set default demuxer and protocol limitations
[03:07:12 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:e4b4f9e2dc30: avformat: Document urls a bit
[03:07:13 CET] <cone-985> ffmpeg 03Paul B Mahol 07release/2.6:f71e0b798a50: avcodec/flacenc: fix calculation of bits required in case of custom sample rate
[03:07:14 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:361af0a47cee: avutil/opt: check for and handle errors in av_opt_set_dict2()
[03:07:15 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:684e189eb3a5: avcodec/jpeg2000dec: More completely check cdef
[03:07:16 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:d91a03d46ce2: MAINTAINERS: remove unmaintained releases
[03:07:17 CET] <cone-985> ffmpeg 03Michael Niedermayer 07release/2.6:a3a5aedc0782: Update for 2.6.8
[03:27:03 CET] <BBB> omg auto-vectorization is useful after all?
[03:27:16 CET] <BBB> I remember some blog post that the only thing it optimized was memset(0) calls in init functions
[03:27:18 CET] <BBB> (a few years ago)
[03:29:29 CET] <TD-Linux> I used it about a year ago and it worked okay
[03:29:42 CET] <TD-Linux> especially with adequate use of restrict etc
[03:43:23 CET] <cone-985> ffmpeg 03Stephen Hutchinson 07n2.6.8:HEAD: libx265: Enable 12-bit encoding
[03:57:02 CET] <Timothy_Gu> the autovectorization patch makes ffmpeg get 3MB bigger (not --enable-small)...
[04:05:07 CET] <jamrial> after stripping?
[04:19:42 CET] <Timothy_Gu> ya after stripping
[07:28:59 CET] <Gramner> Timothy_Gu: 64-bit registers outside of addressing brackets (and 32-bit registers inside addressing brackets) require the use of a REX-prefix, so the code becomes larger when using 64-bit regs when it's not neccessary. also some more complex instructions are slower when using 64-bit registers on some older cpus
[10:45:27 CET] <cone-454> ffmpeg 03Hendrik Leppkes 07master:f85cc3bf1223: hevc: set profile based on the profile compatibility flags if needed
[10:45:27 CET] <cone-454> ffmpeg 03Bruce Dawson 07master:09b3a42495b4: riffdec: Explicitly null-terminate array to work around VC++ bug
[10:52:20 CET] <cone-454> ffmpeg 03Carl Eugen Hoyos 07master:9ca5b2724010: lavf/adxdec: Add Autodetection.
[11:31:07 CET] <ubitux> so, how are we supposed to behave when facing negative timestamps?
[11:32:44 CET] <wm4> in which situations?
[11:34:02 CET] <ubitux> http://dpaste.com/029FFGR.txt
[11:34:04 CET] <ubitux> this one
[11:35:03 CET] <andrey_utkin> could maillist moderators please approve tonights letter "doc/examples/transcoding cannot into libx264" to ffmpeg-devel and libav-user?
[11:37:15 CET] <wm4> ubitux: depends what the hell mov/mp4 specifies here
[11:46:38 CET] <ubitux> FFS apple decided that a pdf for a specs wasn't a good thing so they put it on a fucking website with shitty js folding and the inability to search in it
[11:46:42 CET] <ubitux> fuck these assholes seriously
[11:46:48 CET] <ubitux> fuck them, i'm so annoyed
[11:47:42 CET] <wm4> I'm sure Daemon404 or Paranoialmaniac know about negative timestamps in mov/mp4
[12:00:26 CET] <nevcairiel> andrey_utkin: i see those mails on the ML already
[12:02:28 CET] <durandal_1707> ubitux: negative timestamps should be properly scaled if not passed unchanged 
[12:02:44 CET] <ubitux> what is the player supposed to do?
[12:03:01 CET] <cone-454> ffmpeg 03Paul B Mahol 07master:0a7379d9cfc0: avfilter/vf_yadif: make use of ctx pointer
[12:03:09 CET] <wm4> probably depends in the player
[12:05:37 CET] Action: durandal_1707 will push nnedi
[12:06:09 CET] <ubitux> wm4: it's a general question, how do you interpret it?
[12:06:35 CET] <ubitux> like, you have the first video ts at -320, and let's say you have the audio at -640, or even +640
[12:06:40 CET] <ubitux> what do you do in these situations?
[12:07:10 CET] <wm4> mpv doesn't care
[12:07:11 CET] <ubitux> do you start by offsetting everything by min_ts?
[12:07:30 CET] <ubitux> and fill the gaps with either silence for audio and black frames for videos?
[12:07:32 CET] <wm4> well if libavformat reports a start time, then all timestamps are offset by that after the demuxer
[12:07:51 CET] <wm4> but either way the player core doesn't care if the first timestamps isn't 0
[12:08:01 CET] <wm4> it's similar to the case after seeking, where the timestamp isn't 0 either
[12:08:04 CET] <ubitux> ah, so you do a pts + stream->start_time?
[12:08:22 CET] <wm4> yeah (pts - start_time)
[12:08:34 CET] <ubitux> why does mpv have this laggy effect then?
[12:08:39 CET] <ubitux> is it another issue?
[12:08:46 CET] <wm4> but it doesn't matter, unless a player 1. expects start time = 0, or 2. has unsigned pts, or 3. has special logic for pts<0
[12:09:09 CET] Action: durandal_1707 gonna write filter that creates own filtergraph
[12:09:11 CET] <wm4> dunno
[12:09:14 CET] <wm4> maybe some bug
[12:10:54 CET] <wm4> ubitux: which was the sample again?
[12:13:14 CET] <ubitux> http://b.pkh.me/LZZX7353.MOV
[12:13:25 CET] <ubitux> ok so, the "laggy" effect is probably normal
[12:13:40 CET] <ubitux> the frame are actually that way
[12:13:54 CET] <ubitux> it appears smooth in QT because they're actually dropping the frames
[12:14:02 CET] <wm4> what
[12:14:15 CET] <ubitux> the first frame in qt is the first non negative one
[12:14:19 CET] <wm4> anyway, might be the camera reducing framerate due to bandwidth then?
[12:14:41 CET] <wm4> maybe there's an edit list thing that discards the frames?
[12:14:48 CET] <wm4> I have no clue about mov though
[12:15:05 CET] <ubitux> yes there is an edit list
[12:15:09 CET] <ubitux> with a single entry
[12:15:19 CET] <ubitux> value=320 (the start time basically)
[12:16:04 CET] <ubitux> so we basically if there is a start time, we should just discard anything before it?
[12:18:46 CET] <wm4> is it possible to mark frames as "do not display"?
[12:19:00 CET] <ubitux> well, i guess that's what pts<0 means
[12:19:08 CET] <ubitux> or even the start time
[12:21:52 CET] <wm4> not sure if I agree with such semantics in general, but it could work anyway
[12:24:18 CET] <ubitux> quicktime behaves like this, so apparently that's how you're supposed to
[12:24:49 CET] <wm4> if it's mov
[12:25:05 CET] <wm4> anyway, I guess there seriously aren't any other uses of negative timestamps
[12:37:59 CET] <kierank> ubitux: how familiar are you with the teletext decoder?
[12:39:31 CET] <ubitux> not much, but you can ask
[12:41:02 CET] <kierank> basically i was wondering how the bitmaps work 
[12:41:15 CET] <kierank> when the subtitles "run along"
[12:41:24 CET] <kierank> when you have one word, then another then another appear
[12:44:04 CET] <ubitux> i don't know, maybe ask tmm1_ 
[12:54:19 CET] <wm4> the lavc API is such that a new bitmap subtitle completely replaces the previous one
[13:04:32 CET] <wbs> ubitux: yes, in mov/mp4, pts<0 doesn't exist, and the start time in this case means "start presenting the content from this timestamp" - mainly used to cut out e.g. audio preroll
[13:05:06 CET] <ubitux> i don't like the "in this case" :(
[13:05:31 CET] <wbs> (and to handle delays with e.g. b-frames)
[13:07:17 CET] <wm4> for audio we have AV_FRAME_DATA_SKIP_SAMPLES
[13:07:20 CET] Action: JEEB is just debugging something similar
[13:07:24 CET] <wm4> which handles this in a container-agnostic way
[13:07:34 CET] <wm4> probably could be used for video too?
[13:07:38 CET] <JEEB> tfxd and timestamp differences between tracks
[13:07:59 CET] <ubitux> i suppose it's a video preroll in this case as well, like, the camera isn't yet hot/in sync (that's why it's "laggy" when you play it)
[13:08:11 CET] <ubitux> (there is no audio)
[13:08:31 CET] <ubitux> i don't understand why they just don't drop the frames though, but whatever
[13:09:15 CET] <Venti^> s
[13:12:50 CET] <cone-454> ffmpeg 03Paul B Mahol 07master:75f3e5e08226: avdevice/lavfi: replace deprecated avpicture_layout
[13:43:20 CET] <cone-454> ffmpeg 03Paul B Mahol 07master:79991b2288a9: avfilter: add nnedi filter
[13:56:32 CET] <JEEB> does the nnedi help or documentation point you towards the weights?
[13:59:14 CET] <wm4> I wonder if this slower or faster than the mpv shaders
[14:00:58 CET] <nevcairiel> C will be slower than a GPU could ever be
[14:03:56 CET] <wm4> not sure about that, the shader is extremely slow and unoptimized
[14:04:54 CET] <cone-454> ffmpeg 03Umair Khan 07master:9effa0125531: doc/ffmpeg: explain properly how -fs works
[14:08:56 CET] <kierank> can someone push my fuzz patch
[14:11:15 CET] <durandal_1707> wm4: its nowhere realtime for SD
[14:16:51 CET] <J_Darnley> kierank: did you post it on the ML?  Is it the "Fixes tickets #5208 and #5209" one?
[14:17:02 CET] <J_Darnley> "[FFmpeg-devel] [PATCH] avcodec/cfhd: Make sure we have an end of header tag before allocating a frame."?
[14:17:14 CET] <durandal_1707> that one
[14:17:14 CET] <kierank> yes
[14:17:21 CET] <kierank> I can't push it currently
[14:17:26 CET] <kierank> on the wrong pc
[14:17:45 CET] <J_Darnley> Sure.  I will.  Let just switch back to master.
[14:21:32 CET] <cone-454> ffmpeg 03Kieran Kunhya 07master:bdd8e02b72e7: avcodec/cfhd: Make sure we have an end of header tag before allocating a frame.
[14:22:52 CET] <J_Darnley> There you go
[14:25:55 CET] <Daemon404> [10:47] <+wm4> I'm sure Daemon404 or Paranoialmaniac know about negative timestamps in mov/mp4 <-- too much
[14:25:57 CET] <Daemon404> and also
[14:26:00 CET] <Daemon404> theyre onlt negative in lavf
[14:26:11 CET] <Daemon404> due to the way it handles edit lists
[14:26:24 CET] <Daemon404> theyre not actually coded negative 
[14:26:44 CET] <Daemon404> indeed timestamps are unsigned in isobmff
[14:27:28 CET] <wm4> fascinating
[14:38:27 CET] <Daemon404> ubitux, pingpong
[14:39:23 CET] <jrosser> isnt there a subtlety that one of mov/mp4 allows -ve timestamps and the other doesnt?
[14:39:58 CET] <Daemon404> not sure what -ve does
[14:40:53 CET] <jrosser> i dont remember this very well (looong time ago) but it was due to coded vs display order
[14:41:09 CET] <jrosser> and what happened at the very start of the file
[14:41:34 CET] <Daemon404> sure, the first PTS in mp4 for e.g. h264 is usually non-zero
[14:41:39 CET] <Daemon404> due to frame reordering
[14:41:49 CET] <Daemon404> usually there is an edit list to compensate
[14:42:00 CET] <Daemon404> lavf applies this in such a way taht it shifts everything back to the fitst entry of teh edit list
[14:42:06 CET] <Daemon404> making the dts negative
[14:42:32 CET] <Daemon404> of course, this is wrong for edit lists with more than one entry.
[14:42:33 CET] <jrosser> my very distant memory is that the defintions of the mov vs mp4 containers are slightly different wrt this
[14:42:36 CET] <Daemon404> yes
[14:42:40 CET] <Daemon404> mov timestamps are signed
[14:42:42 CET] <Daemon404> mp4 are not
[14:42:50 CET] <Daemon404> however, you can do the same in mov
[14:42:56 CET] <Daemon404> multiple ways to do everything in mov/mp4
[14:42:59 CET] <Daemon404> it's like the perl of containers.
[14:45:03 CET] <wm4> write-only?
[14:46:47 CET] <Daemon404> that said its still way less painful than most containers
[14:46:49 CET] <Daemon404> if you use a good subset
[14:46:52 CET] <JEEB> yeah
[14:46:56 CET] <JEEB> ENOBIFS
[14:47:02 CET] <JEEB> iä iä BIFS f'thagn
[14:47:16 CET] <JEEB> (that's what they recommend to use for chapters, crazy folk)
[14:50:37 CET] <Paranoialmaniac> on player, negative timestamps in mp4/mov make no sense. you can see there are always implicit or explicit edits on presentation and the spec does not refer to the start time i.e. arbitrary. there are maps from media timelines to a presentation timeline
[14:51:02 CET] <Paranoialmaniac> that's all
[14:51:35 CET] Action: Paranoialmaniac hides
[14:51:38 CET] <Daemon404> :P
[14:51:40 CET] <Daemon404> hmm
[14:51:41 CET] <Daemon404> "Dear FFmpeg developer,
[14:51:41 CET] <Daemon404> we are conducting a short survey to study the characteristics of
[14:51:42 CET] <Daemon404> developer roles that exist in open-source communities."
[14:51:58 CET] <Daemon404> sure do get a lot of emails about FOSS studies
[14:52:03 CET] <Daemon404> 2-4 a year
[14:56:42 CET] <BBB> Daemon404: I think we all got that one
[14:56:50 CET] <Daemon404> BBB, yeah
[14:56:56 CET] <BBB> Daemon404: typical hi Im doing a masters in business school please help me do my homework
[14:57:04 CET] <BBB> heres a question my teacher is asking me to ask someone"
[14:57:09 CET] <BBB> $lawyer_legal_business_language
[14:57:14 CET] <BBB> ktnxbye"
[14:57:18 CET] <ubitux> Daemon404: timecode?
[14:57:24 CET] <Daemon404> ubitux, aye
[14:57:30 CET] <Daemon404> michael reviewed 2 of 3
[14:57:34 CET] <Daemon404> ffprobe change needs review
[14:57:39 CET] <ubitux> Daemon404: any reason not to set it to every frame btw?
[14:57:43 CET] <ubitux> :p
[14:57:51 CET] <Daemon404> because its not the timecode for every frame?
[14:57:55 CET] <Daemon404> it only applies to one frame per gop
[14:57:59 CET] <ubitux> but you can compute the missing ones :)
[14:58:15 CET] <Daemon404> that doesnt seem like our job.
[14:58:27 CET] <ubitux> depends if you want to print them in drawtext :°
[14:58:31 CET] <ubitux> anyway, no problem :)
[14:58:38 CET] <ubitux> if you obtain matching timecode with ffprobe, it's fine with me
[14:58:43 CET] <Daemon404> yeah i do
[14:58:44 CET] <Daemon404> i checked
[14:58:49 CET] <Daemon404> i actually think it is more useful
[14:58:54 CET] <Daemon404> since the old way only showed the first
[14:58:55 CET] <Compn> Daemon404 : we should ask the students for the teachers name and then talk to the teacher and ask him why hes making his studnets bug FOSS projects
[14:58:55 CET] <ubitux> well then it's fine with me
[14:58:59 CET] <ubitux> please add a Changelog entry though
[14:59:02 CET] <Daemon404> i did
[14:59:07 CET] <Daemon404> oh wait i didnt
[14:59:07 CET] <ubitux> great
[14:59:09 CET] <Daemon404> lol.
[14:59:18 CET] <ubitux> no more comment from me, thanks
[14:59:21 CET] <Daemon404> al lright
[14:59:25 CET] <ubitux> we have ffprobe tests btw
[14:59:30 CET] <Daemon404> i know
[14:59:37 CET] <ubitux> you might want to add one at some point, but i feel like you're fed up with it :p
[14:59:38 CET] <Daemon404> the only part im unhappy about is where i put the code in mpeg*.c
[14:59:41 CET] <Daemon404> as noted in the patch
[14:59:45 CET] <Daemon404> nobody suggested a better place though
[14:59:57 CET] <ubitux> at the end at got output seems ok with me
[15:00:09 CET] <ubitux> but i'm not a mpeg expert :p
[15:00:11 CET] <Daemon404> seemed kinda "high level"
[15:00:13 CET] <Daemon404> but eh.
[15:01:34 CET] <Daemon404> BBB, weirdly this has a siemens url
[15:01:46 CET] <kierank> probably a student
[15:01:49 CET] <Daemon404> yeah
[15:18:09 CET] <BBB> do we have a way to skip the first N frames from an input file in ffmpeg?
[15:18:16 CET] <BBB> like -ss, but in frame number instead of time
[15:18:45 CET] <Daemon404> BBB, -r 1, then 1s = 1 frame
[15:18:52 CET] <Daemon404> ive ysed this trick to trim e.g. .h264 before
[15:18:57 CET] <Daemon404> but YMMV
[15:19:01 CET] <Daemon404> depending on use case
[15:19:03 CET] <BBB> does that work for input which has a framerate?
[15:19:07 CET] <BBB> my input is ffv1/mkv
[15:19:16 CET] <Daemon404> probably not
[15:19:21 CET] <Daemon404> remember that lavf is not frame accurate
[15:19:36 CET] <Daemon404> (i use an external library for this)
[15:22:22 CET] <Compn> BBB : if ffmpeg has framestep filter ?
[15:22:37 CET] <Daemon404> filter doesnt do jack if you want to remux.
[15:22:42 CET] <Compn> ah
[15:22:45 CET] <Daemon404> or dont want to decode all the preceding frames
[15:23:09 CET] <BBB> hm I guess I have to write my own helper for this then
[15:23:34 CET] <Daemon404> frame -> time is pretty simple
[15:24:58 CET] <Daemon404> assuming CFR.
[15:25:10 CET] <Daemon404> otherwise youll need the index
[15:25:41 CET] <BBB> I did frame/fps
[15:25:45 CET] <BBB> and its off by a few frames
[15:26:55 CET] <Daemon404> not sure how we seek in mkv
[15:27:49 CET] <Daemon404> (and i never use -ss)
[15:31:23 CET] <cone-454> ffmpeg 03Michael Niedermayer 07master:af24b1c0cd3c: Revert "avformat/hls: Require the file extension to be m3u / m3u8 for probing to succeed"
[15:36:07 CET] <BtbN> ah, so that's why twitch streams stopped working
[15:47:42 CET] <wm4> what did twitch do?
[15:48:15 CET] <BtbN> Nothing, but they have a massive amount of query strings behind the .m3u8
[16:13:40 CET] <cone-454> ffmpeg 03James Almer 07master:77b5b952470a: avcodec/dca_core: rename get_vlc function
[16:13:41 CET] <cone-454> ffmpeg 03James Almer 07master:f781bf3e101b: fate: re-enable dca-xll test
[16:55:38 CET] <atomnuker> michaelni: I got an email about the dirac HQ profile decoding
[16:56:43 CET] <atomnuker> the tables we have support up to 128 quantization values
[16:57:09 CET] <atomnuker> but the HQ profile spec does allow up to 8 bits to be used (different from the low delay mode which only uses 7 bits)
[16:58:24 CET] <atomnuker> michaelni: can you generate new quanziation tables (with taking overflows in account like last time) if I sent you the part of the spec that specifies how?
[17:01:18 CET] <michaelni> if the equations are the same then further values would not fit in 32bit ints
[17:01:37 CET] <atomnuker> oh yeah, forgot about that
[17:50:44 CET] <Timothy_Gu> kierank: is avdev still open to general use?
[17:50:55 CET] <kierank> erm probably but I reinstalled it
[17:54:25 CET] <atomnuker> can someone compile without optimizations and test if fate-dirac still passes?
[18:00:45 CET] <atomnuker> nvm it works
[18:00:51 CET] <BBB> blegh that sucked
[18:00:59 CET] <BBB> I ended up writing my own ffmpeg.c sort of thing :D
[18:01:19 CET] <BBB> I have to admit that our API usability has gone up a fair deal, I find it fairly easy to get things done with it nowadays
[18:53:17 CET] <cone-454> ffmpeg 03Sebastian Dröge 07master:e3a125c970db: Revert "do not write f2 if not interlaced"
[19:30:04 CET] <wm4> nevcairiel: time to push that main10 patch?
[19:33:43 CET] <cone-454> ffmpeg 03Derek Buitenhuis 07master:66e9d2f44ee0: avutil: Add GOP timecode frame side data
[19:33:44 CET] <cone-454> ffmpeg 03Derek Buitenhuis 07master:792a5cefbe53: mpeg12dec: Export GOP timecodes as side data
[19:33:45 CET] <cone-454> ffmpeg 03Derek Buitenhuis 07master:b62825a48051: ffprobe: Deprecate stream timecode field and add frame side data timecode field
[19:35:43 CET] <kierank> Gramner: can you review J_Darnley's patch when you have time please :)
[20:41:46 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:d1fc87529f38: swscale/swscale-test: Fix slice height in random reference data creation.
[20:41:47 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:b515a23f7628: avcodec/aacenc: Check both channels for finiteness
[20:41:48 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:262192a48b59: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_rgb24_wrapper()
[20:41:49 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:93c675d6a6c9: swscale/swscale_unscaled: Fix odd height inputs for bayer_to_yv12_wrapper()
[20:41:50 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:9631209eeaf0: swscale/x86/rgb2rgb_template: Fix planar2x() for short width
[20:41:51 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:0f956cde937b: swscale/swscale: Add some sanity checks for srcSlice* parameters
[20:41:52 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:dee25a5fa5da: avcodec/tiff: Check subsample & rps values more completely
[20:41:53 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:22e20a1d833e: avcodec/put_bits: Assert buf_ptr in flush_put_bits()
[20:41:54 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:9f30eafd0f31: avcodec/gif: Fix lzw buffer size
[20:41:55 CET] <cone-454> ffmpeg 03Derek Buitenhuis 07release/2.5:dd957b56e614: mov: Add an option to toggle dref opening
[20:41:56 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:7ee0b1937a25: avcodec/ass_split: Fix null pointer dereference in ff_ass_style_get()
[20:41:57 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:30463a0c9985: avformat/avio: Limit url option parsing to the documented cases
[20:41:58 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:3fc75e79cf03: avcodec/mpeg12enc: Move high resolution thread check to before initializing threads
[20:41:59 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:65bb07d4bed2: avcodec/wmaenc: Check ff_wma_init() for failure
[20:42:00 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:2df2c0aab0ca: avformat/avformat: Replace some references to filenames by urls
[20:42:01 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:0ec1ffcb4db4: avcodec/mpegvideo_enc: Check for integer overflow in ff_mpv_reallocate_putbitbuffer()
[20:42:02 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:58ea532cad81: avcodec/mjpegdec: Check for end for both bytes in unescaping
[20:42:03 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:e46999ccf440: doc/demuxers: Document enable_drefs and use_absolute_path
[20:42:04 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:dde76f2d0414: avformat/concat: Check protocol prefix
[20:42:05 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:0251cd6cf328: avformat/libquvi: Set default demuxer and protocol limitations
[20:42:06 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:7bcf142c02e0: avformat: Document urls a bit
[20:42:07 CET] <cone-454> ffmpeg 03Paul B Mahol 07release/2.5:d7c0287fbdac: avcodec/flacenc: fix calculation of bits required in case of custom sample rate
[20:42:08 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:3fa6ecca7632: avutil/opt: check for and handle errors in av_opt_set_dict2()
[20:42:09 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:69e191f854f7: avcodec/jpeg2000dec: More completely check cdef
[20:42:10 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:5eca7ba16b8c: MAINTAINERS: remove unmaintained releases
[20:53:49 CET] <cone-454> ffmpeg 03Michael Niedermayer 07release/2.5:60eebbbf22f3: Update for 2.5.11
[22:17:31 CET] <jamrial> nevcairiel: tiny_pnsr seems only able to compare s16 pcm audio
[22:17:53 CET] <nevcairiel> thats what all those lossy tests use
[22:18:01 CET] <nevcairiel> converts to s16 and then compares that
[22:18:29 CET] <jamrial> wouldn't it be better to stick with s24 if possible?
[22:18:49 CET] <nevcairiel> it comes out of the decoder as float
[22:19:08 CET] <nevcairiel> (unlike libdcadec, he stuck to a pure float chain)
[22:19:17 CET] <jamrial> ah true
[22:19:39 CET] <nevcairiel> tiny_psnr actually supports float
[22:19:45 CET] <jamrial> did you reuse the pcm reference files (merged into multichannel files) or made new ones?
[22:19:55 CET] <nevcairiel> make fate-dca GEN=1 =p
[22:20:09 CET] <nevcairiel> his are wav's
[22:20:13 CET] <jamrial> haha
[22:20:15 CET] <nevcairiel> would need careful mangling :(
[22:20:32 CET] <nevcairiel> but i confirmed its bitexact to libdcadec wrapper
[22:20:36 CET] <nevcairiel> so ..
[22:20:38 CET] <jamrial> well, you could always use amerge filter taking all mono files as input and barfing a multichannel pcm file
[22:23:36 CET] <nevcairiel> his files are also 24-bit wavs which  we cant use anyway
[22:23:40 CET] <nevcairiel> so not sure there is anything to gain
[22:37:53 CET] <jamrial> nevcairiel: forcing the tests to use f32le results in somewhat bigger reference pcm files. about ~550kb extra
[22:37:56 CET] <jamrial> not sure if worth it
[22:38:25 CET] <nevcairiel> and would probably need careful tuning of the compare functions
[22:38:35 CET] <nevcairiel> who knows how big the difference in float can be on some systems
[22:39:09 CET] <nevcairiel> all the float decoders use the pcm s16 mode and oneoff compare, so it seems fine for the lossy tests
[22:39:20 CET] <nevcairiel> i made the lossless tests stick to their original bitdepth
[22:39:42 CET] <jamrial> yeah
[22:40:14 CET] <nevcairiel> curious that there is no stereo f ile in the bunch
[22:40:16 CET] <nevcairiel> but oh well
[22:40:42 CET] <jamrial> one of the 5.1 files has stereo downmix coeffs, so your idea to test that is good
[22:40:52 CET] <nevcairiel> yeah i noticed
[22:41:08 CET] <nevcairiel> those with _0 _1 endings usually have one with coeffs
[22:43:26 CET] <nevcairiel> i guess i'll just pick 2 or 3 which I know have coeffs to test that code path
[23:10:57 CET] <nevcairiel> there posted a new patch with a select number of downmix tests
[23:52:09 CET] <Daemon404> :D
[00:00:00 CET] --- Tue Feb  2 2016