[Ffmpeg-devel-irc] ffmpeg-devel.log.20190707

[burek]

[02:13:07 CEST] <yuzo555> Hi, is there any news about when FFmpeg 4.2 will be released?
[07:43:21 CEST] <cone-728> ffmpeg 03Calvin Walton 07master:2bd21b960963: vf_drawtext: Add pkt_pos, pkt_duration, pkt_size as variables
[11:36:27 CEST] <cone-120> ffmpeg 03Paul B Mahol 07master:ca576833e46c: avcodec/cngdec: set .caps_internal
[13:34:40 CEST] <Lynne> lol, youtube provide ttml subs
[14:30:07 CEST] <kierank> durandal_1707: http://samples.mplayerhq.hu/V-codecs/CFHD/
[14:30:12 CEST] <kierank> MT bear one has pframes
[14:31:12 CEST] <durandal_1707> thanks!
[14:31:31 CEST] <kierank> patches are on ml somewhere
[14:35:45 CEST] <durandal_1707> kierank: don't be against bayer in swscale
[14:35:58 CEST] <kierank> durandal_1707: there are so many different variants
[14:36:02 CEST] <kierank> are you going to have  new pixel format per camera
[14:36:04 CEST] <kierank> per bitdepth
[14:36:20 CEST] <durandal_1707> there is only 12bit in cineform
[14:36:42 CEST] <kierank> durandal_1707: sure but it's a slippery slope
[14:37:13 CEST] <durandal_1707> i could output it into 12bit gray, and let users do debayer
[14:37:38 CEST] <Lynne> also an issue: what exactly do you do to bayer to make it normal rgb? bilinear interpolation? spline? neural nets?
[14:38:37 CEST] <kierank> exactly
[14:38:49 CEST] <kierank> https://twitter.com/kierank_/status/1147847340134457346
[14:39:06 CEST] <kierank> all the bayer related stuff is a whole library in itself
[14:39:29 CEST] <kierank> probably as big and complex as ffmpeg
[14:40:31 CEST] <kierank> durandal_1707: getting pframes in would be interesting
[14:40:39 CEST] <kierank> I had a whole hard drive of samples somewhere
[20:17:51 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:c1a430885dc7: lavc/h265_profile_level: Fix DPB size calculation
[20:17:51 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:f344bb2466f8: lavc/h265_profile_level: Do not allow high tier at level < 4
[20:17:51 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:c7142875c38c: h265_metadata: Add option to set the level of the stream
[20:17:51 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:e3424575f588: lavc/h265_profile_level: Add unit test
[20:17:51 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:cd3578a8e4e1: lavc/h265_profile_level: Add new profiles
[20:38:08 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:451a51124d5d: lavu/frame: Improve ROI documentation
[20:38:09 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:d76e2aaf0896: libx264: Update ROI behaviour to match documentation
[20:38:10 CEST] <cone-677> ffmpeg 03Mark Thompson 07master:f6c572bd89e6: libx265: Update ROI behaviour to match documentation
[22:40:05 CEST] <cone-677> ffmpeg 03Amir Pauker 07master:70ba8e2fa0d7: avcodec/error_resilience: set AVFrame decode_error_flags in case of decoding error
[22:40:06 CEST] <cone-677> ffmpeg 03Amir Pauker 07master:f6dece864de4: avcodec/h264dec: set AVFrame decode_error_flags in case of decoding error
[23:17:32 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:1e93f5060f6f: cbs: Allow non-blank packets in ff_cbs_write_packet
[23:17:33 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:7549f0ac1baf: av1_metadata: Avoid allocations and copies of packet structures
[23:17:34 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:45fd7e44a4dd: filter_units: Unref packet on failure
[23:17:35 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:83be17cfcdd9: filter_units: Don't use fake loop
[23:17:36 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:57f9bc90ae09: filter_units: Reindent after previous commit
[23:17:37 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:b0810454e473: filter_units: Avoid allocations and copies of packet structures
[23:17:38 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:36fcdc3fbe08: av1/h264_metadata, filter_units: Count down when deleting units
[23:17:39 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:a72cc47a275a: h264_metadata: Avoid allocations and copies of packet structures
[23:17:40 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:3c8a2a1180f0: h264_metadata: Localize code for display orientation
[23:32:23 CEST] <mkver> jkqxz: Mark, looking again at the changes to ff_cbs_delete_unit, I am thinking whether it would not be better to assert that the unit to be deleted actually exists and make ff_cbs_delete_unit not return anything.
[23:33:22 CEST] <jkqxz> I'm wondering roughly the same thing looking at 12.
[23:33:50 CEST] <jkqxz> ff_cbs_h264_delete_sei_message() is thematically equivalent, so them having different styling is kindof weird.
[23:34:31 CEST] <mkver> So you agree to it?
[23:35:04 CEST] <jkqxz> Are there any current callers of ff_cbs_delete_unit() which might fail?
[23:35:56 CEST] <mkver> Not if av1_metadata errors out if the fragment is empty.
[23:38:06 CEST] <jkqxz> Right.  You've removed most people who look at the return value, and the odd cases left are all looking at a specific unit.
[23:40:01 CEST] <jkqxz> And s/most/all/, once I look more carefully.
[23:40:20 CEST] <jkqxz> Then yes, making it void and av_assert0()ful sounds good to me.
[23:40:30 CEST] <mkver> Ok.
[23:42:00 CEST] <jkqxz> Can you redo 11 with that?
[23:42:18 CEST] <mkver> Yes, I will do it.
[23:42:48 CEST] <jkqxz> 10 is good anyway, I'll apply it now.  Rest I can still look through but some rebasing is probably needed.
[23:43:24 CEST] <cone-677> ffmpeg 03Andreas Rheinhardt 07master:3f81cc8c9458: av1_metadata: Error out if fragment is empty
[23:49:54 CEST] <jkqxz> Hmm.  On 14, how did that ever work?
[23:52:12 CEST] <jkqxz> Oh, ha.  Looks like the bluray streams always have an SEI between the PPS and the slices, which would be the skipped unit.
[23:53:58 CEST] <mkver> Even without the SEI, the slices would have been modified during the same iteration of the loop (there was no continue, after all).
[23:55:01 CEST] <mkver> The change only matters if there are multiple PPS in sequence.
[00:00:00 CEST] --- Mon Jul  8 2019