[Ffmpeg-devel-irc] ffmpeg-devel.log.20190909

[burek]

[00:43:28 CEST] <Lynne> here's the new vulkan patches for anyone to test: https://0x0.st/zJ47.patch https://0x0.st/zJ4h.patch
[09:57:16 CEST] <cone-595> ffmpeg 03Hendrik Leppkes 07master:4e759b56754a: configure: check for a sufficiently recent enough AMF version
[10:04:49 CEST] <cone-595> ffmpeg 03Jun Zhao 07master:a0abcb4a31c3: lavfi/concat: fix logic error in framerate check
[11:08:40 CEST] <cone-595> ffmpeg 03Paul B Mahol 07master:7985430c7091: avfilter/vf_v360: factor vector normalization out
[11:36:56 CEST] <BtbN> We do not have support for svt vp9, do we? Or rather, any svt encoder. All I see is patches for the hevc one on the ML, but they are untouched for a month.
[11:37:47 CEST] <nevcairiel> no i dont think anything of the svt stuff was ever  added
[11:38:14 CEST] <JEEB> yeh
[11:38:26 CEST] <nevcairiel> the code quality was pretty bad when it was initially submitted and fixing took a long time, so by now it probably got a bad reputation :p
[11:39:33 CEST] <BtbN> I'd be very interested in the svt-vp9 encoder. Given how horrible libvpx performs.
[11:40:22 CEST] <BtbN> There is patches here even: https://github.com/OpenVisualCloud/SVT-VP9/blob/master/ffmpeg_plugin/0001-Add-ability-for-ffmpeg-to-run-svt-vp9-with-svt-hevc-av1.patch
[11:40:40 CEST] <JEEB> yea htey had them in their repos all the time
[11:40:57 CEST] <BtbN> A Patch with massive chunks of commented out code
[11:41:15 CEST] <nevcairiel> and some hacks in general code
[11:46:53 CEST] <BtbN> #define AV_PKT_FLAG_SVT_VP9_EXT_ON  0x10000 // Indicating SVT VP9 frame header ext on
[11:46:58 CEST] <BtbN> what? Why not just an option? oO
[11:47:46 CEST] <BtbN> The code itself doesn't look too horrible. But everything around it...
[14:20:39 CEST] <cone-085> ffmpeg 03Paul B Mahol 07master:976617c7d21e: avfilter/vf_v360: extend stereographic projection
[18:07:41 CEST] <cone-453> ffmpeg 03Paul B Mahol 07master:799c07e5c864: avfilter/vf_v360: partialy revert previous commit
[18:07:41 CEST] <cone-453> ffmpeg 03Paul B Mahol 07master:0c9e01011b54: avfilter/vf_v360: add guards against NaNs
[18:21:43 CEST] <cone-453> ffmpeg 03Andrey Semashev 07master:6d9d053edbed: tests: Fix bash errors in lavf_container tests.
[18:21:44 CEST] <cone-453> ffmpeg 03Limin Wang 07master:29bde4b3b6d7: swscale/swscale: delete unwanted assignments
[18:41:16 CEST] <cone-453> ffmpeg 03Paul B Mahol 07master:558265d4bc1a: avfilter/vf_v360: add padding to u/v/ker
[19:08:34 CEST] <J_Darnley> What is the value in h264 that avcodec stores in "first_mb_addr"?  And how does it change with mbaff coding?
[19:10:00 CEST] <J_Darnley> It is read in h264_slice_header_parse()
[19:10:55 CEST] <J_Darnley> The value is sanity checked with a shift using FIELD_OR_MBAFF_PICTURE
[19:11:45 CEST] <J_Darnley> Is it coded as half the value I am expecting?
[19:48:32 CEST] <J_Darnley> adjusting my the of it with FIELD_OR_MBAFF_PICTURE does seem to do what I want
[21:08:39 CEST] <durandal_1707> test
[21:09:25 CEST] <BradleyS> durandal_1707: beep boop
[21:22:13 CEST] <BradleyS>  /topic version
[21:29:16 CEST] <taliho> just saw this today - 
[21:29:17 CEST] <taliho>  https://www.phoronix.com/scan.php?page=news_item&px=FFmpeg-Avformat-ZeroMQ
[21:40:38 CEST] <durandal_1707> typical moronix
[21:50:36 CEST] <tmm1> what is the correct way to handle H264_NAL_PREFIX in cbs_h264_read_nal_unit
[22:07:06 CEST] <Lynne> philipl: are semaphore imports into cuda expensive too?
[22:26:33 CEST] <cone-001> ffmpeg 03Aman Gupta 07master:425be35c2648: avcodec/mediacodecdec_common: warn when PTS is missing
[22:26:33 CEST] <cone-001> ffmpeg 03Aman Gupta 07master:f821d1953151: avcodec/mediacodecdec_common: improve trace logging for end-of-stream
[22:26:33 CEST] <cone-001> ffmpeg 03Aman Gupta 07master:ca613dcc387e: avcodec/mediacodecdec_common: ensure current input buffer is always used correctly
[22:26:33 CEST] <cone-001> ffmpeg 03Aman Gupta 07master:d0e676276856: avcodec/mediacodecdec_common: log codec name during configure/start failures
[22:26:33 CEST] <cone-001> ffmpeg 03Aman Gupta 07master:35b0e75b545d: avcodec/mediacodec_surface: drop unnecessary local variable
[22:48:17 CEST] <cone-001> ffmpeg 03Paul B Mahol 07master:2f071767d9b4: doc/filters: fix typo in v360
[22:57:24 CEST] <philipl> Lynne: probably. I never had a reason to profile it but it would be in-character for them.
[22:57:49 CEST] <philipl> I'll try and find some time to try out your latested patches soon.
[23:05:17 CEST] <cone-001> ffmpeg 03Michael Niedermayer 07master:9d1e98afeecb: tools/target_dec_fuzzer: Adjust threshold for LSCR
[23:12:53 CEST] <cehoyos> https://github.com/Swaraj1998/FFmpeg/commit/7b69e66d
[23:13:18 CEST] <cehoyos> The heif student has produced a patch that allows decoding both for advanced library users and for whoever just wants to decode the images.
[23:14:06 CEST] <BradleyS> very cool
[23:28:47 CEST] <thardin> not all heif variants i presume?
[23:30:27 CEST] <thardin> tho that would be neat
[23:30:29 CEST] <cehoyos> Not all variants?
[23:30:49 CEST] <thardin> heif is a huge standard last time i looked
[23:31:09 CEST] <cehoyos> I tested with a few samples that we had
[23:31:38 CEST] <thardin> it's not just h26* in isobmff, it supports all kinds of shenanigans
[23:31:50 CEST] <cehoyos> Do you have samples?
[23:31:59 CEST] <thardin> no i do not
[23:32:15 CEST] <cehoyos> (We do not support all shenanigans of many standards)
[23:32:24 CEST] <thardin> true
[23:32:37 CEST] <thardin> baseline h264 to name one
[23:32:41 CEST] <cehoyos> My goal was to support what iDevices produce, that may (or may not) be achieved
[23:32:48 CEST] <JEEB> I would be OK with supporting the stuff actually in use in the wild
[23:32:51 CEST] <cehoyos> Or j2k or vc1
[23:33:14 CEST] <JEEB> most iDevices are HEVC I think, although it might also output HEIF with JPEG? although I never got a confirmation of that :)
[23:35:02 CEST] <thardin> do they use that weird album mode?
[23:35:45 CEST] <thardin> like instead of individual files there's support for tossing a whole album into a single file, for some reason
[23:38:10 CEST] <JEEB> yes the spec supports all sorts of "interesting" things, but I would only start worrying about things after finding samples
[23:38:20 CEST] <JEEB> and I'm fine with starting with the most popular sub-set
[00:00:00 CEST] --- Tue Sep 10 2019