[Ffmpeg-devel-irc] ffmpeg-devel.log.20180825

burek burek021 at gmail.com
Sun Aug 26 03:05:04 EEST 2018

[13:37:50 CEST] <durandal_1707> how should i name my blog site?
[13:46:34 CEST] <thardin> I see kostay updated his blag
[13:46:39 CEST] <thardin> kostya*
[13:46:55 CEST] <thardin> https://codecs.multimedia.cx/2018/08/nihav-some-progress-to-report/ tl;dr rust still not as fast as c
[13:47:35 CEST] <nevcairiel> all that safety doesnt come for free, shocker =p
[13:49:47 CEST] <thardin> heh
[13:54:34 CEST] <durandal_1707> huh, x262 still wants yasm
[13:54:57 CEST] <JEEB> I don't think that branch was updated
[13:55:04 CEST] <durandal_1707> since ages
[13:55:12 CEST] <JEEB> yup
[14:05:24 CEST] <durandal_1707> still uses old APIs
[14:23:50 CEST] <durandal_1707> also really broken, mpeg mode causes SEGV here
[14:26:48 CEST] <durandal_1707> disabling asm make it not crash
[15:57:03 CEST] <kierank> durandal_1707: what command
[15:57:39 CEST] <kierank> Shouldn't segfault afaik
[16:01:12 CEST] <durandal_1707> kierank: ./x262 --mpeg -o out.264 ~/Videos/derf/akiyo_cif.y4m
[16:01:41 CEST] <kierank> Yes that produces mpeg4
[16:01:46 CEST] <kierank> I want to produce mpeg2
[16:02:18 CEST] <kierank> iirc we had to make some breaking changes to h264 encoding
[16:03:39 CEST] <durandal_1707> well, it is mpeg2(disabled asm): Stream #0:0: Video: mpeg2video (Main), yuv420p(tv, progressive), 352x288 [SAR 1:1 DAR 11:9], 29.97 fps, 29.97 tbr, 1200k tbn, 59.94 tbc
[16:05:46 CEST] <kierank> Ok weird
[16:10:39 CEST] <cone-283> ffmpeg 03Jun Zhao 07master:00286673d97c: doc/examples/hw_decode: re-licensed to MIT-license.
[16:10:40 CEST] <cone-283> ffmpeg 03Jun Zhao 07master:5f090727adda: doc/examples/vaapi_encode: re-licensed to MIT-license.
[16:10:41 CEST] <cone-283> ffmpeg 03Jun Zhao 07master:d6915d0d4510: doc/examples/vaapi_transcode: re-licensed to MIT-license.
[17:08:00 CEST] <durandal_1707> finally configure fix
[17:09:29 CEST] <JEEB> oh, avih posted his thing
[17:11:50 CEST] <JEEB> but yea, if things still work for everyone then I must say that's a nice speed-up
[17:12:36 CEST] <kierank> what is point of weird indentation rukes
[17:26:58 CEST] <jamrial> "(...)commit XXX changed it to keep the first. This commit restores the original order due to *TBD*."
[17:27:03 CEST] <jamrial> from one of those patches :p
[17:27:53 CEST] <JEEB> yes, the e-mail notes that if the order matters there's a patch to get the pre-patchset order
[17:28:03 CEST] <JEEB> and if it is not required then that patch doesn't have to be applied
[17:28:19 CEST] <JEEB> pre-emptive posting
[17:46:29 CEST] <kierank> michaelni: how can I submit a fate test
[17:47:47 CEST] <kierank> michaelni: also are you going to commit your patch
[17:56:51 CEST] <jamrial> kierank: if you mean a sample, just upload it somewhere, link it here, and i'll add it to the samples suite
[17:57:35 CEST] <kierank> jamrial: thanks, I'm not fully sure how to add an mpeg4 fate test anyway
[17:59:12 CEST] <jamrial> what is it meant to test? decoding bitexactness, frame/stream props?
[17:59:30 CEST] <jamrial> look at other tests and basically just copy paste what they do
[17:59:37 CEST] <kierank> mpeg4 only has vsynth I think
[18:03:39 CEST] <kierank> jamrial: https://0x0.st/syD9.mkv --> can you call it mpeg4_sstp_dpcm.m4v
[18:12:03 CEST] <jamrial> kierank: done
[18:26:06 CEST] <michaelni> kierank, which patch do you want me to commit ? 
[18:26:27 CEST] <kierank> The one with the padding checks
[18:26:32 CEST] <kierank> I think it's needed
[18:30:45 CEST] <durandal_1707> that patch iirc just silences red text
[18:31:11 CEST] <michaelni> kierank, will commit, thx for reminder
[18:31:36 CEST] <cone-283> ffmpeg 03Timo Rothenpieler 07master:0c71c6d66f9a: avcodec/utils: document allocation requirement of extradata
[18:31:43 CEST] <kierank> durandal_1707: dunno I don't want to try and understand that crazy code
[18:37:44 CEST] <durandal_1707> i RE yet another screen capture codec, and whole algo is again in my head
[18:43:09 CEST] <atomnuker> speaking of that, doesn't zlib do RLE anyway?
[18:43:34 CEST] <durandal_1707> atomnuker: tell them
[18:45:27 CEST] <kierank> durandal_1707: maybe you're right michaelni code is useless
[18:49:33 CEST] <michaelni> without correct slice end detection error concealment will not function completely and there would be artifacts from the concealment code thinking that there is an error. If it does work without the patch then thats probably due to the EC code not being run
[18:50:29 CEST] <kierank> durandal_1707: yes you are right patch is useless
[18:50:43 CEST] <kierank> error concealment doesn't work in 10-bit mode anyway
[18:50:46 CEST] <kierank> it's disabled for studio
[18:51:01 CEST] <cone-283> ffmpeg 03Michael Niedermayer 07master:168d8d56bfb0: avcodec/mpeg4videodec: Fix slice end detection in mpeg4_decode_studio_mb()
[18:54:35 CEST] <durandal_1707> bitpacked decoder is misplaced in libavcodec/alldecoders.c
[18:57:31 CEST] <michaelni> it would make sense in principle to extend the EC code to support all variants the decoder supports
[18:58:42 CEST] <kierank> that would require a 10-bit aware mpegvideoc
[18:58:47 CEST] <kierank> which is a huge piece of work
[19:03:16 CEST] <michaelni> the 10bit EC code could skip over the complex mpegvideo involving part and just do a simple memcpy from the previous frame where damage is found
[19:03:41 CEST] <michaelni> i mean copy on a macroblock base not copy of the whole frame
[19:06:11 CEST] <kierank> I wonder if i've been blocked by videolan git again
[19:07:11 CEST] <kierank> ER needs a complete redesign
[19:08:41 CEST] <kierank> durandal_1707: can you push my patch
[19:08:45 CEST] <kierank> I think I am blocked again
[19:09:02 CEST] <durandal_1707> im busy writing decoder
[19:09:27 CEST] <durandal_1707> after i finish it
[19:11:28 CEST] <kierank> ok
[19:38:59 CEST] <cone-283> ffmpeg 03Rostislav Pehlivanov 07master:2e23cfbb4246: allcodecs: reorder ff_bitpacked_decoder
[19:39:03 CEST] <atomnuker> durandal_1707: thanks for telling me, fixed
[21:42:57 CEST] <JEEB> oh the 0p1pp1 guy rebased his stuff on top of 4.0.x
[21:51:28 CEST] <cone-283> ffmpeg 03Michael Niedermayer 07master:acba153a1487: avcodec/snowdec: Fix integer overflow with motion vector residual
[21:51:29 CEST] <cone-283> ffmpeg 03Michael Niedermayer 07master:1cbac9ce20d3: avcodec/vb: Check for end of bytestream before reading blocktype
[21:51:30 CEST] <cone-283> ffmpeg 03Michael Niedermayer 07master:c9222b972d6c: avcodec/hq_hqa: Check remaining input bits in hqa_decode_mb()
[21:51:31 CEST] <cone-283> ffmpeg 03Michael Niedermayer 07master:323095a6db67: avformat/mpegts: add missing null pointer checks in ff_parse_mpeg2_descriptor()
[22:10:02 CEST] <cone-283> ffmpeg 03Kieran Kunhya 07master:844ff494696e: mpeg4video: Add Studio DPCM support
[22:23:42 CEST] <durandal_1707> JEEB: who?
[22:26:43 CEST] <JEEB> the guy who added DVB protocol and some other broadcast stuff
[22:27:13 CEST] <JEEB> too bad his ISDB caption stuff requires the linuxtv custom iconv
[22:27:31 CEST] <JEEB> the DVB and ISDB/ARIB caption text encodings didn't get merged upstream because they were not "major enough"
[22:27:47 CEST] <JEEB> https://github.com/0p1pp1/FFmpeg/commits/isdb-4.0
[22:28:01 CEST] <JEEB> and then libaribb24 seemed to have issues with the license?
[22:28:27 CEST] <JEEB> I wonder if that was ever fixed or is the best alternative to make libbroadcasttext or something to decode/encode DVB and ARIB text codings :P
[22:28:38 CEST] <JEEB> since those iconv patches were clearly under LGPL or so
[22:29:16 CEST] <durandal_1707> please review my mt prores decoder patch asap
[22:30:35 CEST] <durandal_1707> and i still have no idea how to name blog
[22:30:57 CEST] <JEEB> I don't see such a patch and I'm not sure if I'd even grasp it depending on the complexity
[22:30:57 CEST] <atomnuker> JEEB: just implement in in lavu, I don't see an issue
[22:31:27 CEST] <JEEB> atomnuker: yea that's another thing I thought of, but I wondered if people would go "ewwww broadcast text coding stuff in lavu"
[22:31:29 CEST] <atomnuker> durandal_1707: name it - "The trials and tribulations of a XXX"
[22:32:15 CEST] <atomnuker> JEEB: would it be possible to implement it in the demuxer instead? if its not going to be used elsewhere might as well
[22:32:32 CEST] <durandal_1707> https://patchwork.ffmpeg.org/patch/10056/
[22:32:54 CEST] <JEEB> atomnuker: sure, the code could be pretty much everywhere. but the user would be both the demuxer as well as the subtitle decoder
[22:33:10 CEST] <atomnuker> ah, right
[22:33:11 CEST] <durandal_1707> atomnuker: trials? WTF
[22:33:13 CEST] <JEEB> since ARIB captions as well as the broadcast metadata effectively utilize the ARIB caption text coding
[22:33:23 CEST] <JEEB> they have a flag for UTF-8 but I've not seen that utilized in real life
[22:33:26 CEST] <JEEB> seems like a theoretical flag
[22:37:54 CEST] <atomnuker> JEEB: if you implement it as a .h file in lavc you can use it in both lavc and lavf
[22:38:09 CEST] <atomnuker> its how some opus things are needed in mpegts.c
[22:39:43 CEST] <JEEB> I will poke the aribb24 folk first because I'd really rather not start reimplementing it based on iconv stuff - although that way you'd also gain DVB text coding since the same guy implemented both the ARIB thing and the DVB one
[22:40:47 CEST] <durandal_1707> >>>>  https://patchwork.ffmpeg.org/patch/10056/  <<<<
[22:43:30 CEST] <durandal_1707> blog name: Bink it!
[23:02:08 CEST] <durandal_1707> bye
[00:00:00 CEST] --- Sun Aug 26 2018

More information about the Ffmpeg-devel-irc mailing list