[Ffmpeg-devel-irc] ffmpeg-devel.log.20160619

burek burek021 at gmail.com
Mon Jun 20 02:05:04 CEST 2016

[00:02:38 CEST] <Illya> ubitux how can I set the default layout to stereo? in the AVOptions I mean
[00:02:39 CEST] <Zeranoe> Compn: What do you mean "win10 is not even an os yet"?
[00:51:51 CEST] <michaelni> ubitux, the patch: http://pastebin.com/jDdAbkUM
[00:54:12 CEST] <Compn> Zeranoe : i mean it sucks shit!
[00:54:18 CEST] <Compn> slow, not many drivers out yet
[00:55:25 CEST] <Zeranoe> Compn: I've used it for a while and I'd disagree. What metric do you find it slow in? Many benchmarks would disagree with it being slow
[00:56:12 CEST] <nevcairiel> dont listen to Compn, he used windows 2000 until last year
[01:06:08 CEST] <Compn> its so slow compared to win2k
[01:06:14 CEST] <Compn> like watching paint dry
[01:06:23 CEST] <Compn> the ui is painful
[01:06:33 CEST] <Compn> (but also vista and xp)
[01:06:54 CEST] <Compn> i mean vista and xp are faster than w10
[01:06:58 CEST] <Illya> I like XP
[01:07:12 CEST] <Compn> Zeranoe : metric is playing some video game. 30 fps on win10 :\
[01:09:00 CEST] <Compn> also cant disable metro ui stuff.
[01:09:06 CEST] <Compn> but yeah dont listen to me, what do i know
[01:09:44 CEST] <Compn> i was able to use an operating system (win2k) for 15 years. :P
[01:10:19 CEST] <Compn> i guess winxp is nearing 15 year mark too :P
[01:11:25 CEST] <Zeranoe> Compn: What video game are you playing that worked on win2k and on win10? I'm not surprised it's slower on 10....
[01:11:30 CEST] <Compn> Zeranoe : how many ffmpeg binary downloads per month? just curious. 
[01:11:49 CEST] <Compn> i was comparing rocket league on vista vs win10. 
[01:12:00 CEST] <Zeranoe> too busy to pull the metric now
[01:12:05 CEST] <Compn> no prob
[01:12:17 CEST] <Compn> i need to stop arguing on the net
[01:12:19 CEST] <Compn> bbl
[01:16:04 CEST] <michaelni> ubitux, better patch: http://pastebin.com/FLE0hSjL
[01:31:26 CEST] <Illya> ubitux: I took a different approach to the probing, using file-based stream operations as opposed to trying to make my own above some data, and the probing works properly now. 
[01:44:41 CEST] <Illya> ubitux: I forgot to put a v2 on the new patch, sorry
[02:04:30 CEST] <cone-360> ffmpeg 03Michael Niedermayer 07master:0fcc252829a5: Revert "avfilter/trim: add compatibility layer to not break ABI used by ffmpeg"
[03:05:59 CEST] <cone-360> ffmpeg 03Andrey Turkin 07master:63adb3602d3b: libavcodec: factor out SEI generation for A53 captions
[06:21:01 CEST] <Zeranoe> Commit 63adb3602d3b35c5d1df14cf1e477bc458f96b7b doesn't compile with the changes made to qsvenc_h264.c. 'qsvenc_h264.c:46:9: error: q undeclared (first use in this function)'
[06:41:25 CEST] <Zeranoe> Not really sure why the declaration of q was moved to the inside of a conditional statement that's checking for a member of q...
[09:46:01 CEST] <cone-378> ffmpeg 03Paul B Mahol 07master:8a135a55b357: avcodec/magicyuv: check dimensions
[11:37:28 CEST] <ubitux> durandal_170:  ff_set_dimensions(avctx, bytestream2_get_le32(&gb), bytestream2_get_le32(&gb))
[11:37:43 CEST] <ubitux> the calling order of bytestream2_get_le32() is not defined
[11:38:02 CEST] <ubitux> michaelni: nice, i'm integrating this into the tree
[11:38:12 CEST] <ubitux> thank you
[11:38:26 CEST] <ubitux> michaelni: any other issue or i can push?
[11:38:53 CEST] <ubitux> Illya: i'll have a look
[11:38:56 CEST] <michaelni> ubitux, that was the one and only issue i saw
[11:39:04 CEST] <durandal_170> ubitux: for what compilers?
[11:39:15 CEST] <ubitux> durandal_170: C standard
[11:40:35 CEST] <ubitux> durandal_170: real case issue: ab80d7e8c52d912b94aeca52f77c03492c71ea94
[12:22:25 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:98c97994c5b9: h264: decouple extradata parsing from the decoder
[12:22:26 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:0bf5fd2e19b0: Merge commit '98c97994c5b90bdae02accb155eeceeb5224b8ef'
[12:41:34 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:48ea5433c813: lavc/h264_parser: pass logctx to h264_find_frame_end()
[12:41:35 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:34ec084b8481: lavc/h264: move history parsing variable to H264ParseContext
[12:51:28 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:72da8d9bb24d: h264_parser: remove the remaining dependencies on the h264 decoder
[12:51:29 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:0ed14bba1232: Merge commit '72da8d9bb24d1b1bf74c2f1108650c0da0054d2e'
[12:59:21 CEST] <ubitux> is UNCHECKED_BITSTREAM_READER in lavc/h264_parser.c still needed?
[12:59:44 CEST] <nevcairiel> if it uses get_bits.h then probably
[12:59:50 CEST] <nevcairiel> not that it ever really was needed
[12:59:58 CEST] <nevcairiel> its jut an optimization
[13:01:12 CEST] <ubitux> ah, that's for get_bits, ok
[13:01:21 CEST] <ubitux> yeah, makes sense
[13:01:40 CEST] <ubitux> i wonder if we shouldn't add a parser class in that file
[13:01:50 CEST] <ubitux> (for logging)
[13:03:10 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:0ba471d7d864: h264: eliminate copy_fields
[13:03:11 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:98456d4d69e0: Merge commit '0ba471d7d864c712f45d7ac6aca4829aba025adc'
[13:05:50 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:a2fd547839d3: h264_refs: reorder functions to avoid forward declarations
[13:05:51 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:902862bb4be4: Merge commit 'a2fd547839d3ef8f342b6f2c096d0887675d3b1a'
[13:08:27 CEST] <ubitux> ah, seems merging the next one is not going to be possible
[13:09:05 CEST] <nevcairiel> why not?
[13:09:26 CEST] <nevcairiel> at least partially where it is possible?
[13:10:00 CEST] <ubitux> ff_h264_decode_ref_pic_list_reordering() modifies the H264Context
[13:10:11 CEST] <ubitux> it writes to the default_ref
[13:10:27 CEST] <nevcairiel> then merge the other hunks only, skip that one?
[13:10:27 CEST] <ubitux> introduced in 4da2ac5c7a491b20be62ad19d77526e62aa57c69
[13:10:44 CEST] <ubitux> yeah, but can we do better?
[13:11:01 CEST] <ubitux> is this the appropriate place for default_ref?
[13:13:21 CEST] <nevcairiel> the only thing i can think of is having default_ref in the slice context, the places where its written seems otherwise ok
[13:14:41 CEST] <Illya> ubitux: yes it is propability....
[13:14:47 CEST] <ubitux> Illya: lol
[13:15:13 CEST] <Illya> also, good suggestion with the extensions
[13:15:27 CEST] <ubitux> Illya: ah btw, i forgot, what's the minimum version we support?
[13:15:38 CEST] <ubitux> you might want to add a version check
[13:15:51 CEST] <ubitux> require_pkg_config "libopenmpt > 1.2.3" for instance
[13:16:12 CEST] <ubitux> (">=" generally)
[13:16:17 CEST] <Illya> I didnt know of llrint either
[13:19:16 CEST] <Illya> what should I return if fopen() fails? an AVError?
[13:20:59 CEST] <Illya> ubitux: also, I think it could probably be null if the tracked data is passed in from another filter/device
[13:22:36 CEST] <ubitux> return AVERROR(errno)
[13:23:30 CEST] <ubitux> Illya: so anyway, does it play http://www.keygenmusic.net/?page=pack/packlist&lang=en these?
[13:23:41 CEST] <ubitux> :)
[13:30:36 CEST] <ubitux> michaelni: any comment on having default_ref in the slice context instead of the h264 one?
[13:30:55 CEST] <Illya> 0.2 looks like the first version supported, no functions have been added since afaict
[13:31:37 CEST] <ubitux> okay
[13:31:58 CEST] <Illya> it was also the first version released
[13:32:24 CEST] <Illya> So I guess that doesnt require a version check, as all versions should be compatible
[13:32:35 CEST] <ubitux> alright
[13:35:36 CEST] <Illya> it plays all the files I've tested so far
[13:36:10 CEST] <Illya> from your keygenmusic link
[13:37:57 CEST] <nevcairiel> ubitux: imho just merge as good as possible right now, and if we decide to move it later we can then add the const, future improvements shouldnt hold this up
[13:38:13 CEST] <ubitux> yeah that was what i was doing
[13:38:32 CEST] <nevcairiel> :)
[13:41:23 CEST] <Illya> ubitux: here's the next iteration: http://sprunge.us/EgZB Would you prefer if I put it on the ML as well?
[13:42:40 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:755f79f84cbe: h264_refs: make the H264Context const where possible
[13:42:41 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:222e7d23d0fb: Merge commit '755f79f84cbeb5d749fb120e55e0098a2d7663a0'
[13:42:58 CEST] <ubitux> Illya: so, last question, it can't be stream oriented?
[13:43:08 CEST] <ubitux> 'cause you're mapping the whole file in memory here
[13:43:48 CEST] <Illya> Well, it can, but it requires callbacks for read/seek/tell, and I wasn't sure how to implement that at all.
[13:44:30 CEST] <ubitux> ok
[13:44:31 CEST] <Illya> For the same reason, I'm using the file callbacks in probe, when it can handle streams.
[13:44:45 CEST] <ubitux> i guess it's fine for a first version
[13:45:01 CEST] <ubitux> at some point you probably want to look into making a custom aviocontext
[13:45:03 CEST] <ubitux> i guess
[13:45:36 CEST] <Illya> sounds good :)
[13:46:34 CEST] <ubitux> better send it to the ml anyway
[13:46:45 CEST] <ubitux> poke me at the end of the day if no one applied it
[13:47:32 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:89ae244e78a8: h264_refs: remove an unused parameter from ff_h264_fill_mbaff_ref_list()
[13:47:33 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:d5cfe233240d: Merge commit '89ae244e78a8d636fbb8427a2da052b04cf710bd'
[13:49:08 CEST] <cone-378> ffmpeg 03Anton Khirnov 07master:5b35b290dd76: h264: remove a stale comment
[13:49:09 CEST] <cone-378> ffmpeg 03Clément BSsch 07master:c3afe64d0cbe: Merge commit '5b35b290dd76db64f079b3940f6148c8f273fc17'
[13:50:26 CEST] <ubitux> should i merge the removal of low_delay?
[13:51:06 CEST] <ubitux> 99c554efc8b09c3f1bb2fb41c3da5431085f7470
[13:51:53 CEST] <nevcairiel> if its equally unused in our code as in the patch, it probably servers no purpose
[13:52:29 CEST] <kierank> durandal_170: did you see kostya documented apple pixlet
[13:52:58 CEST] <ubitux> nevcairiel: it seems we have much more code around it
[13:54:47 CEST] <michaelni> ubitux, not really an oppinion on default_ref location ATM
[14:00:50 CEST] <ubitux> ok
[14:05:04 CEST] <ubitux> michaelni: opinion on low_delay?
[14:19:47 CEST] <durandal_170> kierank: yes, some details are missing
[14:21:38 CEST] <michaelni> ubitux, if AV_CODEC_FLAG_LOW_DELAY has no users and it doesnt break anything then its better to remove as simpler
[14:21:56 CEST] <michaelni> i couldnt find a user of AV_CODEC_FLAG_LOW_DELAY quikly but that doesnt say much
[14:23:21 CEST] <michaelni> if its left we should probably add a fate test for AV_CODEC_FLAG_LOW_DELAY
[17:06:12 CEST] <cone-281> ffmpeg 03James Almer 07master:5f5a97dc3d29: avcodec/qsvenc_h264: declare the QSVEncContext variable before using it
[18:52:48 CEST] <cone-281> ffmpeg 03Carl Eugen Hoyos 07master:e274113292ac: lavc/cafenc: Allow QDMC muxing.
[19:01:21 CEST] <cone-281> ffmpeg 03Clément BSsch 07master:fd1d84bcf6f4: lavc/magicyuv: fix undefined behaviour introduced in 8a135a55b
[21:40:10 CEST] <Illya> ubitux: you told me to ping you at the end of the day if no one had looked at it?
[22:04:12 CEST] <cone-414> ffmpeg 03Thomas Mundt 07master:8b7b52c863f0: avfilter/vf_bwdif: Change default to deinterlace all frames
[22:31:15 CEST] <ubitux> Illya: sure, give me a moment
[22:42:12 CEST] <ubitux> Illya: i was ready to apply but it seems to break fate
[22:42:18 CEST] <ubitux> (because it probes too much files)
[22:45:06 CEST] <ubitux> i'm going to do a simple return ret > 0.5 ? AVPROBE_SCORE_EXTENSION : 0;
[22:45:13 CEST] <ubitux> actually that's not enough
[22:46:26 CEST] <ubitux> maybe ret * AVPROBE_SCORE_EXTENSION might do
[22:47:27 CEST] <ubitux> yeah well i don't want to rush it
[23:22:35 CEST] <cone-414> ffmpeg 03Marton Balint 07master:0a065694877e: avformat/udp: do not accumulate packet_gap delay errors
[23:22:36 CEST] <cone-414> ffmpeg 03Marton Balint 07master:cc16229d914a: avformat/udp: replace packet_gap with bitrate option
[00:00:00 CEST] --- Mon Jun 20 2016

More information about the Ffmpeg-devel-irc mailing list