[Ffmpeg-devel-irc] ffmpeg-devel.log.20150211

burek burek021 at gmail.com
Thu Feb 12 02:05:02 CET 2015

[00:01] <Timothy_Gu> also are the trig functions in radians or degrees?
[00:22] <kierank> radians
[00:23] <kierank> oh aevalsrc erm might not be
[00:23] <kierank> but I assume it uses libc sinc
[00:23] <Daemon404> ... why would you ever make it degrees
[00:42] <llogan> michaelni: also i forgot to mention the typo in the commit title: s/frame_drop_threashold/frame_drop_threshold </nitmongering>
[00:44] <michaelni> llogan, do you want to push it with the typo fixes? if so dont hesitate
[00:44] <michaelni> otherwise ill do it later
[01:15] <cone-716> ffmpeg.git 03Michael Niedermayer 07master:26db9d3b77fb: doc/ffmpeg: Document frame_drop_threshold
[01:16] <michaelni> llogan, thx
[01:41] <cone-716> ffmpeg.git 03Anton Khirnov 07master:0232ba62a285: frame: clarify buf documentation
[01:41] <cone-716> ffmpeg.git 03Michael Niedermayer 07master:1cca12488617: Merge commit '0232ba62a28527579e34d44700b48017cc727e5f'
[01:50] <cone-716> ffmpeg.git 03wm4 07master:6a5b8ca43290: lavf: remove unused code
[01:50] <cone-716> ffmpeg.git 03Michael Niedermayer 07master:702cf8d0de9e: Merge commit '6a5b8ca4329039fad44ad50b6496948f4bfacb4c'
[02:05] <cone-716> ffmpeg.git 03Anton Khirnov 07master:1509c018bd5b: mpegts: relax restrictions on matching the packet start in read_header
[02:05] <cone-716> ffmpeg.git 03Michael Niedermayer 07master:b61170f51d2e: Merge commit '1509c018bd5b054a2354e20021ccbac9c934d213'
[02:50] <cone-716> ffmpeg.git 03wm4 07master:9deaec782810: lavf: move internal fields from public to internal context
[02:50] <cone-716> ffmpeg.git 03Michael Niedermayer 07master:2924514721bc: Merge commit '9deaec782810d098bca11c9332fab2d2f4c5fb78'
[11:38] <cone-823> ffmpeg.git 03Michael Niedermayer 07master:08509c8f8662: avcodec/mjpegdec: Skip blocks which are outside the visible area
[11:38] <cone-823> ffmpeg.git 03Michael Niedermayer 07master:238493398558: ffmpeg: Print negative times like "-00:05:01.22" instead of "00:-5:-1.-22"
[12:46] <cone-823> ffmpeg.git 03Paul B Mahol 07master:3a8801eb8d72: Changelog: mention more ported filters
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:20be3ea442ab: avcodec/mjpegdec: Check escape sequence validity
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:fd6a9fcd427b: avcodec/mjpegdec: Check number of components for JPEG-LS
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:42d9a7010f2d: swscale/utils: Limit filter shifting so as not to read from prior the array
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:3508aa142766: avformat/thp: Check av_get_packet() for failure not only for partial output
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:4ef5605fc91d: avcodec/h264_ps: More completely check the bit depths
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:763c7533483b: avcodec/h264: Be more strict on rejecting pps/sps changes
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:d9618b964b17: avutil/opt: Fix types used to access AV_OPT_TYPE_PIXEL_FMT
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:5b4a79ee0217: avutil/opt: Fix type used to access AV_OPT_TYPE_SAMPLE_FMT
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:64e50b2f2a68: avcodec/h264_slice: Do not change frame_num after the first slice
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:4de40be50773: avcodec/h264_slice: Check picture structure before setting the related fields
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:ac8ef33f9a39: avcodec/h264_slice: ignore SAR changes in slices after the first
[13:24] <cone-823> ffmpeg.git 03Michael Niedermayer 07release/1.2:7e7772c13ad8: avcodec/mjpegdec: Skip blocks which are outside the visible area
[13:37] <cone-823> ffmpeg.git 03Paul B Mahol 07master:ac494e5a6650: avformat/isom: add X-Com Radvision fourcc
[13:52] <cone-823> ffmpeg.git 03Alexey Titov 07master:a05a7373166c: avfilter/unsharp: OpenCL unsharpen filter optimization: substitute N^2 filter computation with 2N+C
[17:21] <cone-823> ffmpeg.git 03Paul B Mahol 07master:edf217ebb7d5: avfilter: add dcshift filter
[17:36] <cone-823> ffmpeg.git 03Jean Delvare 07master:194165aed862: cmdutils: fix success path
[18:18] <akira4> ubitux, how do I decide the value of the flag CODEC_FLAG2_RAWTEXTSUB?
[18:20] <ubitux> 0x40000000 should work
[18:20] <ubitux> just after SKIP_MANUAL
[18:20] <ubitux> hex(0x20000000<<1)
[18:20] <ubitux> '0x40000000'
[18:26] <akira4> okay. Could you have a look at the diff? http://pastebin.com/AHD05xYC
[18:27] <ubitux> try to call it with the decoded strings
[18:27] <akira4> hmm. as in inside the loop?
[18:28] <ubitux> yeah, typically
[18:29] <akira4> do I need to add a field in AVCodecContext too?
[18:30] <ubitux> why do you think so?
[18:30] <akira4> it was written in the task description
[18:30] <ubitux> the option already exists, you're just adding a flag
[18:30] <ubitux> let me check
[18:31] <ubitux> akira4: that's a mistake from me :)
[18:31] <ubitux> fixed
[18:31] <akira4> cool.
[18:32] <ubitux> btw, wm4 wrote an ASS stripper in mpv, which he agreed to relicense to LGPL
[18:32] <ubitux> you can probably take inspiration from it
[18:32] <Daemon404> ASS striper :o
[18:32] <Daemon404> stripper
[18:32] <ubitux> :)
[18:32] Action: Daemon404 runs
[18:32] <wm4> the bad puns never end
[18:32] <ubitux> perfectly fit for opw, thanks guys :p
[18:33] <akira4> haha. sure.
[18:33] <ubitux> akira4: here it is: http://git.srsfckn.biz/mpv/tree/sub/sd_ass.c#n181
[18:33] <ubitux> you'll probably use bprint api instead
[18:33] <ubitux> but that's the same thing
[18:33] <ubitux> that is, assuming wm4 is still OK to relicense to LGPL
[18:33] <wm4> yep
[18:34] <Daemon404> if he didnt, you could steal aegisub's
[18:34] <wm4> I was about the paste the useful parts: http://privatepaste.com/6cf91914ca
[18:34] <Daemon404> which is MIT
[18:34] <wm4> Daemon404: post it anyway, it might be a good reference too
[18:34] <Daemon404> effort to fidn it right now
[18:34] <Daemon404> almost lunch
[18:35] <ubitux> wm4: anyway, just fyi, as you might guess, -flags2 +rawtextsub will fill the AVSubtitle->rects[0..N]->text
[18:35] <akira4> :D thanks 
[18:35] <ubitux> in addition to ->ass
[18:35] <ubitux> so akira4 is doing that right now
[18:35] <wm4> ubitux: not sure what you mean, but sd_ass.c doesn't involve lavc at all
[18:36] <ubitux> the idea is that you can configure ffmpeg to decode the text sub and get them without markup
[18:36] <ubitux> if you have a limited renderer
[18:36] <ubitux> like stdout, or an old OSD, or whatever
[18:36] <wm4> uh that sounds very awkward
[18:36] <ubitux> why?
[18:36] <wm4> is it additional to the ass?
[18:36] <ubitux> yeah because we're not patching all decoders
[18:37] <ubitux> we just decide to strip the markup after the decoders
[18:37] <ubitux> (if requested)
[18:37] <wm4> well, as long as I can always just use the ass (and strip the tags if needed)...
[18:37] <ubitux> sure
[18:37] <ubitux> -flags2 +rawtextsub will just fill the ->text field
[18:37] <ubitux> that's all
[18:38] <wm4> why doesn't it always do this?
[18:38] <ubitux> zomg speed
[18:39] <ubitux> could be set by default but well
[18:39] <ubitux> you probably don't want the raw version in many cases
[18:39] <ubitux> so no need to call this for every single subtitles for nothing
[18:40] <wm4> wouldn't it be better to make the ZOMG SPEED version the non-default (i.e. swap the semantics of the flag)
[18:40] <nevcairiel> doesnt seem like something that the majority of users would w ant
[18:40] <wm4> at least you should stick a big fat warning on the text field's doxygen
[18:40] <wm4> well, it's all about expectations
[18:40] <ubitux> yeah we should probably hint the option above
[18:41] <wm4> "you need to set this obscure flag in some completely different place to get lavc to set this field"
[18:41] <ubitux> :)
[19:02] <llogan> haha. word censoring fun. http://ffmpeg.gusari.org/viewtopic.php?f=11&t=625&p=5512#p5512
[19:03] <wm4> is it spam?
[19:03] <llogan> yes. let me look up the replaced word...
[19:04] <llogan> iDealshare VideoGo
[19:05] <llogan> i'll delete the post, but i wonder what the humonbots think (if they can read).
[19:05] <wm4> nice name (not)
[19:05] <nevcairiel> my father installed some random video converter some months ago and his PC got  infested with evilness in the process
[19:05] <nevcairiel> why someone would even use those..
[19:05] <llogan> they probably come up high in search and are "free"
[19:06] <llogan> nevcairiel: a cow-orker did the same a few years ago. don't know why they didn't ask me
[19:06] <nevcairiel> i gave him handbrake instead, all he wanted was convert a few videos so they play on his TV from a usb drive (which is rather picky about supported formats)
[19:07] <jamrial> nevcairiel: person gets a weird "mkv" file handed to them. notices windows media player or quicktime doesn't play them, so they google "mkv to avi" or "mkv to quicktime"
[19:07] <jamrial> happens all the time
[19:08] <llogan> they seem to release dozens of "programs" each titled for each task (but probably the same shitty shit)
[19:08] <jamrial> yeah
[19:08] <nevcairiel> and in most cases just a wrapper for cli ffmpeg or some such
[19:09] <llogan> yes!
[19:09] <llogan> usually with --non-free
[19:09] <nevcairiel> because writing a tiny dumb GUI with ads in it is a 1 hour job :d
[19:09] <llogan> enable-nonfree
[19:10] <llogan> here is the naming scheme: i*soft
[19:50] <Plorkyeran> <@ubitux> akira4: here it is: http://git.srsfckn.biz/mpv/tree/sub/sd_ass.c#n181 <-- the check for in drawing is wrong there, btw
[19:50] <Plorkyeran> \p with no argument leaves drawing mode
[19:50] <Plorkyeran> and leading zeroes are supported
[19:50] <ubitux> a mission for wm4
[19:51] <Plorkyeran> also I think that incorrectly drops text after { with no trailing }
[19:51] <Plorkyeran> (ass is dumb)
[19:54] <wm4> sucks
[19:56] <Plorkyeran> https://github.com/Aegisub/Aegisub/blob/master/src/ass_dialogue.cpp#L199 is aegisub's shitty block parsing for comparison
[19:58] <wm4> I guess I'll fix my code tomorrow or so
[20:03] <rcombs> ass is, indeed, dumb
[20:31] <Daemon404> s/ass/vsfilter/
[20:45] <rcombs> no, ass is dumb too
[20:45] <rcombs> a lot of it is vsfilter's fault, and vsfilter is very dumb, yeah
[20:53] <wm4> keep in mind that the original application generating ass was written in visual basic
[21:20] <Daemon404> thats not ass
[21:20] <Daemon404> thats ssa
[21:43] <aetasx> you guys at all familiar with -fprofile-generate?  unfortunately the gcc channel is where questions goto die
[22:51] <compn> no longer allowed in mkv ;D
[22:51] <compn> ehe
[22:54] <cone-823> ffmpeg.git 03Michael Niedermayer 07master:7366bb38f10e: avcodec/mpegvideo_enc: Add intra_chroma_ac_vlc_length, it will be needed for mjpeg
[22:54] <cone-823> ffmpeg.git 03Michael Niedermayer 07master:7cac568b3970: avcodec/mjpegenc: support trellis quantization
[22:54] <cone-823> ffmpeg.git 03Michael Niedermayer 07master:7d24ccaf002b: fate: mjpeg trellis test
[22:54] <Daemon404> so mjpegenc supports trellis, but not, you know, more than 1 quant matrix.
[22:54] <Daemon404> lol
[23:32] <kierank> Daemon404: that's why it's so slow
[23:32] <kierank> iirc it does subme 9/10 by default
[23:33] <Daemon404> and yet it will still suck because it uses the same qmat for chroma and luma
[00:00] --- Thu Feb 12 2015

More information about the Ffmpeg-devel-irc mailing list