[Ffmpeg-devel-irc] ffmpeg-devel.log.20141015

burek burek021 at gmail.com
Thu Oct 16 02:05:02 CEST 2014


[03:54] <jamrial> anyone else having trouble accessing videolan (and therefore the git repo)?
[03:56] <rcombs> looks down from here
[04:39] <michaelni> jamrial, rcombs yes videolan is down i dont know why, use github ffmpeg mirror for checkout/fetch/pull and keep your changes locally or push to your own public repo
[06:09] <cone-787> ffmpeg.git 03Michael Niedermayer 07master:e3b28f10bb9b: avformat/mxfdec: Fix termination of mxf_data_essence_container_uls
[06:09] <cone-787> ffmpeg.git 03Michael Niedermayer 07master:aae4f5108d04: avformat/mxfdec: Fix memleak of primer_packs
[06:09] <cone-787> ffmpeg.git 03Michael Niedermayer 07master:0834f2056c7f: avformat/mxfdec: Fix memleak of packages_refs
[06:09] <cone-787> ffmpeg.git 03Michael Niedermayer 07master:676110e6d4c4: avformat/mxfdec: cleanup if parsing the header metadata fails
[10:07] <ubitux> lol
[10:07] <ubitux> Cc: Chinese, god, of, Technology, Co., Ltd.
[10:07] <ubitux> riiiight.
[10:42] <ubitux> nicolas hasn't the permissions to close tickets?
[10:42] <ubitux> oh, misread
[11:50] <rcombs> https://gist.github.com/203026397f53729c9595 <-- on a scale of 1~10, how poorly did I explain/name this new option?
[11:57] <wm4> hmm
[11:57] <wm4> I'd probably call it "copyts_base_zero"
[11:57] <wm4> /bikeshedding
[11:57] <rcombs> how about this: https://gist.github.com/8ddcb0fd5f40f8a11029
[11:58] <rcombs> which instead gives us a tri-state of `-copyts`/`-nocopyts`/(default)
[11:58] <wm4> hm
[11:58] <wm4> ask on the ML
[11:59] <rcombs> where copyts is the current -copyts behavior, the default is the current default, and -nocopyts makes the input file's timestamps start at 0 but doesn't have the other effects of copyist
[11:59] <rcombs> *copyts
[11:59] <rcombs> (I think)
[12:03] <rcombs> I've phrased this really badly, actually
[12:03] <rcombs> I'm clearly too tired to be doing anything useful
[12:06] <rcombs> https://gist.github.com/8988f495215740aaf5ee does this description seem saner?
[12:06] <rcombs> (if it does, I'm chucking that off to the ML)
[12:08] <wm4> rcombs: how does the user set the option here? numeric?
[12:08] <wm4> rcombs: also, usually in git, you add a newline to separate subject line and body
[12:08] <wm4> even ffmpeg follows that convention
[12:09] <rcombs> ah yeah, I keep forgetting to do that
[12:10] <rcombs> wm4: and no, it's [no arg] for the default behavior (i.e. copy_ts = -1), `-copyts` for the current `-copyts` behavior (i.e. copy_ts = 1), and `-nocopyts` for the new behavior (copyts = 0)
[12:10] <rcombs> a tri-state similar to [no arg] vs -stats vs -nostats
[12:15] <cone-288> ffmpeg.git 03Rémi Denis-Courmont 07master:e09a947c1a85: doc: fix typing mistake
[12:15] <cone-288> ffmpeg.git 03Michael Niedermayer 07master:a72e75bf980c: Merge commit 'e09a947c1a856d9d34199a13ee378fe46be53d3a'
[12:26] <wm4> rcombs: that seems strange (no arg?)
[12:27] <wm4> rcombs: in any case, just post your patch; it doesn't have to be perfect, and others will make improvement suggestions
[12:27] <rcombs> SGTM
[12:29] <cone-288> ffmpeg.git 03Rémi Denis-Courmont 07master:c1724623ce04: vdpau: have av_vdpau_bind_context() fail on unsupported flag
[12:29] <cone-288> ffmpeg.git 03Michael Niedermayer 07master:fd937866f7b7: Merge commit 'c1724623ce0433c6a9ee72133b1fd4db75ec7193'
[13:18] <cone-288> ffmpeg.git 03Rémi Denis-Courmont 07master:d565fef1b83b: vdpau: add AV_HWACCEL_FLAG_IGNORE_LEVEL to skip the codec level check
[13:18] <cone-288> ffmpeg.git 03Michael Niedermayer 07master:51c810e62b63: Merge commit 'd565fef1b83b6c5f8afb32229260b79f67c68109'
[16:21] <cone-712> ffmpeg.git 03Anton Khirnov 07master:7ea1b3472a61: lavc: deprecate the use of AVCodecContext.time_base for decoding
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:17085a0251a1: Merge commit '7ea1b3472a61de4aa4d41b571e99418e4997ad41'
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:801876fb079e: avcodec/utils: fix time_base after the framerate addition
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:620e7f0f3b3d: avcodec/mpeg12dec: fix time_base and framerate
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:09450c550925: avcodec/h264: fix time_base and framerate
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:220a15c07487: avcodec/vc1: fix time_base and framerate
[16:21] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:18255441dd74: avcodec/dpx: fix framerate
[17:13] <ubitux> can anyone activates the spam filter on opw at ff?
[18:23] <sophiejjj> hi. I am interested in participating as opw interns in ffmpeg.
[18:24] <J_Darnley> Do you have a specific task in mind?
[18:24] <sophiejjj> J_Darnley: YUP!
[18:24] <sophiejjj> I find the basic server and the "browsing content" fits me most.
[18:28] <J_Darnley> For the "basic server" I suggest you contact the mentor(s) directly and/or the opw at ffmpeg.org address...
[18:29] <sophiejjj> J_Darnley: do the project lists on the opw page have equal priority?
[18:29] <cone-712> ffmpeg.git 03Anton Khirnov 07master:2d6e58497e76: lavf: switch to AVCodecContext.framerate for demuxing
[18:29] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:2b37864e7f2b: Merge commit '2d6e58497e76836604364b037df9b00ba3d75b69'
[18:29] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:a92192d2e4f7: avformat/utils: use av_reduce() to simplify code and avoid overflow in ff_compute_frame_duration()
[18:29] <J_Darnley> for the "browsing content" task I suggest you you send an email to both the opw and ffmpeg-devel lists asking for some guidance/mentors.
[18:30] <J_Darnley> I don't think one project vs. another will affect your chances of getting a spot.
[18:31] <sophiejjj> J_Darnley: I am interested in both, and is now thinking on how to make my application nice looking. Do you have any advice on what I should I care about in the application?
[18:32] <J_Darnley> As far as I know it depends mostly on how well you do the qualification task and to a lesser extent, how good your "work ethic" is.
[18:32] <sophiejjj> :). work ethic.
[18:33] <J_Darnley> I think the OPW trac page says some thing along the lines of "we don't care how fancy your application is, we are more interested in the code"
[18:33] <wm4> there are many applicants, and only 1 (or maybe 2-3) slots
[18:33] <wm4> so the idea is the select the best programmer out of the applicants
[18:33] <J_Darnley> Also that ^
[18:34] <wm4> and that's also what the "we are more interested in the code" refers to
[18:34] Action: sophiejjj hopes it's not to late to start.
[18:35] <sophiejjj> /s/to/too
[18:35] <wm4> there's a deadline, probably on the wiki
[18:35] <sophiejjj> wm4: yup. 25/OCT
[18:35] <wm4> if you get done with the qualification task until then, then I suppose it's not too late
[18:35] <wm4> still 10 days
[18:36] <sophiejjj> wm4: my fault. 10.22.    "10.25" is my network course assignment ddl. my bad.
[18:37] <wm4> so the 22nd? still a week
[18:37] <J_Darnley> A week...  I hope you have some experience with network programming.
[18:38] <wm4> aren't qualification tasks supposed to be simple? (I didn't look)
[18:38] <sophiejjj> wm4: en. I see already some applicants are there. https://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12-Qualis
[18:38] <J_Darnley> the server task asks for a "proof-of-concept" which I would assume means basic listening and responding to commands
[18:40] <J_Darnley> As I said before, I strongly suggest you contact mentors directly.
[18:40] <sophiejjj> J_Darnley: yup. I am going to do that right now.
[18:41] <sophiejjj> J_Darnley: btw, by work ethis, you mean?
[18:41] <J_Darnley> How well you communicate with mentors and the project..
[18:42] <J_Darnley> How "happy" you seem to be working on the task...
[18:43] <J_Darnley> Your attitude...
[18:43] <J_Darnley> things like that.
[19:01] <cone-712> ffmpeg.git 03Anton Khirnov 07master:557d4c4ecae0: avconv: replace AVCodecContext.time_base with framerate
[19:01] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:137f7a630b5f: Merge commit '557d4c4ecae080729e68df54ed2bd8420d0f037a'
[19:08] <llogan> Compn: i got the ""Timetable and Fares Information". and some stickers. thanks!
[19:11] <cone-712> ffmpeg.git 03Martin Storsjö 07master:9b7f932ee568: rtpdec_hevc: Parse out of band vps/sps/pps/sei from fmtp lines
[19:11] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:6afc319eee79: Merge commit '9b7f932ee568cadfc0f556a061fcc00cb63f9780'
[19:14] <sophiejjj> btw.
[19:14] <sophiejjj> J_Darnley: how should I contact Cigaes. seems he's not online.
[19:14] <sophiejjj> Nicolas George.
[19:14] <sophiejjj> email addres would be the best, but it's not listed on the opw page.
[19:17] <cone-712> ffmpeg.git 03Martin Storsjö 07master:e5cfc8fdad90: sdp: Provide out of bound parameter sets for HEVC if extradata is set
[19:17] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:a9ed4f1d1567: Merge commit 'e5cfc8fdad901c9487fe896421972852f38bcf5'
[19:19] <cone-712> ffmpeg.git 03Lou Logan 07master:b629b421d742: doc/filters: update link to OpenCV docs
[19:31] <J_Darnley> sophiejjj: unless there is more than one nicholas george: george at nsup.org
[19:32] <Daemon404> i really wish people were not such sticks in the mud
[19:32] <Daemon404> mentoring via email is sad.
[19:33] <relaxed> irc should be a requirement for mentors
[19:34] <J_Darnley> Well, that might be his preference anyway.  He apparently oesn't list the address on the trac page
[19:37] <cone-712> ffmpeg.git 03Clément BSsch 07master:8022bb1d9271: fate: make subtitles tests output to stdout instead of md5
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:d97d4795fd26: fate: add rawdiff and use it for subtitles
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:b243393e8b4b: fate: fix some subtitles tests dependencies
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:bf700a995da4: avformat/assenc: support remuxing old V4 ASS form (SSA)
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:40b9f28641b6: avcodec/ass: output missing fields in AVSubtitles and output files
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:d9f272fe3306: avcodec/ass_split: extend recognized fields in ASS splitter
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:3e86ead383c1: avcodec/ass_split: assume default order when "Format:" line is not found
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:462f2876ce41: avcodec/ass_split: realign after addition of new fields
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:d281a87ae37b: avformat/assdec: typo fix
[19:38] <cone-712> ffmpeg.git 03Clément BSsch 07master:52b0a0ecaa02: avcodec/ass: add FFmpeg signature
[19:38] <relaxed> ubitux: :)
[19:39] <ubitux> ...and now i can work on the important part, hopefully
[19:43] <ubitux> relaxed: please check if it fixes properly your issue though :)
[19:43] <ubitux> like, if rendering is correct or stuff like that
[19:47] <cone-712> ffmpeg.git 03Lou Logan 07master:8a9537999029: doc/encoders: add basic libx265 documentation
[19:48] <akira4> ubitux, I added the decoder. Here is the diff http://pastebin.com/TGtWc1yz. Checked with some stl files and its working fine.
[19:51] <wm4> ubitux: wow!
[19:51] <wm4> so ffmpeg outputs correct ASS now?
[19:51] <ubitux> akira4: doesn't look bad :)
[19:51] <ubitux> akira4: i'll have a deeper look in a moment
[19:51] <ubitux> wm4: yep, at least it should
[19:51] <akira4> alright :)
[19:52] <wm4> ubitux: finally... good job
[19:52] <relaxed> llogan: is it frowned upon to have urls in the man page? If not, http://x265.readthedocs.org/en/default/cli.html#standalone-executable-options might be helpful.
[19:53] <J_Darnley> akira4: you have an small alignment issue in the makefile
[19:53] <ubitux> wm4: note that AVSubtitles->rects[0]->ass also is valid ass now
[19:53] <ubitux> wm4: i didn't mention it in the api changes, i probably should
[19:53] <ubitux> (i bumped though)
[19:53] <myra_> hello, I have written a patch based on Symmetric ciphers  . When I run a tools/patcheck on my patch it gives me an error that missing changes in ChangeLog . Can somebody tell me what does this mean ? and What is ChangeLog used for
[19:53] <akira4> I'll fix the alignment issue
[19:53] <J_Darnley> Also, you might want to put your own name in the copyright header.
[19:54] <wm4> ubitux: yeah, probably... it doesn't matter for me (I think), but other users might break...
[19:55] <llogan> relaxed: i don't believe there is an official stance about links. obviously it just sucks when they change/die (such as the one i fixed 36 minutes ago)
[19:55] <akira4> can I put my name on the copyright header?
[19:55] <wm4> ubitux: but it still outputs text-ASS, right? not Matroska-ASS
[19:56] <J_Darnley> You wrote it didn't you?
[19:56] <ubitux> wm4: yes
[19:56] <J_Darnley> I mean in the new file you added
[19:56] <ubitux> that's what i'm going to fix now
[19:56] <ubitux> "now"
[19:56] <akira4> I did.
[19:56] <ubitux> TIL akira4 is named Clément BSsch
[19:57] <ubitux> just like me!
[19:57] <llogan> is boesch a french word for "German"?
[19:57] <J_Darnley> Then perhaps you forgot to add the change before committing.
[19:57] <ubitux> llogan: bosch is a french word for german nazi
[19:58] <michaelni> myra_, the Changelog lists changes, like "avutil: cast128 support" or something like that
[19:58] <akira4> :D I'll change it to my "real" name then 
[19:58] <llogan> ubitux: oh, sorry... i think it predated nazi though.
[19:58] <ubitux> llogan: but bSsch/boesch is the french bastardization of a german name bösch, AFAIK
[19:58] <ubitux> llogan: maybe :P
[19:58] <wm4> Klement Bösch?
[19:59] <ubitux> maybe :p
[19:59] <wm4> man I bet you get many typos of your name and also UTF-8 fuckups
[19:59] <ubitux> yes
[19:59] <ubitux> it triggers all kind of shit with administrative stuff
[19:59] <ubitux> or shitty websites
[19:59] <llogan> ubitux: i tihnk i had some family ancestor named Putnam who was renamed to Buttman when they moved to US
[20:00] <myra_> @michaelni ... so if i have to submit my patch is it necessary that add the changes in the ChangeLog right away as there is a lot of work on the algorithm that is yet to be done
[20:00] <ubitux> llogan: bad luck M. Buttman
[20:00] <michaelni> myra_, your patch should add 1 line to the Changelog file
[20:01] <myra_> Will do !!
[20:02] <akira4> fixed the alignment and added my name http://pastebin.com/hif3hRSi 
[20:03] <ubitux> akira4: so far it looks mostly good, so how about you send it to the mailing list?
[20:03] <ubitux> we'll make a round of real review over the mailing list
[20:03] <ubitux> so you can see how we iterate
[20:04] <akira4> sounds good. I'll do that.
[20:04] <ubitux> you can use git send-email, or just git format-patch -1, and send the file
[20:04] <akira4> alright.
[20:05] <akira4> ubitux, btw I kept your name on the copyright boilerplate because you helped so much :)
[20:05] <ubitux> as you wish, you can take full authorship on it, i really don't mind
[20:06] <akira4> cool.
[20:11] <akira4> Is there a naming convention for patch files that are submitted?
[20:12] <jamrial> you should use git send-email, but if you can't then attach the output of git format-patch
[20:13] <wm4> akira4: the patch files? doesn't matter
[20:14] <wm4> the patch format and contents are important
[20:14] <akira4> that's true. thanks 
[20:14] <cone-712> ffmpeg.git 03Michael Lynch 07master:c69f230a3afb: avformat/rtsp: check memory allocation
[20:21] <Daemon404> i wonder whats with the sudden surge of patches from samsung
[20:21] <wm4> where
[20:22] <Daemon404> there have been a few lately
[20:22] <ubitux> aren't they sponsoring one slot for ffmpeg?
[20:22] <Daemon404> ubitux, maybe
[20:22] <ubitux> maybe they just discovered the project
[20:22] <Daemon404> ubitux, considerign they shipped lgpl violating binaries for years and continue to do so
[20:22] <Daemon404> says they didnt
[20:22] <Daemon404> ;)
[20:22] <ubitux> reynaldo might know more
[20:23] <jamrial> Some guy mention a day or so ago here that he approached Samsung about those violations and got them to admit it, or something like that
[20:24] <jamrial> guess it worked
[20:24] <cone-712> ffmpeg.git 03Yu Xiaolei 07master:8091fe3071db: build: mov_muxer requires ac3_parser
[20:24] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:1b54e6b12bcf: avformat/mov: auodetect "use_mfra_for"
[20:29] <jeetu> hello,i want  to Misc Libravfilter extension project
[20:29] <jeetu> i want to know is their anyone else who is working on this project
[20:30] <llogan> of course when i try to contact dolby through their web form it has a "server error"...
[20:32] <michaelni> jeetu, i think theres someone working on it but it seems she is not listed on https://trac.ffmpeg.org/wiki/SponsoringPrograms/OPW/2014-12-Qualis
[20:32] <michaelni> jeetu, but do not hesitate to add yourself to that wiki page if you start working on a qualifiation task
[20:32] <jeetu> shall i start work on this project
[20:33] <Daemon404> llogan, why are you contacting dolby
[20:33] <Daemon404> theyre the #1 ffmpeg and vlc haters
[20:33] <JEEB> isn't that dts?
[20:33] <Daemon404> dts is 2nd i think
[20:33] <llogan> Daemon404: i'm updating old, dead links in the docs and wanted the new URL to 46_DDEncodingGuidelines.pdf (but I didn't mention why i wanted it)
[20:34] <Daemon404> oh ok
[20:34] <Daemon404> theyll probably invite you to pay for it
[20:34] <llogan> then i'll invite them to piss off
[20:35] <michaelni> jeetu, if you want
[20:35] <llogan> basically i just don't feel like cleaning up the video equiment scattered in the office so i'm looking busy...
[20:36] <jeetu> ofcouse i want
[20:36] <michaelni> jeetu, do you have some idea what qualification task you want to do or i should suggest something ?
[20:38] <jeetu> ikk. sir i need your suggestion
[20:39] <jeetu> plz suggest somethind
[20:40] <michaelni> hmm, well, you could write a audio source filter that turns numbers into DTMF tones and a filter that turns DTMF audio back into numbers 
[20:40] <jeetu> sure sir,,thanx for the suggestion
[20:41] <michaelni> np, and dont call me sir please
[20:41] <jeetu> ok
[20:42] <jeetu> michaelni
[20:42] <jeetu> : D
[20:42] <Daemon404> Sir Michael
[20:42] <Daemon404> ill remember that,
[20:44] <Compn> llogan : ok, did you see they are embossed sticker things? so you peel it off and only the lettering remains...
[20:44] <Compn> for ffmpeg, dunno bout the vlc stickers
[20:46] <llogan> next are temporary (neck and trampstamp) tattoos
[20:47] <Daemon404> well
[20:47] <Daemon404> the ffmpeg stickers i took home
[20:47] <Daemon404> look like tabs of E.
[20:47] <Daemon404> <_<
[20:49] <J_Darnley> You've got a party brewing?
[20:59] <prabhy> hey i m new to this organisation n i m interested to work on  FFv1p frame support n Postproc optimisations.can someone please help me which project should be chosen n what qualification task can be performed
[21:03] <J_Darnley> prabhy: I suggest you contact the mentor directly.
[21:03] <J_Darnley> michaelni was here just 20 min ago so he is likely to see this.
[21:05] <michaelni> prabhy, do you know x86 asm ?
[21:08] <prabhy> yes i have done assembly language
[21:09] <Loriker> ### Under http://www.goo.gl/forms/AHVFvPxp8X you find the survey about the Facebook url. This is your chance to your statement there about your personal opinion...
[21:11] <michaelni> prabhy, what exactly ? do you have written some asm code that we could take a look at ? the postproc asm work requires some moderate asm knowledge probably
[21:14] <prabhy> i have learned about this,but i have come to know about this prog very late ,so i havent done any specific work on postproc,but if you ll give me some guidelines ,i will surely give you output as soon as possible
[21:16] <akira4> ubitux, I've sent the patch to the mailing list. 
[21:16] <ubitux> yep, that's good
[21:16] <ubitux> akira4: i will review maybe tonight or tomorrow
[21:17] <akira4> okay. cool
[21:17] <ubitux> i think it can be applied in one iteration
[21:17] <ubitux> anyway, if you want to go further now, we can start to think about the styles
[21:17] <akira4> okay.
[21:18] <ubitux> there are multiple  thing
[21:18] <ubitux> things*
[21:18] <michaelni> prabhy, take a look at lets say do_a_deblock() in postproc, do you think you could convert this from gcc asm to yasm and SSE2 as qualification ? you could also pick a smaller function but if do_a_deblock() just confuses you then maybe you should pick another task without asm
[21:19] <ubitux> akira4: i'll think about the best way to handle this, and will explain
[21:19] <akira4> alright.
[21:19] <ubitux> akira4: for now, i haven't much to suggest; maybe you can start looking at how to use bprint to save all the $foo=bar stuff
[21:20] <ubitux> so they can be transmitted later to the decoder
[21:20] <akira4> oh. okay.
[21:20] <ubitux> the decoder will have to parse them to create an ASS style
[21:20] <ubitux> so whatever way we do that, you need to transmit them to the decoder, and thus you should keep them in a buffer somehow
[21:21] <akira4> I see.
[21:21] <ubitux> i'll try to figure out something because it's not exactly a common model we have here
[21:21] <ubitux> (there is inheritance of the styles, with this weird state changing format)
[21:21] <ubitux> anyway, we'll talk later
[21:22] <akira4> sure. 
[21:27] <iive> if you are looking for tasks, I have one in mind. A filter that cuts video frames/audio samples in given ranges. This would allow e.g. cutting out advertisements in one encode.
[21:27] <ubitux> select/aselect? trim/atrim?
[21:28] <iive> aka you give it a file, with start time and durations, the filter drops the frames/samples in these ranges and corrects the timestaps for these after the range.
[21:28] <ubitux> select/aselect and setpts/asetpts
[21:29] <prabhy> michaelni, yes , i believe i can do it,but i ll ask u about another task if i get totally stuck
[21:30] <prabhy> i ll inform u soon
[21:30] <michaelni> prabhy, ok, 
[21:31] <michaelni> prabhy, also add yourself to the wiki when you start working on a qualification task so other applicants know who works on what
[21:31] <prabhy> so dat means i can start working on it right now,thanx :)
[21:31] <michaelni> sure
[21:31] <iive> ubitux: it might be possible to do this with combination of these filters, but it is not easy or trivial.
[21:32] <ubitux> iive: why?
[21:32] <iive> especially if you have to do multiple cuts.
[21:32] <ubitux> select=not(between(t,30,45)+between(t,123,150))
[21:32] <ubitux> sounds not unbearable to me
[21:33] <iive> what about the setpts and vfr input?
[21:34] <ubitux> i don't know
[21:34] <ubitux> but if you have way to fix/improve this, it's in these filters
[21:34] <ubitux> not by adding more filters
[21:36] <iive> cutting adv is common task, it is good idea to have special filters for common tasks.
[21:39] <cone-712> ffmpeg.git 03Vittorio Giovara 07master:1ac5a29b2e5d: imc: fix order of operations in coefficients read
[21:39] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:cacc1dcadc78: Merge commit '1ac5a29b2e5ddeae068deb9d6e0e803a91941d4d'
[22:03] <cone-712> ffmpeg.git 03Vittorio Giovara 07master:74b023779803: mov: Correctly check the color transfer characteristics range
[22:03] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:16f9f7b881d2: Merge commit '74b02377980321934e33969c84733ace7e9f4eeb'
[22:16] <prabhy> michaelni ,what exactly i have to deblock to yasm in postproc
[22:16] <prabhy> can you please explain abit
[22:16] <Compn> [15:18] <michaelni> prabhy, take a look at lets say do_a_deblock() in postproc, do you think you could convert this from gcc asm to yasm and SSE2 as qualification ? you could also pick a smaller function but if do_a_deblock() just confuses you then maybe you should pick another task without asm
[22:17] <Compn> hmm, the do_a_deblock function in postproc
[22:18] <Compn> so libpostproc/postprocess.c
[22:18] <Compn> around line 438 is the do_a_deblock function
[22:18] <wm4> btw. why is there no "take intrinsic code from openhevc and port it to lavc yasm" task?
[22:18] <wm4> that would be much more useful than libpostproc crap
[22:18] <wm4> because who the fuck uses libpostproc, it's 100% useless
[22:18] <Compn> wm4 : sure, sounds good idea
[22:19] <Compn> wm4 : can you pick out a function to port ?
[22:19] <prabhy> k..thanx :)
[22:19] <wm4> Compn: I don't know the openhevc code, but I heard significant optimizations are missing from lavc
[22:21] <Compn> wm4 : are there any sample commits which port openhevc code to lavc so that students dont have to figure out the api and cpu switching code ? maybe ? :)
[22:21] <Compn> i'm just trying to make it easier, this is not a requirement
[22:22] <michaelni> wm4, do you volunteer to mentor such a openhevc project ?
[22:22] <Compn> i was going to ask that but i didnt want to scare wm4 off :(
[22:23] <wm4> michaelni: no, but I bet there are some who could+would do that
[22:23] <wm4> by "many" I mean more than 1
[22:23] <Compn> who are the hevc devs in here
[22:23] <michaelni> wm4, dont bet, bring them here and make them say they want and will do it
[22:23] <Compn> (we should keep a list or something)
[22:24] <wm4> Daemon404, jamrial: cc ^
[22:24] <Compn> porting hevc stuff would be nice :)
[22:24] <wm4> and whoever Christophe Gisquet is
[22:24] <J_Darnley> I've looked at the openhevc instrinsics and it is not at all clear what has already been done.
[22:25] <J_Darnley> Not to mention I can hardly decipher the intrinsics themselves
[22:25] <Compn> has anyone benchmarked it ?
[22:25] <Compn> vs ffmpeg that is
[22:26] <Compn> i dont doubt its faster, but also i doubt things for fun
[22:26] <nevcairiel> over 100% faster in 64-bit
[22:26] <kierank> it is much faster
[22:27] <nevcairiel> but porting the intrinsics to clean asm isn't an easy task
[22:28] <michaelni> prabhy, Compn do_a_deblock() is in libpostproc/postprocess_template.c
[22:28] <michaelni> line 2547
[22:28] <Compn> oops :)
[22:28] <wm4> well, wouldn't it make a good project?
[22:29] <wm4> I find the whole "Postproc optimizations" project very questionable
[22:29] <wm4> we don't need AVX optimization of state of the art open source tech from 2001
[22:29] <Compn> wm4 : did you remove -vf pp from mpv ?
[22:29] <wm4> not yet
[22:29] <Compn> do it!
[22:29] <Compn> :)
[22:30] <Compn> or are you waiting for replacement first ?
[22:30] <michaelni> wm4, postprocess is SIMD code from MMX times theres alot that can be improved, hevc can be improved too but ATM HEVC has no mentor and i suspect its harder
[22:30] <wm4> I never heard of anyone using it, and the replacement is vapoursynth (portable avisynth-like), which is _actually_ used
[22:30] <Compn> oh ok
[22:30] <Compn> forgot about vaporsynth
[22:30] <J_Darnley> Is that still written in c++?
[22:30] <wm4> J_Darnley: vapoursynth? yes, but it has a C API
[22:30] <wm4> and filters can be written in C
[22:32] <J_Darnley> From the 2 c++ projects I've actually looked at, "can be" seems to mean "never" in practice.
[22:32] <wm4> I'm using the API from C code
[22:33] <Compn> michaelni : maybe you can email prabhy and make sure shes looking at postprocess-template :(
[22:33] <Compn> blame me for misdirection 
[22:33] <wm4> J_Darnley: this is the API: https://github.com/vapoursynth/vapoursynth/blob/master/include/VapourSynth.h
[22:34] <J_Darnley> Sure, that's nice for people using vapoursynth, but what dissuaded me from contibuting when it first started was everything being c++
[22:35] <nevcairiel> why do people care what something is implemented in, as long as the external API can be used without any pain
[22:35] <wm4> entire C projects are being converted to C++ these days... could be worse
[22:35] <Compn> you guys love to bikeshed :P
[22:35] <Compn> [16:28] <michaelni> prabhy, Compn do_a_deblock() is in libpostproc/postprocess_template.c
[22:35] <Compn> [16:28] <michaelni> line 2547
[22:35] <Compn> prabhy ^^
[22:35] <Compn> i got it wrong :P
[22:35] <J_Darnley> nevcairiel: because I wasn't interested in the external API
[22:36] <prabhy> got do_a_deblock() 
[22:36] <wm4> not avisynth, that had a C++ API
[22:36] <wm4> *now
[22:36] <michaelni> well, the code in postprocess.c is do_a_deblock too, just the c code
[22:36] <prabhy> i found it
[22:36] <wm4> and it was (is) terrible
[22:36] <J_Darnley> I agree, that is one of the 2 c++ projects I have actually seen the source code for.
[22:37] <wm4> there's even a C filter in the main repo (with a funny name): https://github.com/vapoursynth/vapoursynth/blob/master/src/filters/assvapour/assvapour.c
[22:37] Action: J_Darnley wonder whether it is subtitles
[22:38] <prabhy> k...so postprocess.c  do_a_deblock is not wat i have to search
[22:38] <prabhy> so got dat wrong
[22:38] <wm4> J_Darnley: yes, of course
[22:38] <prabhy> k...going to search it again
[22:42] <iive> prabhy: to put it simply. the current code uses inline assembler, that is in .c file.  you have to get this assembler and convert it to an .asm file that could be compiled using yasm or nasm. we also use some macro written by the x264 project.
[22:43] <prabhy> k...got dat
[22:43] <iive> the macro is in libavutil/x86/x86inc.asm
[22:44] <iive> macros, there are more than one macros there.
[22:45] <kierank> are you all out of your mind
[22:45] <kierank> asking opw students to work on postproc
[22:45] <iive> I wonder if we have a documentation for it somewhere. I might be able to dig an old coversation where the usage was explained to BBB
[22:45] <prabhy> k...but we need to use particular macro to convert it into .asm file
[22:46] <iive> kierank: would you mentor HEVC asm?
[22:46] <prabhy> if you can arrange that ,it would be really helpful
[22:46] <iive> prabhy: well, macros are like preprocesor #defines
[22:46] <prabhy> i know
[22:47] <prabhy> its combination of statements
[22:47] <kierank> iive: if necessary yes but i haven't written asm for a long time
[22:47] <iive> well, there are some basic stuff, like defining function prototype
[22:47] <iive> kierank: at least you have written asm :)
[22:47] <kierank> I've also flown a plane for 10s. Doesn't mean I should be flying 747s
[22:48] <iive> or tricks that help you write one function that could work with 64bit MMX or 128bit SSE or 256 AVX, depending on how it is compiled.
[22:48] <iive> kierank: if there is an emergency we'll call you first.
[22:49] <iive> kierank: at least you know what all these instruments do.
[22:49] <prabhy> kierank: theres no harm in trying :)
[22:50] <kierank> prabhy: read wiki.videolan.org/x264asm
[22:50] <kierank> i mean https://wiki.videolan.org/X264_asm_intro/
[22:50] Action: iive bookmarks this.
[22:51] <prabhy> k i ll check it out :)
[22:52] <J_Darnley> Shit, was that really 4 years go?
[22:53] <jeetu> can anybody tell me whether ,,audio converter to convert number into DTMF tones is same as the one we use  in the keypad of our cell phones??
[22:53] <J_Darnley> I should update it to cover named arguments.
[22:54] <reynaldo> Daemon404: sudden interest -> they actually want to collaborate, well, the Samsung OSG goes
[22:54] <michaelni> jeetu, yes that was the idea 
[22:54] <reynaldo> Daemon404: they doint thing wrong in the past (and still) thats likekly not the OSG (recently formed a year or so ago), but they are trying to help those situations been resolved. I know they are currently working one one particular one pertaining our project
[22:55] <reynaldo> Daemon404: they are a massive company, the OSG means well and is trying to make a difference but inflicting change on the whole structure will take time
[22:56] <reynaldo> they are still trying to and directly supporting us so I'd say there's hope :)
[22:56] <Daemon404> oic
[22:56] <kierank> hmm I wonder if I submit opus in ts patch to ffmpeg, my smart tv will support it
[22:57] <kierank> though actually I guess it will be detected as mp3
[22:57] <kierank> like everything
[22:57] <wm4> wut, hw using ffmpeg ts demuxer?
[22:57] <jeetu> okay,,, u mean i have to convert all the nine numbers into their corrosponding frequency tones,,,,ryt??
[22:57] <kierank> wm4: for the watching things on a usb stick
[22:57] <nevcairiel> wm4: many use ffmpeg for their generic file parsing needs
[22:58] <kierank> the main broadcast ts demuxing goes thorugh hardware
[22:58] <kierank> because of the dynamic changes
[22:59] <wm4> dynamic changes?
[23:00] <kierank> pid changes
[23:00] <kierank> that sort of stuff
[23:00] <kierank> drm too
[23:00] <wm4> is there a reason this couldn't be supported in software?
[23:00] <wm4> oh, drm
[23:03] <michaelni> jeetu, yes one source filter that takes the numbers from the command line (also #) and a 2nd filter that takes recoreded audio and turns the tones back into numbers
[23:03] <jeetu> okay !!!! got it 
[23:03] <jeetu> thanks alot :)
[23:04] <kierank> what i would like to see is a lipsync filter
[23:04] <kierank> i.e vf testsrc outputs audio as well
[23:08] <jeetu> sorry,,, didn't get your point @kierank??
[23:08] <kierank> not really related to your project
[23:08] <kierank> but it's a nice thing to have in libavfilter
[23:08] <jeetu> okay 
[23:09] <wm4> kierank: that sounds like useful idea
[23:09] <jeetu> okay
[23:09] <kierank> wm4: yeah but vfr people will ruin it as always
[23:09] <kierank> it only works properly in cfr mode
[23:09] <wm4> how so?
[23:10] <wm4> wouldn't it just output audio and video frames with timestamps?
[23:10] <wm4> maybe I'm misunderstanding what it's supposed to do
[23:11] <kierank> wm4: something like https://www.youtube.com/watch?v=fQdgNSeJQgg
[23:12] <wm4> why does it not work with vfr?
[23:14] <kierank> because the whole point of a lipsync test is to have an integer number of samples between your beep
[23:14] <kierank> vfr makes this all totally unbounded
[23:15] <wm4> I thought it was matching up audio and video timing
[23:15] <kierank> no it's producing raw audio in video that is known to be in sync
[23:16] <kierank> so that you can figure out where (if any) sync is lost
[23:16] <kierank> e.g aac priming samples
[23:16] <kierank> because for audio timestamps are merely a first order approximation
[23:16] <kierank> because every encoder does it's own priming busines
[23:17] <wm4> what a mess
[23:18] <cone-712> ffmpeg.git 03Martin Storsjö 07master:752e71e74f50: rtpdec_hevc: Rename a variable for clarity
[23:18] <cone-712> ffmpeg.git 03Michael Niedermayer 07master:60a876fe7b50: Merge commit '752e71e74f50e7a6f9a19edb8e775b2ea2fb94d8'
[00:00] --- Thu Oct 16 2014


More information about the Ffmpeg-devel-irc mailing list