[Ffmpeg-devel-irc] ffmpeg-devel.log.20150304

burek burek021 at gmail.com
Thu Mar 5 02:05:02 CET 2015


[00:27] <cone-226> ffmpeg 03Tobias Rapp 07master:b4fbad2634c4: libavformat/avienc: Fix duration of audio segment in OpenDML master index
[01:32] <cone-226> ffmpeg 03Anton Khirnov 07master:db5cc75f8b56: wma.h: #include "avcodec.h"
[01:32] <cone-226> ffmpeg 03Michael Niedermayer 07master:0d20896775a2: Merge commit 'db5cc75f8b5620e7cd2d05762c3b087ab16d8c24'
[02:24] <cone-226> ffmpeg 03Michael Niedermayer 07master:f6f1e2b3be8f: avcodec/pngenc: Remove outer () from macro
[03:17] <cone-226> ffmpeg 03Michael Niedermayer 07master:7da7d269b840: ffplay: Switch to show waves mode if allocation/init of RDFT fails
[04:56] <Timothy_Gu> :q
[04:56] <Timothy_Gu> sorry too much vim recently
[04:58] <Compn> :wq!
[06:25] <pcordes> Hey, did anyone see my email to the list from earlier tonight?  I'm new to contributing to ffmpeg, so I figured I'd hop on IRC
[06:25] <pcordes> in case that would be easier for anyone to point me in the right direction
[06:25] <pcordes> http://ffmpeg.org/pipermail/ffmpeg-devel/2015-March/169679.html
[06:39] <ninten> hi all i am new to ths org and want to contribute i have gone to gsoc2015 ideas page and the project interest me a lot. so someone can guide me with initial steps !!\
[06:42] <pcordes> Well I've lurked around open source projects enough to maybe be able to help.  I'm also new to contributing to ffmpeg, though
[06:43] <pcordes> Did you have any specific project in mind already, or are you looking to talk over ideas for things that need doing, with potential mentors?
[06:45] <ninten> yeah i have gone through FFv1 P frame support project, I have little bit of idea about this and i want to explore more about this !!!
[06:46] <pcordes> Ok, then it sounds like you're pretty much ready to go, and just need to find someone who actually knows anything (not me :/)
[06:47] <pcordes> I think sending an email to the ffmpeg-devel mailing list would be your best bet, as a next step
[06:47] <pcordes> as far as "step 3. Contact them" goes
[06:48] <pcordes> Probably just say who you are, that you want to work on FFv1, and a bit about some your ideas in detail
[06:50] <ninten> yeah thats what i am going to now follow step 3
[06:50] <pcordes> That way, people that wrote the code can see if your ideas look good, or if you might need to rethink something.  And if your ideas are good, then give suggestions for more detailed things you might want to try
[06:50] <pcordes> good luck :)
[10:37] <BtbN> Hm, there still is a "race" between the dash muxer and the aac adtstoasc filter. Is there any way to make sure a muxer is not initialized before a bitstream filter has done its work(of adding the extradata)?
[10:44] <BtbN> In particular, the problem isn't only that the codec string in the manifest wasn't update once adtstoasc fixed the extradata, it's also the initial segment which is wrong, as it is written before the extradata is fixed up.
[10:44] <BtbN> Fixing up the codec string was trivial, but re-writing the initial segment does not look like it can be easily done.
[11:08] <wbs> BtbN: the dash muxer uses the "delay_moov" flag now, which should delay writing the moov tag until it actually writes the first fragment. what's missing is copying the extradata from the dash muxer to the inner chained mp4 muxer though, in case it wasn't present in write_header (the delay_moov feature was mostly designed to get edit lists for b-frames right)
[11:09] <nevcairiel> the underlying problem is that the bitstream filter can only write the new extradata after it already processed the first frame, so during init its just not present .. but for the first segment it should be, i imagine
[11:11] <wbs> yes, it should be - I assume it is set when writing the first packet or so
[11:12] <wbs> and with delay_moov (as the dash muxer uses), the init segment is written only after the first full gop is written
[11:12] <wbs> BtbN added code for updating the codec id string in case extradata wasn't present during init, the same would be necessary to copy the extradata to the inner muxer
[11:12] <wbs> shouldn't be too hard
[11:15] <nevcairiel> there is a similar problem when trying to remux adts-aac from ts to mkv, since mkv doesnt do adts, but during init it doesnt have extradata yet either...
[11:16] <wbs> in general this sounds like something that may be worthwhile fixing on a higher level as well, to avoid having to add extra trickery to all muxers
[11:17] <wbs> (the delay_moov stuff also sounds like such hackery, but it's necessary for more reasons than just late extradata)
[11:17] <wbs> similarly, some encoders might not be able to produce extradata on init, only when the first packet is written
[11:33] <cone-531> ffmpeg 03Martin Storsjö 07master:2a66a580678d: rtpdec_mpa_robust: Fix incrementing split_pos
[11:33] <cone-531> ffmpeg 03Michael Niedermayer 07master:5dce723715e1: Merge commit '2a66a580678dd9401f4d95e01e0958ca51864b6f'
[12:15] <cone-531> ffmpeg 03Mark Reid 07master:ec5a4af560aa: libavformat/mxfdec: refactor reading strong ref array
[12:40] <poste9> can u guys tell me when or why using CONFIG_MEMALIGN_HACK?
[12:41] <nevcairiel> its used when the OS you are building for does not provide an aligned memory allocator
[12:41] <nevcairiel> although most operating systems have one these days
[12:48] <poste9> well.. thats weird setting this metadata av_dict_set(&ost->stream->metadata, "encoder",  "libx264", AV_DICT_DONT_STRDUP_VAL | AV_DICT_DONT_OVERWRITE); when I use avformat_free_context I get a crash trying to free this setting. If I remove it, it works perfectly... do you see any reason for that? 
[12:49] <poste9> doesnt look ok the fact that I cant even see where the crash happens since its inside the free() function
[12:49] <nevcairiel> you should not use dont_strdup for constant strings
[12:49] <poste9> k
[12:50] <nevcairiel> dont_strdup means that it takes ownership of the string and will free it when its time to do so, this doesnt work for constants
[12:50] <poste9> thanks.. that solved the problem
[12:50] <poste9> yea makes sense
[12:56] <wm4> AV_DICT_DONT_STRDUP_VAL should just die
[12:56] <wm4> it's 100% useless
[12:56] <wm4> and a microoptimization that doesn't matter at all
[12:56] <poste9> is there any reason to believe av_dump_format would take a filename != from ctx->filename ?
[12:59] <rmklp> What is the usual workflow to update fate checksums for files affected by a code change? Run make GEN=1 fate and then look at the files that changed using git status and check if the output makes sense? 
[13:16] <poste9> do I need to send pull request to the mail  list?
[13:26] <thardin> rmklp: sounds about right
[13:26] <thardin> then commit && git format-patch
[13:28] <rmklp> thardin: alright, thanks
[13:29] <thardin> you can set up git to e-mail patches directly, which might be useful
[13:56] <rmklp> thardin: yes, I know. Havent done that yet out of pure lazyness. I probably should.
[13:59] <thardin> tbh I haven't either
[14:30] <poste9> can u take a look at this and tell me if the code duplication is an error pls ? https://github.com/FFmpeg/FFmpeg/blob/master/ffmpeg.c#L3601
[14:30] <poste9> line 3601 of ffmpeg.c
[14:36] <ubitux> dts ` pts
[14:38] <poste9> my bad
[14:38] <poste9> thanks
[14:44] <tar> Hi
[14:44] <tar> I am a GSoc 2015 aspirant and wish to contribute in postproc optimizations. Guide me on how to start
[14:52] <Compn> tar : wait around, its a little early for everyone 
[14:52] <Compn> :)
[14:52] <Compn> and welcome.
[14:52] <Compn> have you written asm /cpu instructions before?
[14:53] <tar> I am very proficient in ARM Assembly language
[14:58] <wm4> yeah, let them work on worthless crap like postproc filter that made sense in 2001
[15:00] <Daemon404> never mind stuff like HEVC...
[15:01] <wm4> who needs HEVC? most video content is in mpeg2!!!!!!1111111111
[15:11] <tar> Compn : Can you please guide me further?
[15:11] <Daemon404> the qualification task should be listed on the wiki, no?
[15:14] <kierank> 1:58 PM <"wm4> yeah, let them work on worthless crap like postproc filter that made sense in 2001
[15:14] <kierank> +1
[15:15] <iive> doesn't hevc emply loopfilter?
[15:15] <iive> employ
[15:16] <nevcairiel> every modern codec has a loop filter, but its built into the codec directly
[15:16] <wm4> the point is, HEVC could use some optimizations... or just port them from OpenHEVC
[15:17] <JEEBsv> yeah
[15:31] <kierank> should have a task for reindenting code
[15:31] Action: Daemon404 stabs kierank 
[15:32] <Rodeo> kierank: I thought your said it made cross-project merges harder?
[15:32] <kierank> hence the troll, yes
[15:33] <Daemon404> if you read their ML
[15:34] <Daemon404> even they are telling <obviousperson> to stop reindenting
[15:34] <ramiro> you mean their senior software indentation engineer?
[17:44] <michaelni> kierank, do you want to mentor a hevc optimization task?
[17:44] <kierank> no
[17:44] <michaelni> then dont troll
[17:45] <kierank> I'd do libavfilter
[17:45] <kierank> sure
[17:45] <kierank> I never said hevc anywhere
[17:45] <kierank> and gsoc is not the new deal, it's about making good tasks not as many as possible
[17:45] <michaelni> if you want to mentor some libavfilter work, that would be great too !
[17:46] <michaelni> wm4, do you want to mentor some task ? hevc optimizations or other ?
[17:48] <michaelni> also gsoc is not just about having tasks that are great and usefull and shiny its also about attracting new developers some of whom might stay in the project after gsoc
[17:53] <michaelni> and the number of really great and shiny tasks that are doable for students is not that huge so having easier and less shiny tasks does make sense
[17:54] <kierank> the tasks should not be drudgery
[17:54] <BtbN> wbs, nevcairiel: https://gist.github.com/BtbN/a83cc8b9ba313b697faa haven't yet tested it(Other than that it propperly updates the codec string in the manifest).
[17:55] <michaelni> kierank, yes, which tasks are & how can it be improved ?
[17:55] <kierank> I am going to add a task
[17:56] <himangi> Any suggestions for the name of a inverse telecine filter?
[17:57] <kierank> ivtc
[17:57] <Daemon404> i dont understand the point of suck a task... we have two already
[18:02] <michaelni> Daemon404, i think its part of the qualification task for the VDPAU filter
[18:02] <Daemon404> seems pretty silly...
[18:03] <michaelni> please talk with carl if you have a better idea
[18:05] <himangi> michaelni: I see fieldmatch and pullup.. I was also going to ask what must be different for this one
[18:07] <michaelni> himangi, best ask carl, i dont know what he had in mind exactly, i think he will reply quickly by mail, sadly he isnt on irc
[18:08] <himangi> yes.. I'll ask him
[18:08] <michaelni> himangi, you can probably also suggest carl some qualification task if you have something specific or different in mind
[18:08] <himangi> sure michaelni :)
[18:14] <ubitux> yeah ivtc with vdpau would be interesting
[18:15] <ubitux> but maybe a "useful" task could be good
[18:15] <ubitux> like maybe something else related to vdpau, but dunno
[18:20] <michaelni> kierank, btw, great idea!
[18:23] <ubitux> if someone wants some tasks for lavfi, https://ffmpeg.org/pipermail/ffmpeg-devel/2014-October/164438.html
[18:23] <ubitux> but yeah, api etc.
[18:24] <michaelni> ubitux, i ve the feeling this is too hard
[18:24] <ramiro> do we have a page to track students working on qualification tasks (like we had for opw)?
[18:24] <ubitux> michaelni: yes
[18:25] <michaelni> ramiro, i think theres a dead link, someone has to create the page or remove the link
[18:25] <michaelni> ramiro, dont hesitate to do one of these
[18:28] <Daemon404> eh
[18:28] <Daemon404> dont get students to design APIs
[18:28] <Daemon404> that happened in the past, and it was bad
[18:29] <ramiro> any ideas on how many slots we will have btw?
[18:29] <wm4> Daemon404: which ones?
[18:29] <Daemon404> trying to remember
[18:30] <michaelni> ramiro, no clue at all, anything would be a pure guess
[18:30] <nevcairiel> didnt some gsoc student try to design this new seek api that was never used?
[18:31] <Daemon404> ... was it lavfi that was a stuent api originally?
[18:31] <Daemon404> i swear it was
[18:31] <Daemon404> sorry
[18:31] <Daemon404> "api"
[18:32] <ramiro> michaelni, ok, so about 20? =)
[18:32] <michaelni> ramiro, i suspect it will be fewer ;)
[18:32] <ramiro> dang
[18:34] <Daemon404> 20 would be quite a few spots
[18:35] <kierank> 5 at most
[18:41] <himangi> baptiste_, ping
[18:42] <michaelni> i would guess in the 3-9 range if i had to guess something
[18:50] <Loriker> hi, does anyone have some additional hints how I can maximize the throughput of x264 encoding/transcoding - I have 2 streams running in parallel? (I have already tried to use "-crf 0" for losless compression , "-threads x" to align the number of threads to the number of CPU cores, "-preset ultrafast" to test with low quality and fast processing)
[18:51] <Daemon404> for one, let threads auto do its work
[18:51] <Daemon404> threads==cores is not ideal for libx264
[18:51] <Loriker> Daemon404: yes, you mean cores * 1,5 ...as it is used by "auto"
[18:52] <Daemon404> yes
[18:52] <Daemon404> and lossless will likely be slower than lossy
[18:52] <wm4> what's the optimal multiplier for h264 _decoding_?
[18:52] <Loriker> but I have two streams running in parallel... 
[18:52] <Daemon404> the entropy coder will have a LOT mroe data to compress
[18:52] <wm4> and does lavc do enable threads by default yet?
[18:52] <Daemon404> wm4, for ages
[18:52] <wm4> wow, how progressive
[18:53] <Loriker> :-)
[18:55] <nevcairiel> auto-thread-count for decoding uses nb_cpus+1 btw, not 1.5x
[18:56] <Daemon404> nevcairiel, not for libx265
[18:57] <Daemon404> er libx264
[18:57] <Daemon404> it passes through auto
[18:57] <nevcairiel> *decoding*
[18:57] <nevcairiel> :P
[18:58] <Daemon404> o ok
[18:58] <Daemon404> misread
[18:59] <nevcairiel> whats the best thread count for encoding then? auto?
[19:02] <wm4> <nevcairiel> auto-thread-count for decoding uses nb_cpus+1 btw, not 1.5x <- what's the reasoning behind this?
[19:03] <nevcairiel> heuristics?
[19:04] <wm4> http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/pthread_frame.c;h=5a4ab84a4ac5e1d26b41844c67240168753b61d5;hb=HEAD#l620
[19:04] <wm4> interesting
[19:05] <nevcairiel> i did some benchmarking some time ago and there definitely is a benefit from using more than #cores, but the benefit goes down quite a bit with every extra thread
[19:05] <nevcairiel> if the sweet spot is +1 or 1.5x or something can be argued for a long time
[19:06] <wm4> heh
[19:06] <wm4> wow I was confused with this av_dict "patch" on the ML
[19:06] <wm4> so michaelni reviewed the patches by copying the diffs into his mail client??
[19:06] <nevcairiel> many do that
[19:07] <himangi> reynaldo: hey!
[19:08] <wm4> it's quite confusing
[19:08] <wm4> and like the worst of both worlds of ML reviews and GH PRs+reviews
[19:14] <Daemon404> wm4, its a consequence of git patches being mbox...
[19:14] <Daemon404> but yeah.
[19:15] <cone-531> ffmpeg 03Martin Storsjö 07master:62139b14e621: fate: Specify the idct to use for the aic-oddsize test
[19:15] <cone-531> ffmpeg 03Michael Niedermayer 07master:1c9f2ebdcd07: Merge commit '62139b14e621f096d0f8ed90920d042b92867e40'
[19:16] <wbs> BtbN: looks pretty good - you may want to move the call into the if(!init_range_length) block, that's where the init segment is written
[19:19] <BtbN> wbs, still doesn't seem to work though. Not sure if it's the extradata breaking it though. No idea what else would though.
[19:21] <wbs> BtbN: hmm, ok. I'll see if I can reproduce and fix it, unless you beat me to it
[19:21] <BtbN> Would be very helpfull if ffprobe was able to read any of the segments
[19:21] <cone-531> ffmpeg 03Martin Storsjö 07master:1dc19729e92a: rtpdec_asf: Don't free the payload context in the .close function
[19:21] <cone-531> ffmpeg 03Michael Niedermayer 07master:634ce8ddad6b: Merge commit '1dc19729e92a96620000e09eba8e58cb458c9486'
[19:21] <Daemon404> BtbN, for which info?
[19:22] <BtbN> Daemon404, the extradata, codec info and such. Basicaly everything that's put into the init segment
[19:22] <BtbN> Would make it easier to compare the working and the broken one
[19:23] <Daemon404> im not sure how it could; that info isnt present
[19:23] <Daemon404> (except in init)
[19:24] <BtbN> yes, i the init segment is what i want to look at
[19:24] <Daemon404> try L-SMASH's boxdumper perhaps
[19:24] <Daemon404> i sue it often to inspect isom files
[19:24] <Daemon404> use*
[19:28] <wbs> BtbN: concatenate the init segment with a segment file, and you'll have a working file that any tool should be able to read
[19:28] <BtbN> ah, thanks. Didn't realize that.
[19:29] <wbs> that's basically what the segmenter muxer does; takes the linear output from the mp4 muxer and chops it into separate files
[19:29] <wbs> and since they're independent you can just cat them together, skipping some of the earlier segments if you want
[19:30] <henna_> Hey I am a student from India and am interested in contributing for gsoc 
[19:32] <henna_> I found some interesting ideas.. I was particularly interested in the VDPAU filter project. I see the mentor is :cehoyos . Can I please get some help regarding the same 
[19:36] <BtbN> wbs, the segment chrome refuses to play works fine when played with ffplay.
[19:36] <BtbN> The only difference i can spot is the bitrate beeing just a guess
[19:38] <wbs> BtbN: weird. can you send me the input data and all the stuff for reproducing it, and I'll have a look
[19:38] <BtbN> I'm using random livestreams from twitch as test input
[19:39] <henna_> :cehoyos where can I find the documentation for running the current filter? 
[19:39] <BtbN> cehoyos isn't here.
[19:40] <BtbN> He was asked to join this channel for gsoc, but hasn't shown up so far.
[19:41] <BtbN> wbs, https://btbn.de/mse/index.html is what i use as test-player (Works only in Chrome and IE11) https://btbn.de/mse/stream.sh is the script I use for re-streaming from twitch
[19:43] <BtbN> https://github.com/BtbN/FFmpeg/tree/master is what i'm using for testing, it has my extradata patch on top of a somewhat recent master.
[19:48] <Compn> wm4 / kierank : can make hevc optimizations a qualifying task :)
[19:48] <cone-531> ffmpeg 03Diego Biurrun 07master:25f613f8be3b: dca: Move syncword definitions to a separate header
[19:48] <cone-531> ffmpeg 03Michael Niedermayer 07master:87db7e777262: Merge commit '25f613f8be3b51e4396b93cda131e4631ba54302'
[19:49] <Compn> henna_ : please email carl about vdpau task, here is his email , Carl Eugen Hoyos <ce at hoyos.ws>
[19:50] <BtbN> wbs, "Video Element Error: MEDIA_ERR_SRC_NOT_SUPPORTED" that's what chrome/dash.js has to say about it.
[19:52] <BtbN> If i copy the init segment from when encoding re-audio over the broken one from copy, chrome plays it fine.
[20:02] <BtbN> Daemon404, that l-smash boxdumper tool is nice.
[20:02] <BtbN> wbs, the broken file is missing the "[tag = 0x05: DecoderSpecificInfo]" section in boxdumper output. So i guess the extradata copy doesn't reach the mp4 muxer.
[20:03] <cone-531> ffmpeg 03Vittorio Giovara 07master:87e85a133f3c: aac: Relax reserved_bit validation
[20:03] <cone-531> ffmpeg 03Michael Niedermayer 07master:b5ca15893549: Merge commit '87e85a133f3ce2f037b90e9c7bbca99951df6c15'
[20:08] <tar> Hi..
[20:09] <tar> I wish to contrtibute in this Postproc optimizations project....Please tell me how to start..
[20:13] <Compn> tar : ok, michaelni is the mentor for that project. 
[20:13] <tar> Compn Thank You..
[20:14] <Compn> tar : the qualifying task says to optimize some code for SSE2 and AVX2
[20:14] <Compn> some of the postproc code that is mmx2 
[20:15] <Compn> so pick a mmx2 function in postproc and write it for sse2, and make sure to benchmark the code and also make sure its binary identical
[20:15] <Compn> which means you have to have a build environment to build ffmpeg and test your changes of course
[20:16] <henna_> hey, I'm having some trouble building examples
[20:16] <Compn> what problem ? 
[20:16] <Compn> or paste problems to pastebin 
[20:16] <Compn> if they are longer than a few lines :)
[20:16] <Compn> probably missing some dev header or lib ?
[20:17] <michaelni> tar, yes as Compn said
[20:17] <henna_> :Compn demuxing_decoding.c:34:33: fatal error: libavutil/timestamp.h: No such file or directory
[20:17] <henna_>  #include <libavutil/timestamp.h>
[20:17] <henna_>                                  ^
[20:17] <henna_> compilation terminated.
[20:17] <henna_> make: *** [demuxing_decoding.o] Error 1
[20:17] <Compn> henna_ : yeah paste to pastebin site so we can check it all...
[20:17] <tar> michaelni : I have knowledge of ARM assembly language.....
[20:18] <michaelni> tar, you want to optimize postproc for  ARM ?
[20:18] <henna_> i executed PKG_CONFIG_PATH=pc-uninstalled make. in my doc/examples folder 
[20:18] <michaelni> instead of SSE2/AVX2 ?
[20:19] <henna_> are there any steps required to be executed before this
[20:19] <henna_> :)
[20:19] <tar> I am a new developer....I want to know where to gather knowlege of this instruction set...
[20:20] <michaelni> tar, theres some docs from intel, give me a moment ill try to find a link
[20:20] <tar> Thank you..
[20:20] <henna_> https://www.irccloud.com/pastebin/YmtJfHJ3/error
[20:22] <Compn> henna_ : do you have ffmpeg/libavutil/timestamp.h ? how did you check out ffmpeg source ?
[20:22] <Compn> or are you building examples out of dir or something s trange ?
[20:23] <henna_> i cloned the repo from github
[20:23] <henna_> i don't think any file is missing
[20:24] <Compn> well do you need examples for your project ? if not , dont bother with them ;)
[20:25] <henna_> i wanted to execute the vdpau.c file, but it returns an error -> vdpau.h not found 
[20:25] <henna_> so I thought, maybe checking some examples would help me solve the issue
[20:26] <michaelni> tar, http://www.intel.com/content/www/us/en/processors/architectures-software-developer-manuals.html its a big big though, you wont need most of it
[20:26] <ab_del> Compn: So some examples can be broken/redundant?
[20:26] <michaelni> tar, the instrcution set reference there should contain all sse2&avx2 instructions
[20:27] <Compn> ab_del : examples are not required for building ffmpeg.
[20:27] <tar> I'll take a look...thank you very much..
[20:27] <Compn> i dont know if they are broken or not, i havent built them in a long time.
[20:27] <ab_del> Compn: I understand that. I'm just asking
[20:27] <Compn> henna_ : did you install vdpau headers on your system ? 
[20:28] <Compn> henna_ : also what os and shell are you using ?
[20:28] <henna_> i am using linux 14.04 
[20:28] <Compn> mint or ubuntu i guess ?
[20:28] <henna_> and I just cloned the repo
[20:28] <Compn> ok :)
[20:28] <henna_> ubuntu
[20:29] <Compn> henna_ : what video card do you have ?
[20:29] <Compn> nvidia ? intel ?
[20:29] <henna_> so I think cloning the repo would have installed headers on my system
[20:30] <Compn> well, vdpau is another project
[20:30] <Compn> so we dont include the system vdpau headers...
[20:30] <Compn> sudo apt-get install libvdpau1 vdpau-va-driver
[20:30] <Compn> try running that
[20:33] <henna_> nvidia video card 
[20:33] <henna_> and even after running that command, the code does not compile
[20:33] <henna_> with gcc
[20:33] <Compn> what was your configure line ?
[20:34] <Compn> oops need another lib
[20:34] <Compn> sudo apt-get install libvdpau-dev
[20:34] <Compn> run that
[20:35] <Compn> henna_ : i am guessing here , its been a while since i had to set these things up.
[20:35] <henna_> its fine :) 
[20:35] <Compn> so please forgive any of my ideas :)
[20:36] <henna_> config.h missing now 
[20:37] <Compn> henna_ : ok, did you run ./configure in ffmpeg dir ?
[20:37] <henna_> no, let me try that
[20:37] <Compn> henna_ : or maybe read the build guide on ffmpeg documentation or wiki ?
[20:37] <Compn> because it explains how to build ffmpeg...
[20:37] <henna_> okay
[20:37] <Compn> then you dont have to ask us , questions that are frequently asked... and then it proves to us that you can find, read and follow documentation :)
[20:40] <henna_> sorry, but now I am getting a new problem -> buffer.h not found
[20:41] <ab_del> https://trac.ffmpeg.org/wiki
[20:41] <Compn> http://trac.ffmpeg.org/wiki/CompilationGuide
[20:41] <Compn> http://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu
[20:41] <Compn> for ubuntu ^
[20:42] <Compn> henna_ : follow that ^
[20:46] <poste9> I instaled and reinstaled at several ubuntu machines, server and client. With ^ that compilation guide for ubuntu, and neveer got a single error... just follow and its success garantee
[20:48] <Compn> its good to know those instructions work , thanks poste9 
[20:51] <poste9> does ffmpeg use OutputFilter with this command:?  ffmpeg -f v4l2 -i /dev/video0 -c:v libx264 -t 5 -qmin 30 -qmax 51 -f  mp4 output.mp4
[20:52] <poste9> I supposed I didnt need filters but doesnt look like
[20:52] <Compn> what part are you working on again ?
[20:52] <poste9> Im building my own version of ffmpeg
[20:53] <Compn> ok, which output filter? 
[20:53] <poste9> I dont  need any filter
[20:53] <poste9> I just need to record camera to file
[20:53] <poste9> and I need control of that to make optimization, reports and stuffs..
[20:53] <Compn> then yes that should save camera to file, assuming your input stuff is correct
[20:54] <poste9> yes yes.. It does... but Im coding my version of  ffmpeg myself, and the real question is if I need to use OutputFilter to do  so.
[20:54] <poste9> I dont need audio n/or filters 
[20:55] <Compn> oh you want to build an ffmpeg and disable the stuff you dont need 
[20:55] <Compn> correct, you dont need outputfilter to do that
[20:55] <poste9> I know the output pix_fmt and encoder, so I dont need to "discover it"
[20:56] <poste9> well.. thats weird 
[20:56] <Compn> the ffmpeg_g binary is the one that has symbols in debugging, just fyi...
[20:57] <poste9> @reap_filters function theres this line: if (!ost->filter) continue; and ffmpeg doesnt return with  ^ that command line.
[20:58] <Compn> you want to specify the formats to speed up or skip the probing ? 
[20:58] <wm4> "Unfortunately XBMC is using these semi-private fields, so it gets broken by this change." *rimshot*
[20:59] <nevcairiel> who cares
[20:59] <Compn> about kodi ?
[20:59] <wm4> nevcairiel: the maintainer of ffmpeg, I bet
[21:00] <nevcairiel> xbmc uses a private copy of ffmpeg anyway
[21:00] <nevcairiel> they can adapt
[21:00] <cone-531> ffmpeg 03Luca Barbato 07master:e767c9e8f2ea: flv: Validate and reject unsupported codecs
[21:00] <cone-531> ffmpeg 03Michael Niedermayer 07master:735ab7c5e04e: Merge commit 'e767c9e8f2eaa116b61b8b6881b401b54bd320f5'
[21:00] Action: Compn wonders what nevcairiel uses ffmpeg for
[21:00] <wm4> oh right, the guy who wrote this might be the debian maintainer of ffmpeg
[21:00] <poste9> I just want to understand ^^ but also to speed up
[21:00] <wm4> Compn: are you deaf and blind?
[21:00] <nevcairiel> wm4: he uses win2k, my code wouldnt run on his system
[21:01] <wm4> D:
[21:01] <Compn> hey i upgraded
[21:01] <Compn> no more win2k :(
[21:01] <wm4> XP now?
[21:01] <Compn> xp and vista
[21:01] <himangi> can we close ticker #3153.. I see its already implemented
[21:01] <wm4> himangi: I think anyone registered can close tickets
[21:02] <himangi> wm4: I am new here, can I confirm with someone before closing it?
[21:02] <Compn> it can be closed , best if you reference the git commit that fixed the issue ...
[21:02] <Compn> assuming its fixed
[21:02] Action: Compn did not look
[21:03] <wm4> if it's wrong or met with disagreement, someone will reopen it
[21:03] <himangi> wm4: cool
[21:07] <Compn> haha
[21:08] <wm4> fast, but confusing reply
[21:19] <BtbN> wbs, got it working now, no idea if it's the correct approach though. The problem was that the mov muxer didn't check for an extradata update before writing the moov tag.
[21:19] <BtbN> https://github.com/BtbN/FFmpeg/commit/c1385d8ef95b1e85eeb3c20bc9da5c285b7d9634 that's what i changed
[21:33] <BtbN> I'll just send these patches to the ML for review.
[21:33] <wbs> BtbN: if you can wait a while I'll see if there's a different place in movenc
[21:34] <BtbN> ok
[21:34] <wbs> BtbN: because iirc there is already some sort of copying of late extradata somewhere, so I'd want to see why that doesn't trigger in this case
[21:37] <wbs> BtbN: ah, I see. look for "copy extradata if it exists" in movenc.c - there's already code that does exactly this - but this never is called because the extradata is copied in only right before writing the moov, not before actually writing the packets
[21:37] <BtbN> yep
[21:38] <wbs> BtbN: so if you move back the stuff in dashenc.c where you had it before, before my suggestion, it should hopefully work without the hack to movenc
[21:38] <wbs> (you probably don't need the memcmp of extradata, just check if cur size is 0 and the external is > 0, then you can trigger a copy/reinit)
[21:40] <BtbN> wbs, hm? There'd still be no instruction calling into the mov muxer between av_write_frame(os->ctx, NULL); and the extradata copy.
[21:41] <BtbN> Moving it back to the beginning of the loop doesn't help.
[21:41] <wbs> BtbN: ah, yes. no, instead of having it in dash_flush, move it into dash_write_packet
[21:42] <wbs> so it should be propagated immediately on the first packet write when there actually is extradata (practically, it should be available immediately on the first packet)
[21:42] <BtbN> So just iterate over all streams, and invoke the copy function for it?
[21:42] <wbs> no, just copy for the current stream
[21:43] <BtbN> oh, it's already called for one specific stream
[21:43] <wbs> I'd expect that dash_write_packet is called at least once per stream per segment, otherwise you can't really assume that someone initialized extradata for the stream either
[21:51] <cone-531> ffmpeg 03Peter Cordes 07master:44d347726475: vf_showinfo: minimum widths for some early fields
[21:52] <BtbN> wbs, seems to work, at least the DecoderInfo is present
[21:52] <wbs> BtbN: great. and also a reason for skipping memcmp; movenc also just checks the size so it won't actually propagate an updated one with the same size anyway
[22:13] <cone-531> ffmpeg 03Peter Cordes 07master:4da3a14f3ca4: pixelutils: indent comments in pixdesc.h to be clearer
[22:13] <cone-531> ffmpeg 03Peter Cordes 07master:9e5687adf20a: pixelutils: Comment on (lack of) sad_8x8_sse2
[22:13] <cone-531> ffmpeg 03Peter Cordes 07master:46235f0e0296: libx264: update AQ options help for x264 build 144
[22:51] <cone-531> ffmpeg 03Michael Niedermayer 07master:4ad7b8f64a24: avformat/movenc: mark write_colr as experimental
[22:54] <ubitux> michaelni: how long do i have before the release for writing the release note?
[23:01] <michaelni> ubitux, iam waiting for the avformat_flush() patch, i might release after that
[23:01] <ubitux> ok
[23:01] <ubitux> i'll try to get it done for saturday
[23:02] <ubitux> just need to find an hour or so
[23:06] <nevcairiel> no aacenc patch today? :(
[23:07] <wm4> michaelni: I think this was the latest version of the flush patch https://ffmpeg.org/pipermail/ffmpeg-devel/2014-September/163635.html
[23:07] <wm4> trivial but had docs bikeshedding
[23:51] <cone-531> ffmpeg 03Carl Eugen Hoyos 07master:63c9b6e431b8: lavf/mxf: Support ProRes decoding.
[23:56] <cone-531> ffmpeg 03Carl Eugen Hoyos 07master:83808ee7e9e8: Allow AVFoundation compilation on OS X 10.7.
[00:00] --- Thu Mar  5 2015


More information about the Ffmpeg-devel-irc mailing list