[Ffmpeg-devel-irc] ffmpeg-devel.log.20130919

[burek]

[10:14] <cone-685> ffmpeg.git 03Stefano Sabatini 07master:6230a95659da: doc/muxers: itemize tee examples
[10:14] <cone-685> ffmpeg.git 03Stefano Sabatini 07master:f0606a28deca: ffprobe: add -read_intervals option
[10:50] <cone-685> ffmpeg.git 03Martin Storsjö 07master:0f310a6f333b: rmdec: Validate the fps value
[10:50] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:f54d751f3337: Merge commit '0f310a6f333b016d336674d086045e8473fdf918'
[11:19] <BoR0> hi. I have a question. I am writing a function to extract out a frame at nth position (regardless if position is timestamp or frame number). so I managed it to be a frame number by calling av_read_frame() N times and thus it will catch the Nth frame and extract a screenshot out of that. but this is slow, because it receives 0...N frames to get the Nth frame. so I thought to use avformat_seek_file and then call av_read_frame(), but this only gives me th
[11:19] <BoR0> e first frame always. what can I do?
[11:21] <saste> BoR0, nothing, seek is based on timestamps, not on the number of frames
[11:21] <BoR0> how can I grab a frame after seek?
[11:21] <saste> and is not accurate, since it relies on key frames position
[11:21] <BoR0> I don't need accuracy
[11:21] <BoR0> I want to build a thumbnails maker
[11:22] <BoR0> may as well be random one
[11:22] <BoR0> I just need speed
[11:22] <durandal_1707> please if you are not working on ffmpeg libs/tools itself do not use this channel
[11:23] <saste> BoR0, => #ffmpeg
[11:23] <BoR0> durandal_1707, sorry. I am aware of that rule and I am aware that I violate it, but in my experience this channel is always more active than #ffmpeg
[11:23] <BoR0> most answers I get are from here
[11:30] <durandal_1707> everything is explained in this channel topic
[11:36] <michaelni> ubitux, mateo`_  did any of you had time to look at the realloc / list freeing issue in mov ?
[11:36] <durandal_1707> there is such issue?
[11:36] <durandal_1707> michaelni: have you time to look at qp patch?
[11:47] <cone-685> ffmpeg.git 03Alexandra Khirnova 07master:5626f994f273: avformat: Use av_reallocp() where suitable
[11:47] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:20dfab33f5ab: Merge commit '5626f994f273af80fb100d4743b963304de9e05c'
[11:47] <cone-685> ffmpeg.git 03Luca Barbato 07master:fbd54827b1be: rtmp: Unbreak handle_notify
[11:54] <cone-685> ffmpeg.git 03Justin Ruggles 07master:9ae53c5860e1: lavc: add ff_u8_to_s8() for safe type punning of uint8_t to int8_t
[11:54] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:ac2fc63ac6f9: Merge commit '9ae53c5860e17d087642d27b00dade071970a2e7'
[12:43] <cone-685> ffmpeg.git 03Justin Ruggles 07master:c4bfa098072b: Add a WebP decoder
[12:43] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:b26ccf218d93: Merge commit 'c4bfa098072ba338d83555d6e2199f7e1e64ffff'
[12:50] <cone-685> ffmpeg.git 03Alex Converse 07master:5cdefc02079a: aacdec: Add support for Error Resilience syntax.
[12:50] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:257f11820bdf: Merge commit '5cdefc02079a5b899bdec09e15e2bb099cb01734'
[12:53] <durandal_1707> why you disabled it? just remove it
[13:00] <michaelni> durandal_1707, its called a merge not a "and various unrelated changes" 
[13:01] <durandal_1707> so you plan to remove it in non-merge?
[13:02] <cone-685> ffmpeg.git 03Alex Converse 07master:1914e6f010b3: aacdec: Add support for LD (Low Delay) AAC
[13:02] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:32ea39f56d90: Merge commit '1914e6f010b3320025c7b692aaea51d9b9a992a8'
[13:02] <michaelni> durandal_1707, i wanted to post a patch doing so but i might forget as ive a few things i wanted to do
[13:03] <durandal_1707> ok
[13:08] <cone-685> ffmpeg.git 03Luca Barbato 07master:ae0f316a4778: rtmp: Unbreak handle_notify
[13:08] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:374a2f456211: Merge commit 'ae0f316a4778e473d407cffa8392595ba0735153'
[13:08] <durandal_1707> if filter returns frames of different size than outlink, what it should do?
[13:15] <saste> durandal_1707, when does it happen?
[13:15] <saste> in theory it should never happen
[13:15] <durandal_1707> say you want to crop different size, what should you do instead?
[13:15] <saste> durandal_1707, there is no way at the moment
[13:16] <saste> missing dynamic reconfiguration
[13:16] <durandal_1707> that sucks
[13:16] <durandal_1707> it SUUUUUUCCCCCCKSSSSS!!!!!!!
[13:16] <saste> i'll probably work on it soon
[13:34] <cone-685> ffmpeg.git 03Martin Storsjö 07master:5372cda67109: rv10: Validate the dimensions set from the container
[13:34] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:b4c88a4c4cea: Merge commit '5372cda67109848d22146289e401669266217e80'
[13:39] <cone-685> ffmpeg.git 03Martin Storsjö 07master:bcbe4f3ceb6e: idroqdec: Make sure a video stream has been allocated before returning packets
[13:39] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:52dc3600a901: Merge commit 'bcbe4f3ceb6ee0210d3a401963518906c8b9b230'
[13:50] <cone-685> ffmpeg.git 03Martin Storsjö 07master:fdbd924b84e8: rv34: Fix a memory leak on errors
[13:50] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:93648ee47b8f: Merge remote-tracking branch 'qatar/master'
[14:06] <durandal_1707> ah that means merges are finally over and i can rm mp=qp
[14:21] <durandal_1707> i wrote nice filter but will not share it with you because of libmpcodecs
[14:22] <wm4_> "How did we know it was even worth porting if we don't know properly what it does?"
[14:22] <wm4_> haaaaaa
[14:22] <wm4_> let's find out what the hell this mplayer scrap did
[14:22] <durandal_1707> be careful what you are saying, creator of qp reside in this channel
[14:25] <saste> durandal_1707, what filter?
[14:25] <durandal_1707> TOP SECRET
[14:26] <saste> too easy, you can't prove its existence
[14:26] <durandal_1707> i will share it once dynamic reconfiguration is there
[14:27] <Compn> is it -vf kiss ?
[14:27] <saste> ok :)
[14:27] <Compn> it makes ffmpeg devels kiss libav devels
[14:27] <durandal_1707> mainly because it calls sws_scale, while that could be done by scale filter
[14:28] <wm4_> something with alpha?
[14:32] <Compn> where is that daemon guy
[14:32] <Compn> :P
[14:36] <durandal_1707> if i create 100 frames from 1 frame, how can i avoid stupid: [ffmpeg/out] 100 buffers queued in out, something may be wrong.
[14:36] <saste> yes that message it's helpless and frustrating
[14:37] <saste> "something may be wrong, figure it out, ahahah"
[14:38] <durandal_1707> it looks like all 100 frames are held in memory...
[14:39] <saste> durandal_1707, what command?
[14:40] <durandal_1707> filter that in filter_frame takes one frame and calls ff_filter_frame 100 time and allocates 100 frames with ff_get_video_buffer
[14:41] <saste> durandal_1707, are you queing them?
[14:41] <saste> i suppose we should implement some queing policy
[14:41] <saste> right now we have a fixed queue size, and now way to handle how the buffers are stored
[14:42] <saste> i think right now it always drop the new incoming frames when the queue is full
[14:44] <durandal_1707> but idea is to get one frame and returnes multiple one without need to queue them internally
[14:49] <durandal_1707> perhaps i could do it with custom request_frame?
[15:24] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:8be56e464bde: avformat/smacker: Check if bufs has been allocated before use
[15:24] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:ad2a08388c82: avformat/concat: revert one hunk of 20dfab33f5ab61fc51de65192ed239f8b62782a2
[18:22] <cone-685> ffmpeg.git 03Paul B Mahol 07master:390c35a13b3e: lavfi/mp: remove mp=qp
[18:58] <cone-685> ffmpeg.git 03Paul B Mahol 07master:1f5d1342ea25: doc: Add missing hashes and dates to APIChanges
[19:08] <durandal_1707> saste: what needs to be done for dynamic link reconfiguration?
[19:08] <saste> durandal_1707, changing all the filters
[19:09] <saste> durandal_1707, seriously, we need to add fields to each pad, and rework the configuration code
[19:09] <durandal_1707> what fields?
[19:19] <saste> durandal_1707, we need to mark the fields which are supposed to change, in each pad
[19:20] <saste> so no, probably we won't need to change all filters
[19:20] <saste> i need to reread the one-year-old rfc and nicolas' reply
[19:53] <cone-685> ffmpeg.git 03Paul B Mahol 07master:0ad5ef674b65: avfilter/vf_decimate: fix crash with gray8
[19:53] <cone-685> ffmpeg.git 03Paul B Mahol 07master:20101f4a8326: avfilter/vf_decimate: add gray16 and yuv440p
[19:57] <durandal_1707> :(
[20:02] <durandal_1707> hey! when vp9 is coming?
[20:03] <GoaLitiuM> coming to where?
[20:05] <durandal_1707> *ffvp9
[20:06] <ubitux> durandal_1707: "soon"
[20:06] <ubitux> waiting for BBB to get his internet back i guess
[20:07] <GoaLitiuM> so how does it compare to h264 in terms of quality?
[20:08] <ubitux> "50% better"©®"
[22:22] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:63139f62f359: avformat/bmv: remove unused variable
[22:22] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:9c8aeacf82a8: avutil: add av_get_colorspace_name()
[22:25] <ubitux> michaelni: no i didn't have time to look at the mov thing
[22:26] <ubitux> a bit busy currently
[22:26] <ubitux> vobsub is really insane :(
[22:48] <ruggles> michaelni: i'm seeing a weird swscale (i think) issue i need to fix and i'm not sure where to look. when converting from rgb24 to argb, pixel 0,0 is getting 0 for alpha instead of 255.
[22:49] <durandal_1707> i doubt such bug is in ffmpeg
[22:49] <ruggles> to reproduce: ffmpeg -loglevel debug -i frame00.ppm -pix_fmt argb -c:v rawvideo -f rawvideo - |hexdump -C |less
[22:50] <ruggles> durandal_1707: try it
[22:50] Action: durandal_1707 really hates swscale
[22:51] <durandal_1707> that is using unscaled variant?
[22:51] <ruggles> yeah
[22:51] <ruggles> i tried to track it down a little but the closest i got was that it's maybe related to ALT32_CORR but i have no idea how that works.
[22:52] <durandal_1707> the unscaled thing that does rgb24 to argb should set it to 255
[22:52] <ruggles> indeed it should
[22:52] <durandal_1707> but perhaps this one use asm?
[22:52] <ruggles> i tried that too. disabled asm. same result.
[23:19] <durandal_1707> i just tried with two ppms manually created and they all have 255 for alpha
[23:21] <durandal_1707> even when encoded to rawvideo
[23:21] <durandal_1707> which should not be needad as png/pam can be used
[23:22] <durandal_1707> i don't see nowhere any sws line which would say what codepath it used
[23:22] <JEEB> fun fact: -f rawvideo is pretty much -f rawstream
[23:22] <JEEB> first found someone dumping subtitle tracks with it
[23:22] <JEEB> (from mpeg-ts)
[23:23] <JEEB> then I used it when I needed to dump the vorbis stream from a file
[23:23] <JEEB> (without a container)
[23:27] <durandal_1707> btw version i use is 2.5.100 
[23:29] <durandal_1707> also you do not need hexdump.....
[23:32] <ruggles> durandal_1707: ok, i'll rebuild HEAD and try again...
[23:32] <ruggles> but png is actually rgba not argb
[23:33] <ruggles> i think...
[23:34] <ruggles> JEEB: and -f u8 works for audio
[23:36] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:1225b67fc9d7: avutil/frame: suppress "comparison of unsigned expression < 0 is always false" warning
[23:36] <cone-685> ffmpeg.git 03Michael Niedermayer 07master:13eff473adb0: swscale/swscale_unscaled: fix alpha values for rgb/bgr -> RGB32_1 / BGR32_1
[23:36] <michaelni> ruggles, should be fixed
[23:36] <ruggles> michaelni: awesome. thanks.
[23:39] <michaelni> np
[23:41] <durandal_1707> michaelni: i'm interested how you reproduced this?
[23:45] <michaelni> by using the provided testcase
[23:45] <michaelni> with lena.pnm IIRC
[23:46] <michaelni> -i tests/lena.pnm -pix_fmt argb -c:v rawvideo -f rawvideo - |hexdump -C |less
[23:51] <durandal_1707> ah yes, i'm blind
[23:55] <ubitux> i'm really happy with that vobsub sample
[23:55] <ubitux> it really summarizes well the insanity inside
[23:58] <durandal_1707> why it needs to be supported?
[23:58] <ubitux> because it seems supported by some other demuxers
[23:59] <ubitux> but i really have trouble figuring out exactly how it's supposed to be handled
[23:59] <ubitux> i can think of various heuristic that will work
[23:59] <ubitux> but i'm not sure if i'm not missing something in obvious
[23:59] <ubitux> -in
[00:00] --- Fri Sep 20 2013