[Ffmpeg-devel-irc] ffmpeg-devel.log.20140503

burek burek021 at gmail.com
Sun May 4 02:05:02 CEST 2014

[01:14] <cone-788> ffmpeg.git 03Lukasz Marek 07master:eb434f712d98: doc: fix message names in opengl doc
[03:14] <cone-788> ffmpeg.git 03mplayer developers 07master:80dd0aff3346: libavfilter/libmpcodecs: sync existing filters with mplayer HEAD (ebcacb8b3ca91ef90acb93785b62fd8c5e5dae41)
[03:42] <cone-788> ffmpeg.git 03Lukasz Marek 07master:8b02e1c21d4d: doc/APIchanges: add hashes and dates
[03:42] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:719e72c971eb: doc/APIchanges: more dates and hashes
[07:00] <cone-140> ffmpeg.git 03Michael Niedermayer 07master:09725c1d2bed: avfilter/vf_showinfo: also calculate and show mean and standard deviation
[15:20] <cone-197> ffmpeg.git 03Carl Eugen Hoyos 07master:4116cbeaf50e: Fix oga muxer standalone compilation.
[15:20] <cone-197> ffmpeg.git 03Carl Eugen Hoyos 07master:a75349d0c833: Fix standalone compilation of dsd planar decoders.
[15:42] <wm4> michaelni: could it be that you accidentally broke yadif?
[15:42] <wm4> michaelni: see #3605
[15:49] <JEEB> wm4, I could replicate something similar
[15:49] <JEEB> on win32
[15:49] <JEEB> so YADIF seems borked in some way
[15:50] <wm4> probably due to my request to handle 1-frame video :/
[16:03] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:7ef01a787814: avcodec/h264: mark recovery_cnt==0 frames as keyframes
[16:22] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:681e008d06d2: vfilter/vf_yadif: fix request_frame after 0f9f24c9
[16:29] <nevcairiel> i had a dumb hack to support 1 frames differently, it seemed simpler at the time, maybe is hould've submitted it :p
[16:30] <wm4> michaelni: thanks, appears to work, this time I actually tested it
[17:24] <cone-197> ffmpeg.git 03Carl Eugen Hoyos 07master:4aea0afbbb5e: Remove two duplicated named asm constraints.
[17:56] <cone-197> ffmpeg.git 03Matt Oliver 07master:3c3e02b8d197: x86/cavdsp: prevent named constraints appearing twice.
[18:07] <ubitux> michaelni: the err recognition patches for h263 are for the recent spacex video thing?
[18:08] <michaelni> yes
[18:08] <ubitux> does it help?
[18:08] <ubitux> (in a significant way)
[18:14] <michaelni> the patches allow you to decode more of the frames by ignoring errors and disabling error concealment, it looks worse but there is more information that can be used for analysis
[18:23] <ubitux> http://ubitux.fr/pub/pics/_edgedetect-paint.jpg meh.
[18:23] Action: ubitux should really do some useful stuff again
[18:23] <ubitux> (edgedetect with colormix mode)
[18:31] <ubitux> btw, we don't have the css anymore for the local documentation?
[18:31] <michaelni> ?
[18:32] <ubitux> dunno, i opened doc/ffmpeg-filters.html
[18:32] <ubitux> and got no css
[18:33] <ubitux> ah, got it, it's probably because i build outside the root dir
[19:09] <cone-197> ffmpeg.git 03Clément BSsch 07master:3f3c3318be41: avfilter/edgedetect: make processing functions context free.
[19:09] <cone-197> ffmpeg.git 03Clément BSsch 07master:b17e98ded0e2: avfilter/edgedetect: add a colormix mode.
[19:09] <cone-197> ffmpeg.git 03Clément BSsch 07master:365c79bd4e2e: avfilter/edgedetect: reindent after previous commit.
[19:11] <cone-197> ffmpeg.git 03Clément BSsch 07master:d53e87a83716: avfilter/edgedetect: update copyright.
[20:13] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:cc861581735b: avcodec/ffv1dec: use av_malloc_array()
[20:13] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:77df9de78c9a: avcdoec/snowenc: use av_mallocz_array()
[20:13] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:00948ff876ee: avfilter/audio: use av_mallocz_array()
[21:18] <Daemon404> i wonder if that will ever be pushed
[21:19] <ubitux> yeah it's a bit like if the guy was maintaining his own fork diff-based on our trac
[21:20] <Daemon404> looks like it is sitll losing to LAME and FAAC
[21:20] <Daemon404> wait thats v7
[21:21] Action: Daemon404 is remidned of how he used to manually generate diff sets on top of stuff and keep it in a dir
[21:27] <BBB> Daemon404: quilt?
[21:27] <Daemon404> BBB, at the time i wasn taware of quilt
[21:27] <Daemon404> patch/diff
[21:28] <Daemon404> i used to have a fork of xchat that was a set of 10-20 patches rebased on top of it manually (CVS days)
[21:28] <nevcairiel> i dont really remember how I did it before Git, but now I just keep a Git repo
[21:28] <Daemon404> since its developer refused to accept them, snce it would negate his shareware scheme
[21:29] <ubitux> crackme.diff
[21:29] <Daemon404> lol
[22:02] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:56cc17848526: avcodec/h263dec: fix "incompatible pointer type" warning
[22:10] <ubitux>         for (end = nptr + (h - 2) * ns, nptr += ns, optr += os, top = 0;
[22:10] <ubitux>              nptr < end; nptr += ns - w, optr += os - w, top ^= 1) {
[22:10] <ubitux> grumbl.
[22:10] <Daemon404> wat
[22:10] <Daemon404> inb4 mplayer code
[22:11] <ubitux> well, that's indeed from mplayer
[22:11] <Daemon404> called it.
[22:13] <JEEB> lol
[22:16] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:bb605ef8ab60: avcodec/xbmdec: fix pointer type warnings
[22:16] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:5918c5a4ccd7: avcodec/kgv1dec: add casts to silence "assignment from incompatible pointer type"
[22:16] <ubitux> i'm actually delighted to find 3 additionnal increment at the end of the loop that could be part of the for
[22:31] <ubitux> i see a lot of (relatively old) recordings with a black line on top with doesn't do the whole width
[22:31] <ubitux> but length is somehow random
[22:31] <ubitux> like half or T
[22:31] <ubitux> any idea where that comes from?
[22:31] <ubitux> s/with/but/
[22:32] <ubitux> (second one)
[22:32] <ubitux> i mean, even this here: http://samples.ffmpeg.org/avi/GoneNutty.avi
[22:33] <ubitux> also here: https://trac.ffmpeg.org/raw-attachment/ticket/2595/ffmpeg-bug-2595.mpg
[22:35] <ubitux> when it's the whole line it's probably because of bad cropping, but here i can't tell
[22:35] <ubitux> it doesn't look random vbi "noise" either
[22:36] <JEEB> seems like a thing with analog captures by your description
[22:36] <JEEB> I've seen it in Japanese TV too
[22:36] <JEEB> where they use an analog input
[22:36] <JEEB> not sure why that line gets cut
[22:36] <JEEB> I think it can be either on top or bottom
[22:38] <Compn> ubitux : i've seen that half line thing
[22:39] <Compn> ubitux : is it because of 'odd field first' setting or something
[22:39] <Compn> ubitux : thinking about making an autocrop filter for that half line ? :P
[22:39] <J_Darnley> Me too.  It is on just about every TV show I have on DVD.
[22:40] <ubitux> so exactly wtf is going on here?
[22:41] <J_Darnley> I thought it was something to do with CRTs starting in the centre
[22:50] <iive> actually the half line is both on top and the bottom
[22:52] <iive> but the top is rarely seen as vbi (teletext) resides there.
[23:14] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:a520d8fd3580: avcodec/snow_dwt: fix "incompatible pointer type" warnings
[23:14] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:6b79ef00fdac: avfilter/vf_deshake: dont pass the wrong context into the SAD functions
[23:14] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:0a5cc20d4aac: avcodec/jpeglsdec: add cast to silence pointer type warning
[23:22] <cone-197> ffmpeg.git 03Luca Barbato 07master:c6b42e854939: flv: K&R formatting cosmetics
[23:22] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:3ea5beab8f73: Merge commit 'c6b42e85493915a3c3891345f57422f9031a0b4f'
[23:22] <kurosu> yeah, analogic is the reason iirc
[23:22] <kurosu> NTSC is originaly 525 lines, but you have two fields
[23:23] <kurosu> guess what, that makes 262.5 lines per field
[23:27] <cone-197> ffmpeg.git 03Luca Barbato 07master:a90d92ce2aee: aac: K&R formatting cosmetics
[23:27] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:0c06ca7a8ab4: Merge commit 'a90d92ce2aeea8eaae5c616958d20e000e20b964'
[23:37] <cone-197> ffmpeg.git 03Luca Barbato 07master:925c37874b61: vp9: Read the frame size as unsigned
[23:37] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:039630631e96: Merge commit '925c37874b617816b6f460c6f229c18b46548b46'
[23:44] <cone-197> ffmpeg.git 03Luca Barbato 07master:3a177a9cca92: swscale: Fix an undefined behaviour
[23:44] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:1c4110be9038: Merge commit '3a177a9cca924e097265b32f9282814f6b653e08'
[23:44] <BBB> so how do we merge these commits?
[23:44] <BBB> I mean, I looked at the vp9: one, and there's nothing to merge
[23:44] <BBB> that code doesn't exist in our code
[23:45] <BBB> also, the bug was already fixed by ubitux' extensive fuzzing
[23:45] <ubitux> the merge commit says "See: c72f587353d9f88c7fbbbf6ef78b1d9ba3d38fd7"
[23:45] <ubitux> nothing to do :)
[23:45] <BBB> so why did we merge it?
[23:46] <ubitux> meta
[23:46] <BBB> how obscure
[23:46] <BBB> ok
[23:46] <BBB> anyway
[23:46] <nevcairiel> because the entire history is merged, or you end up with a silly state
[23:46] <ubitux> it's useful when diffing which commits are in or not
[23:46] <ubitux> since the merge keep the original hash
[23:53] <cone-197> ffmpeg.git 03Luca Barbato 07master:de69aedf9935: mathematics: K&R formatting cosmetics
[23:53] <cone-197> ffmpeg.git 03Michael Niedermayer 07master:bc4b424dfa6e: Merge commit 'de69aedf9935631b7f78e8b8da6e460422a9bc5f'
[23:54] <ubitux> english question, is "[0,10] is the range of values from 0 to 10 inclusive" correct? ("inclusive" in particular)
[23:54] <ubitux> iirc there is a word for this but... :(
[23:55] <ubitux> btw, Daemon404 will be happy, i'm writing a "how to write a filter with lavfi"
[23:59] <BBB> yes
[23:59] <BBB> also if it was exclusive (10) in math context you'd use [0,10> I believe
[23:59] <BBB> that's double notation from what I remember
[23:59] <BBB> long time ago
[00:00] --- Sun May  4 2014

More information about the Ffmpeg-devel-irc mailing list