[Ffmpeg-devel-irc] ffmpeg-devel.log.20111130

burek burek021 at gmail.com
Thu Dec 1 02:05:02 CET 2011


[00:23] <CIA-41> ffmpeg: 03Jean First 07master * r23b4f355bb 10ffmpeg/ffmpeg.c: 
[00:23] <CIA-41> ffmpeg: ffmpeg.c use b:v for video bitrates using a target
[00:23] <CIA-41> ffmpeg: it's more consistent to use -b:v instead of just -b
[00:23] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:08] <ubitux> michaelni: no bump needed with the audio preload?
[01:25] <michaelni> ubitux, thx ill fix it
[02:40] <CIA-41> ffmpeg: 03Aaron Colwell 07master * r39a7a5b8ab 10ffmpeg/libavcodec/pthread.c: 
[02:40] <CIA-41> ffmpeg: pthread: don't increment index on zero-sized packets.
[02:40] <CIA-41> ffmpeg: The next call to decode() will update from an invalid index, which will
[02:40] <CIA-41> ffmpeg: either lead to a memcpy() where dest==src (2 threads), or lead to a
[02:40] <CIA-41> ffmpeg: crash (>2 threads).
[02:40] <CIA-41> ffmpeg: Signed-off-by: Ronald S. Bultje <rsbultje at gmail.com>
[02:40] <CIA-41> ffmpeg: 03Alex Converse 07master * rfdab793a46 10ffmpeg/avconv.c: 
[02:40] <CIA-41> ffmpeg: avconv: Handle audio sync for non-S16 sample formats.
[02:40] <CIA-41> ffmpeg: Also fix reporting of the number of samples added/dropped.
[02:40] <CIA-41> ffmpeg: 03Aneesh Dogra 07master * r464ccb0144 10ffmpeg/libavcodec/indeo3.c: 
[02:40] <CIA-41> ffmpeg: indeo3: check per-plane data buffer against input buffer bounds.
[02:40] <CIA-41> ffmpeg: Fixes : http://bugzilla.libav.org/show_bug.cgi?id=102
[02:40] <CIA-41> ffmpeg: Signed-off-by: Alex Converse <alex.converse at gmail.com>
[02:40] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rc40a35f8a7 10ffmpeg/: (log message trimmed)
[02:40] <CIA-41> ffmpeg: Merge remote-tracking branch 'qatar/master'
[02:40] <CIA-41> ffmpeg: * qatar/master:
[02:40] <CIA-41> ffmpeg:  indeo3: check per-plane data buffer against input buffer bounds.
[02:40] <CIA-41> ffmpeg:  avconv: Handle audio sync for non-S16 sample formats.
[02:40] <CIA-41> ffmpeg:  pthread: don't increment index on zero-sized packets.
[02:40] <CIA-41> ffmpeg: Conflicts:
[03:54] <durandal_1707> i found lost patch for escape130 video codec
[04:01] <michaelni> durandal_1707, where ? :)
[04:01] <durandal_1707> on web
[04:01] <durandal_1707> i downloaded warzone rar
[04:01] <durandal_1707> and i found dec130.dll with debugging symbols
[04:02] <durandal_1707> from quick look in ht, i noticed dec300.cpp
[04:03] <durandal_1707> patch, original work by Eli Friedman just appears to have some color issue, but reading wiki entry i got that apparently fixed
[04:06] <durandal_1707> the wiki entry on multimedia.cx for that codec is also write by Eli
[04:10] <michaelni> durandal_1707, so you have improved elis patch ? do i understand that correctly
[04:13] <durandal_1707> https://github.com/richardpl/ffmpeg/tree/escape130
[04:14] <durandal_1707> it seems duration is incorrectly calculated
[04:18] <michaelni> you mean the duration printed by ffmpeg  ?
[04:18] <durandal_1707> yea
[04:18] <durandal_1707> and mplayer
[04:19] <michaelni> likely some issue in rpl.c
[04:19] <michaelni> is that happening with escape124 too ?
[04:21] <durandal_1707> apparently does
[04:22] <durandal_1707> PYRAMID sample ~9 vs 15sec
[04:22] <michaelni> its quite simply because rpl.c does not set duration
[04:23] <michaelni> so libavformat guesses from bitrates probably and they are maybe wrong
[04:24] <michaelni> number_of_chunks*frames_per_chunk_fps or something like this could be used for duration
[04:24] <michaelni> number_of_chunks*frames_per_chunk * fps
[04:25] <michaelni> note AVStream.duration is in the timebase set by av_set_pts_info
[04:27] <michaelni> also you shoudl set the author of the initial commit to eli if its his code
[04:37] <durandal_1707> do i need to change date too?
[04:39] <michaelni> i dont care about the date, you should improve the commit messages a bit though too when you are already at it
[04:39] <michaelni> "escape124:" prefixes at the least
[04:40] <michaelni> escape130 is what i meant :)
[04:52] <durandal_1707> those tables should probably be generated at decoder initialization
[05:00] <michaelni> could be done but they are not very large, sign_table is just 256 bytes
[05:32] <CIA-41> ffmpeg: 03Joseph Wecker 07master * rd31c68a7ed 10ffmpeg/libavformat/flvenc.c: 
[05:32] <CIA-41> ffmpeg: removes spurious warning
[05:32] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[05:32] <CIA-41> ffmpeg: 03Joseph Wecker 07master * rb55dd10d62 10ffmpeg/libavformat/flvdec.c: 
[05:32] <CIA-41> ffmpeg: Fix typeo in some rarely used memoization
[05:32] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:02] <nathan0n5ire> could some people give me some feedback for this ffmpeg icon I created? http://f.imgtmp.com/Btx89.png
[06:13] <michaelni> nathan0n5ire, its a bad time, i think most devels are sleeping :) i need to go to bed too now
[06:13] <michaelni> but it looks nice
[06:19] <michaelni> nathan0n5ire, one thing i was thinking of was logos for different occasions like google does show on their main page
[06:19] <michaelni> but if this is for google code in, then talk with burek first, he is mentor of these logo tasks
[06:22] <nathan0n5ire> I realized I actually like ffmpeg'
[06:23] <nathan0n5ire> s current icon better then the one I made...
[06:28] <michaelni> nathan0n5ire, you could for example try to modify the current one by adding snow to give it a winterly feel ...
[06:28] <michaelni> but i really need to go to bed now, good luck with your logo task ! :)
[06:29] <michaelni> i mean the current one on ffmpeg.org
[06:29] <buzz_> late night eh michael.. sleep well :)
[06:29] Action: buzz_ just got up 
[06:29] <michaelni> :)
[09:55] <ubitux> michaelni: "AV_PERM_PRESERVE is missing"  as rej_perms right? so if the input pad requires the buffer to be preserved, the buffer will be copied?
[09:57] <ubitux> the images are not modified by the filter (but filtered), does it really apply here?
[09:57] <ubitux> and last question: what use case does it affect?
[09:58] <ubitux> mmh the last N frame feed to the filter will actually output a different frame, so i guess this is what it means by modifying
[14:52] <michaelni> ubitux, it seems preserve is not implemented the way i remembered it, the current implementation would rather need rej_perm=AV_PERM_REUSE2
[15:25] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rdae7ff0416 10ffmpeg/libavcodec/wmalosslessdec.c: 
[15:25] <CIA-41> ffmpeg: wmall: remove tabs that leaked in from a merge and seem to have
[15:25] <CIA-41> ffmpeg: managed to silently bypass the push checks.
[15:25] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:25] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r4710993800 10ffmpeg/libavformat/movenc.c: 
[15:25] <CIA-41> ffmpeg: movenc: memset(0) new cluster elements after realloc()
[15:25] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[15:25] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rdd32a305bd 10ffmpeg/libavcodec/wmalosslessdec.c: 
[15:25] <CIA-41> ffmpeg: wmall: fix return type of dump_int_buffer()
[15:25] <CIA-41> ffmpeg: Found-by: Dominique Leuenberger
[15:25] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:38] <vega13> hi
[16:39] <vega13> is AVPacket a POD or non-POD ?
[16:39] <vega13> : )
[16:41] <vega13> I'm trying to copy it with memcpy but keeps segmentation fault..
[16:42] <vega13> I'm trying to copy it into a custom packet, before i would just memcpy packet.data into the char* buffer
[16:43] <vega13> but since avcodec_decode_video2 now only takes full avpackets i have to copy the whole avpacket
[16:46] <michaelni> vega13, you probably searchj for av_dup_packet() and i dunno what POD is supposed to mean
[16:47] <vega13> okay thanks
[16:48] <vega13> so with av_dup_packet i can copy an avpacket into an char buffer?
[16:49] <michaelni> no
[16:50] <michaelni> but with av_dup_packet() you can keep it while demuxing more packets
[16:50] <michaelni> if you want to put it in a char buffer then
[16:51] <michaelni> you have to store its data&size at minimum
[16:51] <michaelni> you also need dts/pts one way or another
[16:52] <michaelni> the decoders wont care but you need them to display at the right time
[16:52] <michaelni> and flags may be needed by 1 decoder IIRC
[16:54] <michaelni> if you provide avpackets with pts/dts to decoders though then they can reorder the timestamps for you and provide some guess at what timestamp would be correct for the decoded frame
[16:57] <vega13> They are already inside my packet, I get my packet from av_read_frame then I want to do memcpy(triPkg->encDataBuffer(),&packet, packet.size+1);
[16:58] <vega13> the old code worked: memcpy(triPkg->encDataBuffer(), packet.data, packet.size)
[17:01] <michaelni> av_dup_packet(); memcpy &packet, sizeof(AVPacket); and NO freeing of the packet before you are done should work
[17:01] <michaelni> or you can just keep using your old code
[17:02] <michaelni> (old will fail with 1 decoder but it did not work before either)
[17:03] <vega13> thanks !
[17:05] <vega13> gh then they can reorder the timestamps for you and provide some guess at what timestamp would be correct for the decoded frame
[17:07] <vega13> sorry
[17:07] <vega13> accidently pasted inside here
[17:20] <ubitux> michaelni: any idea why this mp3 stream is somewhat misdetected?: http://mp3.live.tv-radio.com/2783/radio-pirate-iledefrance/radio-pirate-iledefrance.mp3 
[17:21] <ubitux> it looks like it "forgets" to send headers
[17:21] <ubitux> i'm not sure the server is at fault
[17:22] <ubitux> i get various "Format mp3 detected only with low score of 25, misdetection possible!" and header missing messages
[17:33] <michaelni> ubitux, i see header missing but no low scores 
[17:33] <michaelni> do you have a dumped file with low score to reproduce ?
[17:33] <ubitux> sure just a sec
[17:33] <michaelni> header missing is probably due to mp3 frames being cut at random byte offset
[17:35] <ubitux> wget 'http://blackhole.ubitux.fr/rp.mp3'
[17:35] <michaelni> confirmed ill look at it
[17:36] <ubitux> thank you
[17:46] <michaelni> ubitux, do you have a actually misdetected one or just low score ?
[17:46] <michaelni> it has 479 mp3 frames and 500 is the threshold for returning a higher score
[17:47] <ubitux> no just low scores + warn/errors; decode and playback are fine
[17:47] <ubitux> oh, so a bigger sample should do the trick?
[17:47] <michaelni> yes
[17:48] <michaelni> the problem with detecting mp3 quickly is that everything contains mp3 and so a truncated mpeg-ps file could be misdetected as mp3 if the threshokd is too low
[17:50] <ubitux> how do you make a distinction between them?
[17:50] <ubitux> if 500 audio frame oly ’ mp3?
[17:50] <ubitux> only*
[17:50] <michaelni> comparing scores frorm the probe functions
[18:20] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rb51eaf3b8c 10ffmpeg/libavformat/mp3dec.c: 
[18:20] <CIA-41> ffmpeg: mp3probe: Detect mp3 stronger with just 200 frames, this should speed up detection
[18:20] <CIA-41> ffmpeg: on mp3 streams.
[18:20] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:20] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r8d5078c10b 10ffmpeg/libavformat/ac3dec.c: 
[18:20] <CIA-41> ffmpeg: ac3probe: Change threshold from 500 to 200 to keep in sync with mp3.
[18:20] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:22] <ubitux> heh, thanks
[18:54] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r8f025f2f60 10ffmpeg/libavfilter/defaults.c: 
[18:54] <CIA-41> ffmpeg: avfilter: align allocated images to 32 for AVX and IFF
[18:54] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[18:54] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rfd3e75d2d4 10ffmpeg/libavcodec/utils.c: 
[18:54] <CIA-41> ffmpeg: lavc: horizontal align for IFF
[18:54] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r957a593cd9 10ffmpeg/libavformat/flvdec.c: 
[21:30] <CIA-41> ffmpeg: flvdemux: export flags for nellymoser through side data.
[21:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r11828b8885 10ffmpeg/libavcodec/nellymoserdec.c: 
[21:30] <CIA-41> ffmpeg: nellymoserdec: Use sidedata to support mid stream sample rate changes.
[21:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[21:30] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * r77af2672b4 10ffmpeg/libavcodec/utils.c: 
[21:30] <CIA-41> ffmpeg: lavc: Fix side data handling for audio decoders.
[21:30] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[23:22] <CIA-41> ffmpeg: 03Michael Niedermayer 07master * rc863d3751f 10ffmpeg/libavformat/movenc.c: 
[23:22] <CIA-41> ffmpeg: movenc: replace cluster memset by zeroing only the needed field.
[23:22] <CIA-41> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Thu Dec  1 2011


More information about the Ffmpeg-devel-irc mailing list