[Ffmpeg-devel-irc] ffmpeg-devel.log.20120921

[burek]

[00:25] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * ra6ec1e49f6 10ffmpeg/libavformat/matroskadec.c: 
[00:25] <CIA-56> ffmpeg: matroskadec: matroska_parse_rm_audio: remove unused parameter
[00:25] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:07] <CIA-56> ffmpeg: 03Andrey Utkin 07master * rd33908a5fc 10ffmpeg/libavcodec/avpacket.c: 
[01:07] <CIA-56> ffmpeg: Move av_dup_packet() copy logic to separate procedure
[01:07] <CIA-56> ffmpeg: Done to simplify next commit introducing av_copy_packet()
[01:07] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:07] <CIA-56> ffmpeg: 03Andrey Utkin 07master * rcc07a7925e 10ffmpeg/libavcodec/ (avcodec.h avpacket.c version.h): 
[01:07] <CIA-56> ffmpeg: Add av_copy_packet()
[01:07] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[01:15] <burek> is this some kind of regression: http://paste.ubuntu.com/1217763/
[01:15] <burek> the same command used to work on previous versions of ffmpeg
[01:16] <burek> just ignore me
[01:29] <burek> can I suggest this page ( http://ffmpeg.org/trac/ffmpeg/report ) to be edited so that the first link "Open+New Tickets" is sorted by creation time?
[01:29] <burek> in descending order
[01:30] <burek> so that newest tickets show first
[02:01] <michaelni> burek, you can change the wiki if such different ordering is more usefull to most people ...
[02:01] <burek> well, that page is not editable
[02:01] <burek> by me, I mean
[02:01] <michaelni> hmm, why??
[02:02] <burek> well those are predefined reports
[02:02] <burek> defined by admin i guess
[02:03] <michaelni> hmm, you can change the way its sorted and then look at the url and copy that into the link 
[02:03] <michaelni> or am i missing sometjing ?
[02:03] <burek> yes, just add &desc=1&order=time
[02:03] <burek> but I don't have an option to change that page
[02:03] <burek> :)
[02:04] <burek> it's not a standard wiki/ticket page
[02:04] <burek> it's kinda special, I guess
[02:05] <michaelni> ahh, i misunderstood you
[02:05] <burek> "View tickets" button at the upper-right part of the page
[02:05] <michaelni> you mean the report #1 
[02:05] <burek> yes
[02:05] <burek> the page that lists all the predefined reports
[02:09] <michaelni> iam a bit unsure if it should be changed, what is the advantage to just change the link ?
[02:10] <burek> well, this way, when you click on the first report it gets all the items that it promised, but look at the order
[02:10] <burek> no order at all
[02:10] <burek> then you click the "created" column and it sorts everything in ascending order, giving you the oldest tickets first
[02:10] <burek> so you have to click again, to get the latest tickets first
[02:11] <burek> I mean, it's not that much important, I just didn't see any other link to a page that will list the latest tickets easily
[02:12] <michaelni> I can add a new report that does exactly that
[02:13] <michaelni> iam just a bit unsure about changing an existing one (its sorted by priority)
[02:13] <michaelni> being sorted by priority developers can see the important issues first
[02:13] <burek> oh, I see, ok then
[02:13] <burek> i didn't realize that
[02:14] <burek> if it's not a big issue for you, can you add a new one that says something like "The newest/latest tickets"
[02:14] <burek> or something like that
[02:19] <michaelni> now if i knew how to make it sort the other direction ;)
[02:20] <burek> :D
[02:20] <burek> just remove desc=1
[02:21] <burek> it will sort in asc ord
[02:26] <michaelni> now if there was a desc=1 in the tracquerry :)
[02:26] <burek> oh, you are generating links using that filter/columns form? :)
[02:27] <burek> or how? :)
[02:27] <michaelni> i added desc=0
[02:27] <michaelni> seems to work
[02:27] <burek> cool :)
[02:28] <michaelni> theres something called a tracquerry (https://ffmpeg.org/trac/ffmpeg/wiki/TracQuery#QueryLanguage) with which the newer reports are specified
[02:28] <michaelni> the older use SQL
[02:29] <michaelni> but using SQL means that its not user editable like changing filters and stuff IIRC
[02:30] <burek> oh, I see, so you just needed to remove "desc" param at all
[02:30] <burek> "order" by default sorts in ascending order
[02:31] <burek> at least that's how the final links look like
[02:31] <burek> but using sql.. geez.. :)
[02:31] <michaelni> the internal thing i edited in a textbox contained no desc, and i donno why
[02:32] <burek> I see now new report :)
[02:32] <burek> it's very descriptive :D
[04:01] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r3dc24600ba 10ffmpeg/libavfilter/vsrc_testsrc.c: 
[04:01] <CIA-56> ffmpeg: vsrc_testsrc: Fix case without a specified duration
[04:01] <CIA-56> ffmpeg: Fixes regression since 98f753ec5198d4f74ba9f68cc85badf8a6257547
[04:01] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:34] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * r6ab1166c77 10ffmpeg/libavformat/nutdec.c: 
[06:34] <CIA-56> ffmpeg: nutdec: fix type of intermediate variable for ts calculation.
[06:34] <CIA-56> ffmpeg: Prior to this it could theoretically become negative and read
[06:34] <CIA-56> ffmpeg: out of array.
[06:34] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[06:34] <CIA-56> ffmpeg: 03Michael Niedermayer 07master * rfb96ac469a 10ffmpeg/libavformat/nutdec.c: 
[06:34] <CIA-56> ffmpeg: nutdec: fix mixup of nb_streams and timebase_count
[06:34] <CIA-56> ffmpeg: Fixes out of array read, should fix ffprove fate failure
[06:34] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[09:31] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r0f40105aaa 10ffmpeg/ffprobe.c: 
[09:31] <CIA-56> ffmpeg: ffprobe: print codec_long_name only when available
[09:31] <CIA-56> ffmpeg: Avoid possible segmentation fault if codec_long_name is defined as
[09:31] <CIA-56> ffmpeg: NULL (for example if --enable-small).
[10:35] <saste> Daemon404, MSVC doesn't like unicode, do you know why?
[12:45] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r31aa5b1a78 10ffmpeg/libavfilter/vf_hue.c: 
[12:45] <CIA-56> ffmpeg: lavfi/hue: fix crash when resetting the saturation expression
[12:45] <CIA-56> ffmpeg: Previously when saturation was reinited, the old value was freed when
[12:45] <CIA-56> ffmpeg: setting options, and freed again in PARSE_EXPRESSION().
[12:45] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r3aa1b53256 10ffmpeg/libavfilter/vf_hue.c: 
[12:45] <CIA-56> ffmpeg: lavfi/hue: set to NULL freed pointers
[12:45] <CIA-56> ffmpeg: Possibly safer.
[14:00] <CIA-56> ffmpeg: 03Stefano Sabatini 07master * r9543cd593e 10ffmpeg/ffprobe.c: 
[14:00] <CIA-56> ffmpeg: ffprobe: print format long_name only if non-NULL
[14:00] <CIA-56> ffmpeg: Fix possible crash when long_name is not defined (for example if
[14:00] <CIA-56> ffmpeg: --enable-small).
[14:48] <CIA-56> ffmpeg: 03Mans Rullgard 07master * ra34a609fc1 10ffmpeg/libavcodec/ (motion_est.c motion_est_template.c): (log message trimmed)
[14:48] <CIA-56> ffmpeg: motion_est: fix use of inline on extern functions
[14:48] <CIA-56> ffmpeg: Inline functions declared without extern do not provide an external
[14:48] <CIA-56> ffmpeg: definition in standard C99. This code only works because most
[14:48] <CIA-56> ffmpeg: compilers do not implement the inline semantics correctly. With a
[14:48] <CIA-56> ffmpeg: stricter compiler, linking fails with unresolved references to these
[14:48] <CIA-56> ffmpeg: functions.
[14:48] <CIA-56> ffmpeg: 03Janne Grunau 07master * r6746cd7f61 10ffmpeg/libavformat/utils.c: 
[14:48] <CIA-56> ffmpeg: lavf probe: prevent codec probe with no data at all seen
[14:48] <CIA-56> ffmpeg: This occurs with fuzzed mpeg-ts files. set_codec_from_probe_data() is
[14:49] <CIA-56> ffmpeg: called with a zeroed AVProbeData since no packet made through for
[14:49] <CIA-56> ffmpeg: specific stream.
[14:49] <CIA-56> ffmpeg: 03Mans Rullgard 07master * r8995d34972 10ffmpeg/libavutil/arm/asm.S: 
[14:49] <CIA-56> ffmpeg: ARM: use 2-operand syntax for ADD Rd, PC in Apple PIC code
[14:50] <CIA-56> ffmpeg:  lavf: zero data/size of the packet passed to read_packet().
[14:50] <CIA-56> ffmpeg: 03Mans Rullgard 07master * rcdb7db5acd 10ffmpeg/libavutil/arm/asm.S: 
[14:50] <CIA-56> ffmpeg: ARM: align PIC offset pools to 4 bytes
[14:50] <CIA-56> ffmpeg: When building Thumb2 code, the end of a function, where the PIC
[14:50] <CIA-56> ffmpeg: offsets are placed, need not be aligned. Although the values
[14:50] <CIA-56> ffmpeg: are only accessed with instructions allowing unaligned addresses,
[14:50] <CIA-56> ffmpeg: keeping them aligned is preferable.
[14:50] <CIA-56> ffmpeg: Signed-off-by: Mans Rullgard <mans at mansr.com>
[14:50] <CIA-56> ffmpeg: 03Mans Rullgard 07master * r0122118ec3 10ffmpeg/configure: 
[14:50] <CIA-56> ffmpeg: configure: update tms470 detection for latest version
[14:50] <CIA-56> (9 lines omitted)
[15:18] <CIA-56> ffmpeg: 03Clément BSsch 07master * r7bc70930dd 10ffmpeg/ (5 files in 2 dirs): lavf/movenc: add F4V flavor.
[16:34] <CIA-56> ffmpeg: 03Matthieu Bouron 07master * r41f7e0649f 10ffmpeg/libavformat/ (mxf.c mxf.h mxfenc.c): 
[16:34] <CIA-56> ffmpeg: mxf: export ff_mxf_get_samples_per_frame to lavf
[16:34] <CIA-56> ffmpeg: Reviewed-by: Tomas Härdin <tomas.hardin at codemill.se>
[16:34] <CIA-56> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[16:50] <Daemon404> [04:35] <@saste> Daemon404, MSVC doesn't like unicode, do you know why? <-- i have a feeling it might be a 'windows thing', but i'd have to test with mingw32 (natively) too
[16:51] <nevcairiel> the other fails are far more interesting
[17:21] <Daemon404> nevcairiel, think i figured it out.
[17:21] <Daemon404> @ saste too
[17:22] <saste> Daemon404, cool
[17:22] <Daemon404> just need to test
[17:22] <Daemon404> fix is 2 lines
[17:22] <Daemon404> unicode is sitll an issue though
[17:28] <CIA-56> ffmpeg: 03Clément BSsch 07master * r405ee405c9 10ffmpeg/ffmpeg.c: ffmpeg: complete mov/mp4 based timebase hack with "f4v" format.
[17:31] <Daemon404> ubitux, that is really ugly
[17:31] <Daemon404> that whole if()
[17:31] <Daemon404> putting such logic in ffmpeg.c is bad...
[17:31] <ubitux> feel free to propose something else
[17:31] <ubitux> i'm just completing with the muxer i added
[17:32] <Daemon404> unset the AVFMT_VARIABLE_FPS flag properly?
[17:32] <Daemon404> in lavf
[17:34] <ubitux> no idea
[17:34] <ubitux> michaelni: changing the QMAT_SHIFT won't cause problems when the optimizations will be re-enabled?
[17:35] <ubitux> i wonder how much more optimizations diego disabled while moving stuff around...
[17:49] <michaelni> ubitux, shouldnt cause problems unless the optimized code is bugy too but we will know for sure once its reenabled
[17:50] <ubitux> well with all the changed tests i was wondering... since at some point with the optim it had the same output afaict
[17:50] <mateo`> a quick fix to reenable optimizations would be to call encode_init_x86 right after common_init_x86 in ff_dct_common_init
[17:51] <mateo`> sounds weird since other arch only call common_init_$(ARCH) ...
[17:54] <Daemon404> saste, damn nvm i didnt
[17:55] <saste> Daemon404, could be that windows uses UTF16 by default?
[17:55] <Daemon404> i dont know
[17:55] <Daemon404> but it turns out i didnt fixt he pos issue
[17:57] <Daemon404> i think its somewhere in value_string
[18:06] <michaelni> ubitux, are you sure the C and MMX quantization routines and DCTs had the same output ?
[18:08] <michaelni> i this does not seem likely to me ...
[18:16] <Daemon404> saste, i dont get it
[18:16] <Daemon404> none of the printing functiond are to blame for it
[18:16] <Daemon404> i dumped pkg->pos before it got printed and it is still wrong
[18:16] <Daemon404> :/
[18:17] <Daemon404> but then wouldnt all the lavf tests be failing
[18:31] <saste> Daemon404, the pos thing... i believe it is related to the unicode thing
[18:31] <Daemon404> why do you think?
[18:31] <saste> the unicode string has not the expected size
[18:32] <saste> maybe we're using some strlen thing which is taking into account the character length, rather than the byte-lenght
[18:32] <saste> but i'm just doing a wild-guess
[18:32] <saste> anyway i bet the pos/unicode issues are somehow related
[18:34] <Daemon404> saste, but why wouldnt lavfi fail
[18:34] <Daemon404> er
[18:34] <Daemon404> lavf
[18:34] <saste> that's a good question
[18:35] <saste> but maybe the tests don't check for unicode strings, or for packet positions
[18:35] <Daemon404> possibly
[18:35] <Daemon404> what are we using in way of unicode functions?
[18:35] <Daemon404> cause windows does use utf16
[18:40] <Daemon404> startcode = ff_crc04C11DB7_update(0, (uint8_t*) &startcode, 8);
[18:40] <Daemon404> wow really?
[18:40] <Daemon404> worst function name ever
[18:41] <saste> aha yes, can't remember why the funny name
[20:08] <Daemon404> saste, im not having an easy time figuring out where we are doing unicode wrong
[21:51] <ubitux> 22:59 < michaelni> i meant ogg_save() + new stream added + ogg_restore() -> nstreams != nb_streams // actually, it seems the number of streams is ok
[21:52] <ubitux> because the context nstreams is kept after ogg_restore
[21:52] <ubitux> and so i need to do this
[21:53] <ubitux> if i don't, in the flac+theora.ogg sample won't work (checking state or headers)
[21:53] <ubitux> http://pastie.org/4775273  here is what it looks like with the flac+theora.ogg sample
[21:55] <ubitux> mmh though no streams are added within the save/restore context...
[21:56] <ubitux> well, i'll add this check then.
[21:59] <ubitux> the new stream code is not supposed to happen at all within the ogg save/restore?
[22:13] <ubitux> it really seems the new stream code is never reached in between ogg context save/restore
[22:13] <ubitux> but i don't understand why
[22:14] <ubitux> since it's just reading pages, as usual
[22:19] <ubitux> no one? :(
[22:38] <michaelni> ubitux, i dont understand your question
[22:39] <ubitux> michaelni: i've worded it differently in the commit annotation
[22:39] <ubitux> i'll explain differently if you still don't understand
[22:39] <teratorn> hi, I'm not getting any help on #ffmpeg - is there any easy way to do multiple "cuts" on the ffmpeg command line. basically I want to specify -ss <foo> and -t <bar> multiple times to skip various portions of the video
[22:39] <ubitux> teratorn: maybe the segmenter format
[22:40] <teratorn> ubitux: where do I go to find out about that?
[22:40] <ubitux> http://ffmpeg.org/ffmpeg.html#segment_002c-stream_005fsegment_002c-ssegment
[22:40] <teratorn> thanks I'll give it a look
[22:42] <teratorn> I guess cutting the video to multiple temp files, then combining them is good enough for me right now :/
[22:43] <michaelni> ubitux, maybe data_packets_seen() prevents it from happening on valid files
[22:43] <teratorn> but I think it would be cool to allow -ss and -t to be given several times. currently it allows it, but only respects one of them
[22:44] <Daemon404> you can probably use the select filter
[22:45] <ubitux> Daemon404: i wonder if you wouldn't mess the ts by doing this
[22:45] <Daemon404> i dunno
[22:45] <Daemon404> i use avisynth
[22:45] <ubitux> and you need to keep it in sync with the audio too
[22:45] <Daemon404> :P
[22:45] <ubitux> :))
[22:45] <Daemon404> tho now... vapoursynth
[22:45] <ubitux> Daemon404: did you see the arguments from nicolas about the frame exact thing?
[22:46] <ubitux> basically he was defending ts over frame id
[22:48] <Daemon404> ubitux, thats fine i can ignore
[22:48] <Daemon404> ffms2 works fine
[22:55] <ubitux> that's not a question about ignoring :p
[22:55] <ubitux> michaelni: mmh ok i think i get it now
[22:56] <Daemon404> ubitux, seems like its about ignoring to me!
[22:57] <ubitux> too bad, the question is pretty interested
[22:57] <ubitux> interesting*
[22:57] <ubitux> and you were likely interested into frame exacteness stuff
[23:05] <burek> is this a bug or just an unimplemented feature: http://sprunge.us/YSAY
[23:05] <burek> ffmpeg -i /home/manizzle/Downloads/Lec1_noon.wmv /home/manizzle/work/ee306/Lec1_noon.avi
[23:05] <burek> that was used as a cmd line
[23:34] <ubitux> burek: i'd say the feature is not implemented, but the stream is still trying to get decode (in case the next packet doesn't require that feature or anything)
[23:35] <ubitux> maybe in that particular case we could raise an EOF or something to make the decode stop
[23:35] <ubitux> the operation not permitted are because there is just a return -1 after the error, nothing to care about
[23:35] <burek> I see
[23:36] <burek> I advised the guy to try to extract the audio with some other tool and to merge it back as wav or something usable
[23:36] <ubitux> that message "If you want to help, upload a sample of this file to ftp://upload.ffmpeg.org/MPlayer/incoming/ and contact the ffmpeg-devel mailing list." looks fine to honor as well :)
[23:37] <ubitux> maybe we should change it to redirect to the bug tracker so it's not forgotten on the ml
[23:37] <burek> :)
[23:38] <burek> I think bug tracker would be a better message
[23:38] <burek> I mean the guy already asked for the bug tracker url :) (#ffmpeg)
[23:41] <ubitux> it's two times in the topic
[23:41] <ubitux> but well, yeah.
[23:45] <burek> who reads the docs, until no other options are available :D
[23:46] <Daemon404> i assume many users do this:
[23:46] <Daemon404> 1) somethign doesnt work
[23:46] <Daemon404> 2) man ffmpeg
[23:46] <Daemon404> 3) OMG OMG OMG
[23:46] <Daemon404> 4) give up
[23:46] <ubitux> you can't say the irc topic is documentation :p
[23:47] <Daemon404> ubitux, irc topics are even less read than docs
[23:47] <ubitux> :(
[23:47] <burek> exactly what i wanted to write :D
[00:00] --- Sat Sep 22 2012