[Ffmpeg-devel-irc] ffmpeg-devel.log.20130923

burek burek021 at gmail.com
Tue Sep 24 02:05:02 CEST 2013


[00:07] <michaelni_> durandal_1707, if nicolas is busy and doesnt react even after reasonable time and you are willing to maintain the code and are sure te code is ok, id say yes
[00:12] <cone-712> ffmpeg.git 03Timothy Gu 07master:ca10d66719f5: doc/filters: add [] labels for the graph of the opening paragraph example
[00:13] <cone-712> ffmpeg.git 03Xi Wang 07release/0.9:4b7036c1d9d1: rtmp: fix multiple broken overflow checks
[00:13] <cone-712> ffmpeg.git 03Xi Wang 07release/0.9:8c0261d6859d: rtmp: fix buffer overflows in ff_amf_tag_contents()
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:03e2e95beb1a: huffyuvdec: Check init_vlc() return codes.
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:21dd8f5baa43: huffyuvdec: Skip len==0 cases
[00:13] <cone-712> ffmpeg.git 03Diego Biurrun 07release/0.9:32a81629dae3: configure: Make warnings from -Wreturn-type fatal errors
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:c681cf34c4a6: apichanges: fix 2 wrong hashes
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:50b260133823: doc/APIchanges: List merge commit hashes and version numbers
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:de2eb6dab101: rmdec: flush audio packet on seeking
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:e4ad9e9377b5: update all trac links to use the trac subdomain
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:edb4aad93dc2: avcodec/flashsv: check diff_start/height
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:4490b3db23bd: avcodec/rpza: Perform pointer advance and checks before using the pointers
[00:13] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:a4b4be7493bd: jpeg2000: check log2_cblk dimensions
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:14f140f36942: matroska_read_seek: Fix used streams for subtitle index compensation
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:40d4e82e2ecd: avcodec/pngdsp: fix (un)signed type in end comparission
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:a62852bae0ee: avcodec/dsputil: fix signedness in sizeof() comparissions
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:a6388942f590: avcodec/ffv1enc: update buffer check for 16bps
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:5198623b0e97: update for 0.9.3
[00:14] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.9:dae552fbca8f: cmdutils: update year to 2013
[00:22] <durandal_1707> michaelni_: well if it fixes 2 bugs and do not introduce any new then i'm gonna apply it
[00:46] <michaelni_> durandal_1707, go ahead if you tested that it does fix 2 bugs
[01:17] <cone-712> ffmpeg.git 03Timothy Gu 07fatal: ambiguous argument 'refs/tags/n0.9.3': unknown revision or path not in the working tree.
[01:17] <cone-712> Use '--' to separate paths from revisions
[01:18] <cone-712> refs/tags/n0.9.3:HEAD: doc/filters: add [] labels for the graph of the opening paragraph example
[01:22] <durandal_1707> michaelni_: your fix for escape124 oom introduced infintive loop
[01:23] <durandal_1707> i really do not see why uint64_t is needed there
[01:23] <durandal_1707> because that check failt if negative signed is compared with unsigned
[01:26] <durandal_1707> well it may be because of overflow somewhere too
[01:27] <durandal_1707> it becomes: -8 >= 1
[01:29] <durandal_1707> and that check is stupid anyway
[01:29] <durandal_1707> there is unchecked reader...
[01:31] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.6:6ad4a116a206: avcodec/rpza: Perform pointer advance and checks before using the pointers
[01:31] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.6:f9225c2884bb: matroska_read_seek: Fix used streams for subtitle index compensation
[01:31] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.6:b51c4df35b52: avcodec/dsputil: fix signedness in sizeof() comparissions
[01:31] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.6:c7ab9de727e5: avcodec/ffv1enc: update buffer check for 16bps
[01:31] <cone-712> ffmpeg.git 03Michael Niedermayer 07release/0.6:243396f06dc4: update for 0.6.7
[01:31] <durandal_1707> and checked one, already used by this codec
[01:31] <durandal_1707> so i will just kill can_safely_read
[01:36] <michaelni_> uint64 looks bad indeed
[01:36] <durandal_1707> when oom happened?
[01:45] <durandal_1707> michaelni_: patch on ml
[02:23] <cone-712> ffmpeg.git 03Paul B Mahol 07master:e494f44c051d: avcodec/escape124: fix infinite loop
[02:41] <cone-712> ffmpeg.git 03Michael Bradshaw 07master:3cfd4df8e722: libopenjpeg: init and close properly; don't config during encode
[02:50] <cone-712> ffmpeg.git 03Michael Bradshaw 07fatal: ambiguous argument 'refs/tags/n0.6.7': unknown revision or path not in the working tree.
[02:50] <cone-712> Use '--' to separate paths from revisions
[02:50] <cone-712> refs/tags/n0.6.7:HEAD: libopenjpeg: init and close properly; don't config during encode
[02:51] <BBB> michaelni_: comparisons
[02:52] <BBB> not comparissions
[02:52] <Compn> BBB : write spell hook into commit script :P
[02:59] <Wheelman_> ˆˆˆ€
[02:59] <Wheelman_> ˆˆ
[02:59] Last message repeated 1 time(s).
[02:59] <Wheelman_> ˆˆˆ„
[02:59] <Wheelman_> ˆˆˆ€
[02:59] <Wheelman_> ˆˆ
[03:00] Last message repeated 1 time(s).
[03:00] <Wheelman_> ˆˆˆ„
[11:36] <cone-114> ffmpeg.git 03Luca Barbato 07master:32a414f316c7: rtmp: Support AMF_DATA_TYPE_MIXEDARRAY
[11:36] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:f3a02f4bc1d0: Merge commit '32a414f316c7f0eea877370e3f9d9f25afbf5da2'
[11:40] <cone-114> ffmpeg.git 03Luca Barbato 07master:e40a0e822801: rtmp: Refactor get_packet
[11:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:eee8cb42d009: Merge commit 'e40a0e822801d2485e4e555909d7a82713fa86a5'
[11:45] <cone-114> ffmpeg.git 03Luca Barbato 07master:584047389044: rtmp: Rewrite embedded flv handling
[11:45] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:c70e647a1587: Merge commit '5840473890440dbe0bd2cce530ebb3d93e187ae6'
[11:57] <cone-114> ffmpeg.git 03Luca Barbato 07master:ba77757a7efa: rtmp: Replace a magic number with a macro
[11:57] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:eaa20afa7608: Merge commit 'ba77757a7efafe38924b6b423acfca8c1e1c62d6'
[11:58] <ubitux> saste: 
[11:58] <ubitux> ffprobe.c:1684:21: warning: start may be used uninitialized in this function [-Wmaybe-uninitialized] end = start + interval->end;
[12:01] <cone-114> ffmpeg.git 03Luca Barbato 07master:c951e4b442df: flv: Make onTextData parsing robust
[12:01] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:09e06cd90fa2: Merge commit 'c951e4b442df9dde6051634a85359eb809bf2142'
[12:10] <cone-114> ffmpeg.git 03Martin Storsjö 07master:5055035670bd: tls: Do not abort on non-fatal TLS alerts with gnutls
[12:10] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:03e9506aae31: Merge commit '5055035670bd1a1eaca64bd3bc71fb07de9df2c3'
[12:17] <ubitux> any opinion on injecting original streams metadata into lavfi for filters init?
[12:18] <ubitux> (and making it possible to add/drop some with the filters)
[12:18] <Compn> sanitized first ?
[12:18] <ubitux> sanitized?
[12:18] <ubitux> what for?
[12:18] <Compn> check inputs are valid
[12:18] <ubitux> the idea is to allow filters to use those data
[12:18] <Compn> ;drop_tables ? no clue :D
[12:18] <Compn> ehe
[12:19] <ubitux> stop your monkeyness Compn :(
[12:19] <ubitux> typical case: -vf rotate=auto
[12:19] <ubitux> mov sets some meta into the video streams for rotation
[12:19] <Compn> ooo, finally fixing peopls' horizontal videos automatically
[12:19] <Compn> :)
[12:19] <Compn> do it do it
[12:19] <ubitux> i'd like to but well...
[12:20] <ubitux> on a side note, would be nice if rotate had fast path for common rotations
[12:22] <Compn> like flip does in some codecs, by reading the lines backward ?
[12:22] <cone-114> ffmpeg.git 03Martin Storsjö 07master:187023f6b2b4: movenc: Set mov->mode earlier in mov_write_header
[12:22] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:b189c699b98c: Merge commit '187023f6b2b43966acb2449a379b4ededdc4b22e'
[12:22] <ubitux> like transpose does
[12:34] <cone-114> ffmpeg.git 03Martin Storsjö 07master:a9553bbb336a: movenc: Set all implicit flags immediately in mov_write_header
[12:34] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:d4574b760834: Merge commit 'a9553bbb336a7186c62b3b14a0499c1b6ba61ebd'
[12:50] <saste> ubitux, will have a look at it later, thx
[12:50] <saste> michaelni_, can remove_extra be used to remove *only* the global extradata in the header?
[12:51] <saste> also, where is the global extradata placed in the first packet, or is set by the codec?
[12:51] <saste> in the latter case, i suppose the bitstream filter is of no use
[13:15] <cone-114> ffmpeg.git 03Martin Storsjö 07master:f1eac2b8a037: movenc: Use keyframes as default fragmentation point in ismv
[13:15] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:341dc5905351: Merge commit 'f1eac2b8a0370b908cd691086d11f51342054730'
[13:19] <mateo`> some merge commits to review :)
[13:28] <cone-114> ffmpeg.git 03Martin Storsjö 07master:91be1103fd1f: wnv1: Make sure the input packet is large enough
[13:28] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:621845feccd1: Merge commit '91be1103fd1f79d381edf268c32f4166b6c3b6d8'
[13:29] <michaelni_> mateo`, yes :)
[13:29] <michaelni_> and thanks
[13:32] <michaelni_> saste, what do you mean by " *only* the global extradata in the header" ?
[13:32] <saste> ubitux, it's a false positive, since has_start => start was set
[13:33] <saste> michaelni_, i tried the remove_extra, but it seems to remove extradata in all key-packets
[13:33] <michaelni_> yes thats the default behavior
[13:33] <saste> i wonder if there is a way to remove only extradata from the global header, which I don't know where it is set
[13:34] <michaelni_> it also can limit its actions to non keyframes
[13:34] <saste> maybe only in the first packet?
[13:34] <saste> i can tweak the filter (and while at it extend the ffmpeg syntax to accept the optional parameter), but I don't know how I can remove global extradata
[13:34] <saste> i'm using libx264 for encoding
[13:35] <michaelni_> i dont think the remove_extra bsf does anything with AVCodecContext.extradata
[13:35] <michaelni_> why do you want to remove the global headers extradata ?
[13:36] <saste> michaelni_, i'm using the tee to create mpegts output and a mov archive at the same time
[13:37] <saste> in order to do that, i'm using +global_header, and then I'm adding extradata in the mpegts packets by using the dump_extra fitler
[13:37] <saste> the problem with this approach is that it is adding overhead, i suppose because i'm sending the global extradata in the stream header
[13:37] <saste> so ideally i would like to remove just the global header extradata
[13:38] <saste> not sure i can really do that, i'm a n00b concerning bitstream/extradata issues
[13:41] <cone-114> ffmpeg.git 03Martin Storsjö 07master:ede508443e4b: vc1dec: Fix leaks in ff_vc1_decode_init_alloc_tables on errors
[13:42] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:eed40b29570b: Merge commit 'ede508443e4bf57dc1e019fac81bf6244b88fbd3'
[13:43] <michaelni_> mpegts doesnt do anything with AVCodecContext.extradata
[13:43] <nevcairiel> the parser extracts it from the video stream
[13:44] <iive> what does it extract from there?
[13:44] <nevcairiel> for h264 it would be sps/pps
[13:53] <cone-114> ffmpeg.git 03Martin Storsjö 07master:b772b0e28eba: vc1dec: Undo mpegvideo initialization if unable to allocate tables
[13:53] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:8b85d9bfb6f1: Merge commit 'b772b0e28eba6abf76d86ee8c6e459a86642db5a'
[14:03] <cone-114> ffmpeg.git 03Martin Storsjö 07master:5e25fdbfe016: vc1dec: Make sure last_picture is initialized in vc1_decode_skip_blocks
[14:03] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:79fbd7a5c1d6: Merge commit '5e25fdbfe01635cfc650ac4adc27d434b2df0d64'
[14:09] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:bb7bf1759514: vc1dec: Don't decode slices when the latest slice header failed to decode
[14:09] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:c74fa39b93ed: Merge commit 'bb7bf175951475235ee7d0504198c4a30cf5d5dd'
[14:10] <saste> ubitux, how do you get that warning?
[14:12] <ubitux> gcc version 4.8.1 20130725 (prerelease) (GCC) 
[14:17] <cone-114> ffmpeg.git 03Stefano Sabatini 07master:7bac6e5cf81b: ffprobe: fix uninitialized variable warning
[14:18] <Daemon404> ubitux, should update
[14:21] <ubitux> saste: thx
[14:23] <cone-114> ffmpeg.git 03Martin Storsjö 07master:ce67f442be0f: lavf: Avoid setting avg_frame_rate if delta_dts is negative
[14:23] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:ef7e8efc6bef: Merge commit 'ce67f442be0f6c4a8794272873852e256b5b8ee4'
[14:34] <cone-114> ffmpeg.git 03Martin Storsjö 07master:f7e616959aff: electronicarts: Check packet sizes before reading
[14:34] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:eeeb3c8fa3a1: Merge commit 'f7e616959aff8706edccdae763c24c897c449f6f'
[14:38] <xlinkz0> should ffmpeg extend segment cuts to the next i-frame after the requested duration was reached?
[14:38] <xlinkz0> or is the fact that it does look for an i-frame to end the segment a bug?
[14:39] <nevcairiel> at the end there really is no special need for a I frame
[14:39] <nevcairiel> as long as there is one at the start
[14:39] <xlinkz0> great, then it is a bug?
[14:39] <nevcairiel> unless you want to start a second segment right after
[14:39] <xlinkz0> or must I use a special condition?
[14:39] <xlinkz0> i don't
[14:39] <nevcairiel> then you should include everything up to the I frame, but not the I frame
[14:40] <xlinkz0> but can't I just include half the gop or just a few frames?
[14:41] <cone-114> ffmpeg.git 03Martin Storsjö 07master:23d6ba2c4454: vc1dec: Check the error handling flags on slice/field header decode errors
[14:41] <nevcairiel> if its closed gop that may not work, depends on the format
[14:41] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:50138a9235b0: Merge commit '23d6ba2c44540eb39c6880e4cbd527d297d264a9'
[14:41] <xlinkz0> h264 codec, mp4 container
[14:41] <xlinkz0> can I force ffmpeg tool to do this?
[14:42] <xlinkz0> to only include frames up until the required duration even though it hasn't reached the end of the gop?
[14:45] <cone-114> ffmpeg.git 03Martin Storsjö 07master:f2b00a5372bf: movenc: Return a proper error code for invalid combinations
[14:45] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:8c766228c180: Merge commit 'f2b00a5372bfa1d82a8ab99cb8eafac8a9473494'
[14:53] <cone-114> ffmpeg.git 03Martin Storsjö 07master:b6c7bd53f4ba: movenchint: Clear size and len if the realloc failed
[14:53] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:552945efe62d: Merge remote-tracking branch 'qatar/master'
[15:03] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:a51f3b53fe7f: avcodec/vc1dec: Fix return code on malloc() failure from ff_vc1_decode_init_alloc_tables()
[15:04] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:356031348c3b: avcodec/vc1dec: fix propagating error codes from various functions
[15:11] <cone-114> ffmpeg.git 03Nicolas George 07master:53319d5c9328: lavfi: add an API to synchronize multiple video inputs.
[15:11] <cone-114> ffmpeg.git 03Nicolas George 07master:231e50157cc9: lavfi/dualinput: reimplement on top of framesync.
[15:11] <cone-114> ffmpeg.git 03Michael Niedermayer 07master:3f3867ca2752: Merge remote-tracking branch 'cigaes/master'
[15:25] <xlinkz0> nevcairiel: are you a developer for ffmpeg?
[15:25] <xlinkz0> i added printf("####         Comparing %"PRId64" with %"PRId64"\n",ist->pts, of->recording_time + start_time);
[15:25] <xlinkz0> to line 1417 in ffmpeg.c
[15:26] <xlinkz0> i get this output : http://codepad.org/NkxlqBaf
[15:26] <xlinkz0> should ist->pts ever be negative in do_streamcopy() ?
[15:36] <xlinkz0> i also printed every input_stream->pts at the begining of transcode_step : http://codepad.org/BrgNPrxV
[15:36] <xlinkz0> can someone please tell me if this is normal?
[15:42] <xlinkz0> ubitux: ?
[15:43] <ubitux> check -debug_ts
[15:43] <ubitux> dunno about your problem, enotime
[15:47] <cone-114> ffmpeg.git 03Ronald S. Bultje 07master:34b429d5ba7b: vp8: fix PPC assembly and bilinear C code to work if src_stride != dst_stride.
[15:48] <xlinkz0> this is the debug_ts output : http://codepad.org/lWomZ9mu
[15:48] <xlinkz0> i still don't understand, why would input_stream->pts be negative?
[16:04] <xlinkz0> cmon it's monday
[16:04] <xlinkz0> why is this channel dead :(
[16:05] <xlinkz0> michaelni__: got 5 minutes?
[16:05] <Daemon404> 1) it's 10am in EST, 7am in PST, and people are still at work in GMT
[16:05] <Daemon404> 2) it's monday.
[16:05] <xlinkz0> when's a good time then? :D
[16:06] <xlinkz0> isn't ffmpeg .. work?
[16:06] <nevcairiel> for most people its not work
[16:06] <nevcairiel> but their free time
[16:07] <xlinkz0> then I am eternally grateful for their development
[16:07] <xlinkz0> have you got 5 minutes nevcairiel ? :D
[16:07] <ubitux> if ppl know they will answer
[16:08] <ubitux> xlinkz0: rule #1, it's sad, but you're most of the time on your own
[16:08] <xlinkz0> i understand, that's why i'm trying to debug the problem on my own.. as my abilities can
[16:08] <Daemon404> and when t comes to timestamps... nobody knows
[16:08] <Daemon404> it's voodoo
[16:09] <xlinkz0> lol
[16:09] <Daemon404> it just exists
[16:09] <xlinkz0> ok i'll just try to ask simple questions.. when cutting a file without -avoid-negative-ts where does the negative start time come from?
[16:09] <xlinkz0> what is its significance?
[16:11] <Daemon404> i dont think you understand... hardly anyone knows WTF ffmpeg.c does
[16:11] <Daemon404> except the people who wrote it (maybe michaelni ?)
[16:11] <xlinkz0> neat
[16:12] <Daemon404> dont worry, i hate it as much as you do.
[16:12] <ubitux> yeah but now it's split into 3 ffmpeg files
[16:12] <ubitux> so it's much better ;)
[16:17] Action: Compn throws some timestamps at Daemon404
[16:36] <xlinkz0> michaelni__: ping me when you have some spare time please..
[17:18] <michaelni__> xlinkz0, i never have time, but whats your question
[17:19] <michaelni__> ?
[17:19] <gnafu> While I would agree with anyone who might say it was in bad taste and wouldn't be appreciated more than once, I quite enjoyed Wheelman's short-lived presence in this channel.
[17:20] <xlinkz0> michaelni__: should ist->pts ever be negative in do_streamcopy() ?
[17:21] <xlinkz0> when cutting, what is the reason of negative timestamps in the resulting video?
[17:22] <michaelni__> i see no reaso why it wouldnt be negative if the demuxer / decoder returned negative timestamps
[17:23] <xlinkz0> michaelni__: but then the problem remains that if ist->pts starts negative, the resulting video segment will not have the exact duration the user requested
[17:27] <michaelni__> xlinkz0, if you request duration X and get something far off then thats a bug, please open a ticket for that with input sample and command line used
[17:28] <xlinkz0> is 1.2 seconds far off?
[17:29] <michaelni__> hard to say, 1.2 seconds could be just due to keyframe spacing
[17:32] <xlinkz0> michaelni__: what does keyframe spacing have to do with duration?
[17:33] <xlinkz0> as far as I can understand ffmpeg tool doesn't look for keyframes when determining if the duration has been achieved
[17:35] <xlinkz0> michaelni__: the problem is keyframe spacing is causing negative timestamps which affect the duration, as far as I can tell from the ffmpeg.c code and my tests
[17:48] <michaelni__> xlinkz0, so if i understand correctly the start time is of due to there being no keyframe and the duration is shorted by the amount that the starttime is off but the end time would be matching ?
[17:50] <xlinkz0> michaelni__: i created a file with a gop of 500 frames, that is about one keyframe every 20 seconds
[17:50] <xlinkz0> so i have a keyframe on frame 0, 20 , 40
[17:50] <xlinkz0> this command : ffmpeg -y -ss 10 -i video.mp4 -t 2 -avoid_negative_ts 1 -c copy test.mp4
[17:50] <xlinkz0> results in a 12 second long video
[17:52] <xlinkz0> if that isn't far off..
[17:52] <xlinkz0> this is proportional to the gop_size of the video
[17:59] <xlinkz0> michaelni__: after testing the latter I am sure this is a bug, I believe it is easy to solve, just adjust output_file->recording_time based on the offset of the first frame of the input stream but I'm scared to do it myself since I don't know what else I might break..
[18:01] <michaelni__> iam a bit curious what the use case is for files for exact duration with quite incorrect starttime
[18:01] <michaelni__> because with -ss 10 and keyframes at 0 and 20 its starttime will be 10 seconds off
[18:01] <xlinkz0> but why should that affect duration....
[18:02] <xlinkz0> it makes no sense
[18:02] <xlinkz0> i just made that above video to show that it is 'far off' as you put it, there are videos out there with gop_size of 5 seconds which could result in duration-increase of up to 4 seconds
[18:02] <michaelni__> you ask ffmpeg for a video from 10 to 12 and it gives you one from 0 to 12 i assume
[18:02] <xlinkz0> which is a lot
[18:02] <xlinkz0> yes
[18:03] <michaelni__> iam just wondering what usecase is there that would prefer 0 to 2 in that case
[18:03] <xlinkz0> what i'm working on relies on cutting user-selected videos
[18:04] <michaelni__> it sure is a bug in the sense that the duration isnt honored
[18:04] <xlinkz0> like snapshots of some important thing they did in some sport, then concatting them together to form a user video
[18:04] <xlinkz0> they might have 10 cuts inside a single user video
[18:04] <michaelni__> and that works with start times being 10seconds off ?
[18:04] <xlinkz0> that 1 second of footage on every cut adds up to a lot
[18:04] <xlinkz0> no, i get 1.2 seconds in surplus
[18:04] <xlinkz0> but since a user might have 10-20 cuts 
[18:05] <xlinkz0> and it's just not professional to tell the user : hey you can cut from x to y but we'll go ahead and put from x to y+1.2 seconds 
[18:05] <xlinkz0> 1.2 seconds the user doesn't care about
[18:05] <xlinkz0> ..
[18:06] <michaelni__> with stream copy you can only start cuts from keyframes
[18:06] <xlinkz0> i'm sure there are other use cases where this ends up screwing someone
[18:06] <xlinkz0> i know
[18:08] <xlinkz0> the thing is i can't control that and that is just... necessary surplus frames because i need to start on a keyframe
[18:08] <xlinkz0> add the duration bug on top of it and it ends up being a lot :)
[18:09] <michaelni__> I dont see how the duration "bug" would be a problem for you, from what you say a fixed version would be alot worse for this usecase
[18:10] <michaelni__> if a user wants a 2sec cut from 10seconds to 12seconds and gets one from 0 to 12 that is too long but at least contains what he wanted
[18:11] <michaelni__> a fixed version would give him a 2 sec cut from 0 to 2 which is outside the part he wanted
[18:11] <michaelni__> or am i misuunderstanding what you want to do ?
[18:11] <xlinkz0> hmm
[18:12] <xlinkz0> on second thought you may be right :D
[18:13] <xlinkz0> i'll investigate further
[18:13] <ubitux> on a totally unrelated note, i e ffmpeg's eval
[18:13] <xlinkz0> sorry for bothering you
[18:16] <ubitux> do we have something for int() in eval?
[18:17] <ubitux> ah, trunc()
[18:18] <durandal_1707> what you use eval for?
[19:05] <ubitux> durandal_1707: i'm not done, but i was trying some nasty thing with scale
[19:06] <durandal_1707> is anybody usinh hrtf from libaf?
[19:07] <ubitux> those who don't know pan syntax to downmix 
[19:08] <durandal_1707> it doesn't do just simple downmix....
[19:09] <ubitux> durandal_1707: yes but the syntax is simpler
[19:09] <ubitux> "close enough"©
[19:10] <durandal_1707> but it is far from pleasant experience
[21:13] <Compn> not at 200 comments yet?
[21:13] <Compn> cmon ticket 2686, get to it!
[21:14] <durandal_1707> tell them to switch to irc
[21:36] <ubitux> what's the range of the global_quality for the aac enc?
[21:36] <ubitux> it looks like it's similar to faac (same default, 120)
[22:10] <durandal11707> anyone uses headphones here a lot?
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:005b38f8f141: avcodec/ffv1dec: check global header version
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:2fd824b4662a: ffv1dec: Check bits_per_raw_sample and colorspace for equality in ver 0/1 headers
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:54bdb5fc8642: avcodec/dsputil: fix signedness in sizeof() comparissions
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:db99c4156711: avfilter/vf_fps: make sure the fifo is not empty before using it
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:ef121a88d571: avcodec/mjpegdec: Add some sanity checks to ljpeg_decode_rgb_scan()
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:aeec1a6430b4: avcodec/truemotion2: Fix av_freep arguments
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:40e52bbb63cf: avcodec/ffv1enc: update buffer check for 16bps
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:0a64b25c7732: avcodec/g2meet: Fix order of align and pixel size multiplication.
[22:31] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:2b06f5f8f15a: avcodec/g2meet: Fix framebuf size
[22:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:c7966bf79564: avformat/oggparsevorbis: fix leak of ct
[22:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:c52a25e03bb4: avformat/oggparsevorbis: fix leak of tt
[22:40] <cone-114> ffmpeg.git 03James Almer 07release/2.0:414d75b8bc06: matroskadec: Improve TTA duration calculation
[22:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:dbb534cea6b2: avformat/matroskadec: check out_samplerate before using it in av_rescale()
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:5753d780b47c: doc/encoders: partially rewrite and reformat libx264 docs
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:68ae344b5e11: doc/encoders: add libxvid doc
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:c08e8ab715bb: doc/encoders: reformat libmp3lame doc
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:01b39884c7f0: doc/encoders: reformat and add some clarification in libtwolame doc
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:9ebfee7ac0f5: doc/encoders: improve libvo-aacenc doc
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:cecb2b39cef1: doc/formats: Add documentation for 3 parameters that have been missing
[22:40] <cone-114> ffmpeg.git 03Timothy Gu 07release/2.0:6911d9e1b0cf: doc/encoders: add doc for AAC encoder
[22:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:211ad5042adc: Merge remote-tracking branch 'jamrial/release/2.0' into release/2.0
[22:40] <cone-114> ffmpeg.git 03Michael Niedermayer 07release/2.0:187297b871a3: Merge remote-tracking branch 'TimothyGu/release/2.0' into release/2.0
[23:25] <dbro> is MP4 -> TS conversion simple enough to do manually without FFMPEG? 
[23:25] <dbro> writing a FOSS HTTP HLS / MPEG-DASH server for Android
[23:26] <dbro> I've gotten the mp4 chunk generation working, and wondering if I've gotta cross compile ffmpeg to do the final mp4 -> ts
[23:26] <dbro> cheers!
[23:31] <kierank> writing a ts mux isn't too hard
[23:31] <kierank> just a bit tedious
[23:32] <dbro> aye, given an mp4 with one audio and video stream?
[23:32] <dbro> thanks
[23:36] <kierank> bit pointless though
[23:36] <kierank> libavformat will be fine
[23:38] <dbro> yeah, it's just always a pain trying to cross compile for android given the headache of different architectures 
[23:38] <dbro> but yeah, that's probably best
[23:53] <cone-114> ffmpeg.git 03Lukasz Marek 07master:23ad4d6f58e9: doc/outdevs: add sdl window_fullscreen option
[00:00] --- Tue Sep 24 2013


More information about the Ffmpeg-devel-irc mailing list