[Ffmpeg-devel-irc] ffmpeg-devel.log.20130705

[burek]

[00:43] Action: durandal_1707 forked FFmpeg
[01:05] <cone-411> ffmpeg.git 03Stefano Sabatini 07master:41fe750f4b13: lavc/bitstream_filter: do not crash in case the argument of av_bitstream_filter_close() is NULL
[01:05] <cone-411> ffmpeg.git 03Stefano Sabatini 07master:d06ae0d791cd: lavf: fix documentation for avoid_negative_ts
[04:09] <cone-566> ffmpeg.git 03Michael Niedermayer 07master:b009267910df: mpegts: only reopen pmt_cb filter if its different from the previous.
[09:42] <saste> You can use libavcodec or libavformat in your commercial program, but any patch you make must be published.
[09:42] <saste> this is BS
[10:11] <burek> <@michaelni> burek, btw, how is the fate server rewrite going ?
[10:11] <burek> well i didnt rewrite the server, i just wrote another set of scripts to display what's already uploaded
[10:12] <burek> i.e. i've avoided to constantly decompress things on each page view
[10:12] <burek> shortly, i just added a different view to logs that already exist on the fate server
[10:12] <burek> and btw, i was ready a couple of months ago to test it on the fate server itself
[10:13] <burek> so, i'm kinda almost finished with it, just we need to find some time to test all that on a real fate server
[10:13] <burek> but i'll need a day/two to remember what was i doing in those php scripts, before we decide to test in a real environment
[10:14] <burek> just let me know when is the good time for testing, so i can prepare before it
[11:05] <cone-688> ffmpeg.git 03Stefano Sabatini 07master:0f85c96091e9: lavc/avcodec.h: document bitstream filter API
[11:37] <michaelni_> burek, dont know when theres a good time, whats the difference between good and bad time for this ?
[11:43] <saste_> michaelni_, difference between good and bad time -> consult your favorite astrologist
[11:46] <cone-688> ffmpeg.git 03Kostya Shishkov 07master:410066986f44: aic: use chroma scan tables while decoding luma component in progressive mode
[11:46] <cone-688> ffmpeg.git 03Jason Garrett-Glaser 07master:d222f6e39e26: cabac: x86 version of get_cabac_bypass
[11:46] <cone-688> ffmpeg.git 03Luca Barbato 07master:e6d8acf6a8fb: indeo: use a typedef for the mc function pointer
[11:46] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:cced6f4d5816: Merge commit 'e6d8acf6a8fba4743eb56eabe72a741d1bbee3cb'
[11:58] <cone-688> ffmpeg.git 03Luca Barbato 07master:f6f36ca8ca1b: indeo: Refactor ff_ivi_dec_huff_desc
[11:58] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:0e2056056fc8: Merge commit 'f6f36ca8ca1b2526d3abff7d7c627322d3bce912'
[12:08] <cone-688> ffmpeg.git 03Luca Barbato 07master:62256010e9bc: indeo: Refactor ff_ivi_init_tiles and ivi_decode_blocks
[12:08] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:9b10440dcd1a: Merge commit '62256010e9bc8879e2bf7f3b94af8ff85e239082'
[12:13] <cone-688> ffmpeg.git 03Luca Barbato 07master:6dfacd7ab126: indeo: Cosmetic formatting
[12:13] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:7d3b55c6a427: Merge commit '6dfacd7ab126aea1392949d1aa10fdc3d3eeb911'
[12:33] <cone-688> ffmpeg.git 03Luca Barbato 07master:6a10142faa1c: indeo: reject negative array indexes
[12:33] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:9f89bebb7bff: Merge commit '6a10142faa1cca8ba2bfe51b970754f62d60f320'
[12:42] <cone-688> ffmpeg.git 03Stefano Sabatini 07master:d880b2bdfff6: doc/filters: clarify documentation for overlay repeatlast option
[12:49] <cone-688> ffmpeg.git 03Luca Barbato 07master:0420c810ceb4: log: pass the correct parameters to missing_feature_sample
[12:49] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:8fdec02fd195: Merge commit '0420c810ceb430003f9f2793c5cfa1fe84657f3d'
[12:59] <kierank> does 24-bit wav exist
[13:17] <cone-688> ffmpeg.git 03Luca Barbato 07master:d3635f3ab0a2: compat: wrap math.h to avoid AIX-specific clashes
[13:17] <cone-688> ffmpeg.git 03Michael Niedermayer 07master:47ca9f041cde: Merge remote-tracking branch 'qatar/master'
[13:50] <cone-688> ffmpeg.git 03Peter Große 07master:c6c2231227c1: libavcodec: use 64bit counter in deprecated audio encoder API to prevent problems regarding negative dts
[14:02] <cone-688> ffmpeg.git 03Jean Delvare 07master:6fc8c4cc0ef6: lavfi/delogo: don't recompute the same difference again and again
[14:02] <cone-688> ffmpeg.git 03Jean Delvare 07master:4e8d6b315c8d: lavfi/delogo: remember left and right samples when interpolating
[15:21] <durandal_1707> there is no Firefox OS for fate
[15:24] <khali> doh
[15:24] <khali> Icleanedupmykeyboardandnowthespacebarnolongerworks
[15:24] <khali> thissucks!
[15:25] <iive> it this joke or is it for real?
[15:26] <durandal_1707> space bar issue?
[15:26] <saste_> khali, so you'll be saving space since now
[15:28] <saste_> michaelni_, could you mentor a socis JPEG2000 project?
[15:28] <durandal_1707> why? someone want to work on j2k?
[15:28] <saste_> durandal_1707, i have a mail on my inbox from a potential student
[15:29] <saste_> SOCIS accepted organizations list is still waited for almost always a week since the official announcement date
[15:30] <saste_> BuxiNess is not sure he will be able to mentor in the program period (August - October)
[15:30] <saste_> i don't even know what it is still missing...
[15:30] <durandal_1707> proper encoder
[15:31] <durandal_1707> saste_: did you checked your spam box?
[15:31] <saste_> DEVILS jpeg2000
[15:31] <saste_> so we miss features?
[15:32] <durandal_1707> well native one supports only rgb
[15:33] <saste_> is it enough we lack to cover an entire socis project?
[15:33] <khali> iive: for_real,_sadly_:(
[15:34] <khali> iive: I_tried_fixing_it_but_could_not
[15:34] <durandal_1707> you could hack your irc client to replace _ with " "
[15:34] <khali> I_guess_my_work_day_is_over_and_I_can_look_for_a_replacement_keyboard_right_away
[15:35] <saste_> khali, try to dump some dirt on your keyboard, to restore the state before you cleaned it up
[15:35] <durandal_1707> install virual keyboard
[15:35] <iive> khali: did you open the keyboard, or you cleaned just the space under the keys?
[15:35] <khali> durandal_1707: I_could_but_IRC_is_not_my_main_concern_;)
[15:35] <ubitux> copy a space, use ^K to put it in the buffer
[15:35] <ubitux> and then use ^Y to add it again
[15:35] <saste_> ubitux, yes that's handy
[15:35] <khali> doh,_right_up_and_down_keys_are_dead_too
[15:36] <khali> iive: I_opened_it
[15:36] <ubitux> ^A ^E for start/end buffer
[15:36] <saste_> who needs to go up and down anyway, just go left and right
[15:36] <khali> I_guess_I_shouldn't_have_:(
[15:36] <BuxiNess> jpeg2000 needs speed ! :-)
[15:36] Action: saste_ stops saying silly things
[15:36] <khali> saste_:right_is_dead_too
[15:37] <iive> khali: then open it again, probably the line that goes to the micro-controller have been moved a little.
[15:37] <saste_> khali, i suppose it is a laptop?
[15:37] <saste_> yeah checking the physical connections may help
[15:38] <khali> no_its_a_regular_keyboard
[15:38] <khali> Key_Tronic_from_1999_or_so
[15:38] <saste_> well at least it is easy to replace
[15:38] <khali> saste_: true
[15:39] <khali> only_I'd_love_to_find_a_Key_Tronic_again...but_they_seem_to_no_longer_exist_:(
[15:40] <iive> khali: the micro controller is the chip under the led.
[15:43] <durandal_1707> are there .jpx samples?
[15:51] <BuxiNess> durandal_1707, I didn't know any
[15:52] <durandal_1707> nothing can create them?
[15:53] <BuxiNess> With START/STOP_TIMER, in a function who is supposed to be constant. If the number of decicycles inrceases. Its not normal, right? 
[15:53] <khali> hallelujah! 
[15:53] <khali> space bar is back :)
[15:54] <khali> I cleaned the conductive sheets with 90° alcool 
[15:54] <durandal_1707> you broke % key
[15:54] <khali> %no I did not% :)
[15:54] <BuxiNess> Mu nb of decicycles increase afters each frame decocding 2 runs < 4 runs < 16 runs < 32 runs ...
[15:55] <durandal_1707> what function?
[15:55] <khali> durandal_1707: alcool concentrations are commonly counted in degrees, at least here
[15:55] <BuxiNess> in jpeg2000dec
[15:55] <khali> (don't ask me why, it makes little sense I admit)
[15:55] <BuxiNess> decode_frame 
[15:55] <Skyler_> typically you want to ignore the first few lines; only the results after a large number of iterations (10000, 100000, 1000000) matter
[15:55] <durandal_1707> yes, but what function?
[15:57] <BuxiNess> START_TIMER after get_buffer; STOP timer after jpeg2000_decode_tile
[15:57] <BuxiNess> in function jpeg2000_decode_frame
[15:58] <BuxiNess> Skyler_, yes but its for one frame, i decode just 240 frames at it takes 10 to 15 seconds
[15:58] <Skyler_> you might want to time a smaller section of the code then (that runs many more times) unless that's impossible
[15:59] <Skyler_> START_TIMER is not really designed for accurate/consistent timing of very large code blocks
[15:59] <Skyler_> e.g. you want to time something that takes on the order of 1 nanosecond to 100 microseconds
[15:59] <Skyler_> not 0.1 seconds
[16:01] <xlinkz0> is copy_tb set when setting -c copy?
[16:07] <BuxiNess> Skyler_, and I wonder how it works in multi-threaded env ...
[16:07] <BuxiNess> is there other helper functions for that ?
[16:07] <BuxiNess> for measuring 0.1 seconds time
[16:08] <Skyler_> it works by using the per-core timer
[16:08] <Skyler_> generally, it will give suboptimal results if the length of time you're timing is close to or larger than one context switch (e.g. 1ms or whatever your kernel uses)
[16:09] <Skyler_> it attempts to ignore runs on which there was a context switch
[16:10] <BuxiNess> okay this why some times there are some non 0 skips
[16:11] <Skyler_> I suggest using gettimeofday or something for really really long timers
[16:25] <Skyler_> basically START_TIMER "skips" the first two or so, plus anything with way longer timing results than the average so far
[16:25] <Skyler_> this is designed to catch context switches
[16:25] <Skyler_> naturally this doesn't really work quite as expected with runs that are 100ms long and will have 10 context switches in them
[16:30] <xlinkz0> do i need to open codecs when stream copying?
[16:32] <BuxiNess> Is there a wrapper function to get the the thread Id ?
[16:34] <durandal_1707> usually you have it when using slice threads
[16:48] <BuxiNess> And also how to get the frame ID in the decoder. 
[16:48] <BuxiNess> ie which frame is it, 1st, 2nd, ....
[16:51] <xlinkz0> what does this mean? Codec for stream 0 does not use global headers but container format requires global headers
[16:52] <BuxiNess> If I can Have in the decoder the frame position or ID, who is currently decoded
[16:55] <durandal_1707> why you need it:
[16:55] <BuxiNess> for debug pupose, I want to know whick frame is decoded in which thread, 
[16:56] <BuxiNess> and how log it takes to decode the frame
[16:59] <BuxiNess> pkt_pts is fine!
[17:09] <cone-352> ffmpeg.git 03Paul B Mahol 07master:9dd14477884c: lavfi/cropdetect: export cropdetect info to frame metadata
[17:31] <cone-352> ffmpeg.git 03Paul B Mahol 07master:413f865a9aa0: lavfi/blackdetect: support 2 more pixels formats
[18:16] <cone-352> ffmpeg.git 03Jean Delvare 07master:72abce3c4672: lavfi/delogo: option show shouldn't affect band
[18:16] <cone-352> ffmpeg.git 03Jean Delvare 07master:4e10d87f384a: lavfi/delogo: band width must be at least 1
[18:44] <durandal_1707> buxiness: rgb48 in jpeg2000 doesn't work at all
[18:44] <durandal_1707> michaelni_: &
[20:33] <cone-352> ffmpeg.git 03Paul B Mahol 07master:48f2750de8a4: get_bits: add get_bits_le()
[20:33] <cone-352> ffmpeg.git 03Paul B Mahol 07master:6516a25f04ed: ADPCM IMA WAV 2, 3 and 5 bits decoder
[20:43] <BBB> what's get_bits_le that BITSTREAM_READER_LE doesn't do?
[20:47] <Daemon404> BBB, work easily in the same file that also needs BE reader
[20:47] <Daemon404> without ugly define/undef all over the place
[20:51] Action: BBB won't object b/c he doesn't care, but raises eyebrows anyway
[20:53] <kierank> BBB: let's you mix and match?
[20:53] <kierank> lets*
[20:56] <durandal_1707> Daemon404: define undef doesn't work, only moving code into separate file ...
[20:57] <Daemon404> right...
[20:57] <Daemon404> i forgot it relies on the include stuff
[20:59] <BBB> kierank: putting a hack in get_bits.h for a design disaster in one obscure shitty codec seems questionable
[20:59] <BBB> maybe put that hack inside that codec's .c file?
[21:00] <Compn> Daemon404 : somehow i dont think your email re yadif will encourage anyone to rejoin the team :P
[21:04] <Daemon404> i dont recall recall that being such a goal
[21:04] <Daemon404> s/recall recall/recall/
[21:05] <durandal_1707> BBB: what is hack in get_bits.h?
[21:11] <durandal_1707> i could move get_bits_le out of get_bits.h, if that is what you're looking for
[22:14] <cone-352> ffmpeg.git 03Nigel Touati-Evans 07master:ef8cc06d6e50: avformat/avidec: Fix incorrect detection of badly interleaved avi
[22:14] <durandal_1707> why you bumped lavc when you changed lavf?
[22:28] <durandal_1707> what about adding generic graph filter which would draw points over time from some frame metadata?
[23:04] <michaelni_> durandal_1707, sloppy mistake :( lucky its not of any practical relevance, ill be more carefull
[23:04] <michaelni_> and such graph filter sounds quite interresting
[23:09] <durandal_1707> not if there is nothing that will use it
[23:34] <durandal_1707> so the only reason psnr filter is not in lavfi is because nobody cares?
[23:35] <nevcairiel> its because psnr is a terrible metric
[23:35] <nevcairiel> well for video anyway
[23:38] <durandal_1707> but when new video codec comes around everybody and he's dog use it
[00:00] --- Sat Jul  6 2013