[Ffmpeg-devel-irc] ffmpeg-devel.log.20120220

[burek]

[01:42] <CIA-48> ffmpeg: 03Anton Khirnov 07master * r5455384219 10ffmpeg/libavcodec/mpegvideo_enc.c: 
[01:42] <CIA-48> ffmpeg: mpegvideo_enc: force encoding delay of at least 1 frame when low_delay=0
[01:42] <CIA-48> ffmpeg: This allows the following commit to extrapolate better dts for the first
[01:42] <CIA-48> ffmpeg: frame. Pts difference between the first two frames is reused as the
[01:42] <CIA-48> ffmpeg: difference between pts and dts of the first frame.
[01:42] <CIA-48> ffmpeg: 03Anton Khirnov 07master * r445a7d48b1 10ffmpeg/libavcodec/ (10 files): mpegvideo_enc: switch to encode2().
[01:42] <CIA-48> ffmpeg: 03Aneesh Dogra 07master * r1a58daed0f 10ffmpeg/libavcodec/sunrastenc.c: 
[01:42] <CIA-48> ffmpeg: sunrastenc: set keyframe flag for the output packet.
[01:42] <CIA-48> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[01:42] <CIA-48> ffmpeg: 03Aneesh Dogra 07master * r566df2eea2 10ffmpeg/libavcodec/sunrast.c: 
[01:42] <CIA-48> ffmpeg: sunrast: use RLE trigger macro inplace of the hard coded value.
[01:42] <CIA-48> ffmpeg: Signed-off-by: Anton Khirnov <anton at khirnov.net>
[01:42] <CIA-48> ffmpeg: 03Paul B Mahol 07master * re04ca1d4ce 10ffmpeg/libavcodec/tta.c: 
[01:42] <CIA-48> ffmpeg: ttadec: cosmetics: reindent
[01:42] <CIA-48> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[02:56] <overflow_0f8b> no way telling gcc not to jump around like a stupid monkey?
[02:57] <overflow_0f8b> tried to force gcc make cmovs, but it likes to jump
[03:22] <Compn> overflow_0f8b : -O0 ? ;P
[03:22] <Compn> unroll all loops ?
[03:22] Action: Compn no clue
[03:31] <overflow_0f8b> used -o3
[09:24] <CIA-17> ffmpeg: 03Clément BSsch 07master * r6e6bdeb137 10ffmpeg/doc/filters.texi: 
[09:24] <CIA-17> ffmpeg: doc: fix channel layout value in abuffer example.
[09:24] <CIA-17> ffmpeg: The channel layout must start with 0x, otherwise it is parsed as a
[09:24] <CIA-17> ffmpeg: number of channels.
[10:34] <ubitux> michaelni: how do you notice the timefilter divergence issues you're talking about with Nicolas?
[15:04] <mateo`> hi, it seems that this merged commit from qatar (9ce2a91b842bdb99fae0f153c5317d786a3f38d8), disable the use of some mpeg2 flag such as ivlc and non_linear_q.
[15:05] <mateo`> is there a new method to specify them ?
[15:05] <av500> try bahrain?
[15:05] <av500> or the emirates
[15:20] <michaelni> mateo`, there are codec specific options, look at the mpeg2_options[] and mpeg1_options[] tables
[15:20] <michaelni> havnt checked if they are docuemented
[15:20] <michaelni> if not a patch that adds them to the docs would be welcome
[15:22] <mateo`> michaelni: thanks
[15:33] <Daemon404> michaelni, in swscale, what does readChrPlanar refer to?
[15:33] <Daemon404> i.e.it's a function pointer, but what is said function supposed to do  
[15:34] <cbsrobot> maybe read chroma planar ?
[15:34] <Daemon404> hmm perhaps
[15:34] <Daemon404> looks like i might need to extend swscale to do what i want...
[15:35] <cbsrobot> what are you working on ?
[15:35] <Daemon404> currently, GBRP is -always- converted to yuv (and scaled too mught i add), and then converted back to packed rgb
[15:35] <Daemon404> when doing a GBRP -> RGB conversion.
[15:36] <Daemon404> obviously this is Bad (TM)
[15:36] <cbsrobot> says who ?
[15:36] <Daemon404> what do you mean?
[15:36] <Daemon404> says who what?
[15:37] <cbsrobot> "this is Bad"
[15:37] <michaelni> Daemon404, you are planing to add a RGB scaling codepath ?
[15:37] <Daemon404> maybe
[15:37] <michaelni> :)
[15:37] <Daemon404> cbsrobot, if youre doing gbrp to rgb conversion, it's simpyl reordering bytes
[15:37] <Daemon404> converting to yuv in between is BAD
[15:37] <Daemon404> fucks up color
[15:37] <cbsrobot> ah I get it
[15:37] <Daemon404> andd introduces roundign errors/.
[15:38] Action: cbsrobot understood adding gbrp->rgb is bad
[15:38] <michaelni> btw, strictly speaking it shouldnt fuck up color (not counting rounding)
[15:38] <Daemon404> it's getting scaled.
[15:38] <Daemon404> to tv levels, i think
[15:38] <Daemon404> (limited range yuv)
[15:38] <cbsrobot> btw do we support rec709?
[15:39] <michaelni> we should use more bits
[15:39] <cbsrobot> I only found rec609 conversion
[15:39] <michaelni> we partially support 709
[15:39] <Daemon404> michaelni, it's using limited range always
[15:39] <Daemon404> from what i udnerstand
[15:39] <Daemon404> and never full range
[15:39] <cbsrobot> michaelni: where ?
[15:39] <Daemon404> [09:37] < michaelni> Daemon404, you are planing to add a RGB scaling codepath ? <-- BBB said he was going to, Eventually (TM)
[15:40] <Daemon404> im merely impatient :x
[15:40] <michaelni> cbsrobot, yuv->rgb has user selectable tables
[15:40] <Daemon404> since this means GBR avc decoding will always have an unneeded yuv conversion when used with e.g. swscale + avisynth
[15:41] <cbsrobot> ah ff_yuv2rgb_coeffs ?
[15:41] <cbsrobot> could we add xyz to that ?
[15:43] <michaelni> cbsrobot, see sws_setColorspaceDetails() for the user API of setting yuv variant
[15:44] <michaelni> about adding xyz, i suspect it would need some additional changes
[15:44] <cbsrobot> michaelni: just found it
[15:44] <cbsrobot> I think so too
[15:44] <cbsrobot> I'm just digging - you know the details
[15:44] <michaelni> as rgb<->yuv have some 0 and 1 in the matrix that i belive are not in xyz
[15:45] <michaelni> and we probably have optimizations for all 0 and 1 multiplies
[15:48] <cbsrobot> also rgb -> xyz would be useful ...
[15:54] <michaelni> yes
[16:35] <ubitux> would anyone be interested in a history for code coverage?
[16:36] <ubitux> let's say 10 or 20 old code coverage reports
[16:36] <ubitux> or it doesn't really matter
[16:36] <ubitux> ?
[16:39] <Compn> mike maybe, ask himmmmmm
[16:39] <Compn> :P
[17:18] <ubitux> Compn: done :)
[17:19] <ubitux> saste: we don't have any audio filter to "buffer" a given number of samples, right?
[17:20] <saste> ubitux: no, unless you count abuffer_src which is not a "filter" 
[17:20] <ubitux> i was wondering for the FFT visualization filter
[17:21] <ubitux> because we need windows of a given number of samples
[17:21] <ubitux> also, it could help user app which are running similar routines
[17:22] <ubitux> the could get buffer of N samples each time, easy to deal with that doing the buffering themselves
[17:22] <ubitux> any thoughts about this?
[17:22] <ubitux> s/the could/they could/
[17:22] <ubitux> (this would suggest buffering with silence for the last frame)
[17:22] <cbsrobot> Compn: I had another idea
[17:23] <Compn> yes?
[17:23] <cbsrobot> for gsoc, but maybe it's too wired
[17:23] <cbsrobot> maybe someone is interested in a audio filter doing similar stuff to http://www.dolby.com/us/en/professional/software/broadcast/dolby-media-meter2.html
[17:23] <cbsrobot> I would be at least
[17:24] <ubitux> oh that would be great
[17:24] <av500> BS.1770-1
[17:24] <ubitux> cbsrobot: a lot of "music information retrieval" stuff could be done in the same idea
[17:24] <cbsrobot> http://www.google.com/url?sa=t&rct=j&q=itu-r%20bs.1770-1&source=web&cd=1&ved=0CC8QFjAA&url=http%3A%2F%2Fwebs.uvigo.es%2Fservicios%2Fbiblioteca%2Fuit%2Frec%2FBS%2FR-REC-BS.1770-1-200709-I!!PDF-E.pdf&ei=s3NCT_eAJ-jT4QSesbSHCA&usg=AFQjCNEFwjkzTRRuis1nj7GKu_Mf8qthtQ&cad=rja
[17:24] <ubitux> such as pitch, beat or key extraction
[17:25] <cbsrobot> ubitux: why not
[17:26] <cbsrobot> maybe kierank would be interested in ^ aswell
[17:26] <kierank> cbsrobot: stuff exists already to do that
[17:26] <kierank> libebu128
[17:26] <kierank> etc
[17:28] <Daemon404> kierank, you forget the great NIH policy 
[17:28] Action: Daemon404 runs
[17:29] <kierank> Daemon404: coming to bristol this weekend?
[17:29] <Daemon404> well its my last weekend here with my the gf...
[17:29] <Daemon404> weekdays work better for me :/
[17:29] <cbsrobot> kierank: you mean http://sourceforge.net/projects/r128gain/ ?
[17:29] <Daemon404> sitll waiting on my google uk results too... sooo ;)
[17:29] <kierank> cbsrobot: yeah and similar
[17:30] <kierank> erm i'm around tomorrow and wednesday
[17:30] <kierank> and friday afternoon i guess
[17:30] <kierank> tomorrow and wednesday afternoons
[17:30] <kierank> [16:28] Daemon404: kierank, you forget the great NIH policy  --> duh
[17:31] <Daemon404> kierank, oic
[17:31] <Daemon404> i could do a day-trip
[17:31] <Daemon404> it's a short train ride away after all
[17:32] Action: Daemon404 checks train prices
[17:32] <kierank> bbiab
[17:32] <Daemon404> k
[17:32] <cbsrobot> kierank: thanks - i wasn't aware of it ....
[17:38] <kierank> ok back
[17:39] <kierank> oh bleh you have to go to bristol parkway from exeter
[17:39] <Daemon404> not temple meads?
[17:39] <kierank> oh you can
[17:39] <kierank> never mind
[17:39] <kierank> temple meads is better
[17:40] <Daemon404> yes
[17:40] <AdamWorld> hi!
[17:40] <kierank> hello
[17:40] <Daemon404> kierank, well then wahts a good time to show up / what would we be doin' around bristol?
[17:41] <kierank> well to be honest i was just thinking of getting a beer
[17:41] <kierank> but i dunno what you want to do in bristol
[17:41] <Daemon404> beer is fine :)
[17:41] <AdamWorld> I run ffserver as in document and in client's command I run ffplay http://192.168.157.129:8090/test.asf but it fails to play.
[17:41] <AdamWorld> what's the problem?
[17:41] <AdamWorld> In other player like VLC works well.
[17:42] <Daemon404> AdamWorld, i think you want #ffmpeg
[17:42] <AdamWorld> Yes.
[17:42] <AdamWorld> Should I ask in #ffmpeg?
[17:42] <Daemon404> yea
[17:43] <AdamWorld> Then What's this #ffmpeg_devel for?
[17:43] <kierank> developing ffmpeg
[17:43] <AdamWorld> Sorry.
[17:44] <kierank> Daemon404: not entirely sure what we can do apart from drink
[17:45] <kierank> and i need to think of a decent place to drink
[17:45] <Daemon404> kierank, drinking is fine
[17:45] <Daemon404> it's actually what i expected :P
[17:45] <kierank> problem is after about 2hrs we won't have much to talk about
[17:46] <Daemon404> well i should probably be back here by ~6:30pm latest
[17:46] <Daemon404> or i will be murdered, etc
[17:46] <kierank> well as long as it doesn't seem like a crap day trip to you
[17:46] <Daemon404> nah
[17:46] <Daemon404> meeting people is never a crap day trip
[17:48] <saste> ubitux: sure a buffer audio filter may be useful
[17:49] <ubitux> could we consider to auto insert with the help of query_formats?
[17:54] <kierank> Daemon404: i think the only places open during the day will be wetherspoons
[17:55] <Daemon404> lol i dont know what that is
[17:55] <Daemon404> but im sure it's better than anything in canada.
[17:55] <kierank> it's a big chain of pubs
[17:55] <CIA-17> ffmpeg: 03Stefano Sabatini 07master * r3293b1adca 10ffmpeg/libavcodec/ (utils.c version.h): 
[17:55] <CIA-17> ffmpeg: lavc: set bit_rate in the decoder context, just after the init stage
[17:55] <CIA-17> ffmpeg: Also move the get_bit_rate() function, in order to avoid an undefined
[17:55] <CIA-17> ffmpeg: symbol reference in avcodec_open2().
[17:55] <kierank> the beer is usually not crazy expensive but the pubs are a bit corporately
[17:55] <CIA-17> ffmpeg: 03Matthieu Bouron 07master * r278d6ab900 10ffmpeg/ (doc/ffprobe.xsd ffprobe.c): 
[17:55] <CIA-17> ffmpeg: ffprobe: report bit rate in stream description
[17:55] <CIA-17> ffmpeg: Signed-off-by: Stefano Sabatini <stefasab at gmail.com>
[17:56] <kierank> corporatey*
[17:56] <Daemon404> oic
[17:56] <Daemon404> but better beer than bud light
[17:56] <Daemon404> re: north america
[17:56] <kierank> oh yes
[17:56] <Daemon404> ;)
[17:57] <kierank> the beer is good and usually they have nice ales
[17:57] <Daemon404> lovely
[17:57] <kierank> http://www.jdwetherspoon.co.uk/home/pubs/v-shed
[17:57] <Daemon404> what's a preferabke time for me to show up then?
[17:57] <kierank> which day?
[17:57] <Daemon404> tomorrow or next
[17:57] <Daemon404> i can do either
[17:57] <kierank> about 2ish
[17:58] <kierank> that place doesn't look too bad
[17:58] <Daemon404> ok lets looky at tickets
[17:58] <kierank> and is on the waterside
[17:58] <Daemon404> getting around Uk is so easy...
[17:58] Action: Daemon404 notes back home is has a 800 km radius of forest on all sides...
[17:58] <kierank> if you want a pub closer to the station or further feel free to say so
[17:58] <Daemon404> it*
[17:58] <kierank> the famous ones won't be open during the day unfortunately
[17:58] <Daemon404> thats fine
[17:59] <Daemon404> i dont care much for "famous"
[17:59] <Daemon404> i care for good beer and friends
[17:59] <kierank> well the pub i had in mind is famous because it has its own beer
[17:59] <kierank> but it's quite far
[17:59] <Daemon404> oic
[17:59] <kierank> and small so not easy to talk
[17:59] <Daemon404> well ill leave it in your hands
[17:59] <Daemon404> im rather impartial
[18:00] <Daemon404> but i do know your bus system as confused me before 
[18:00] <kierank> it may be best to walk from the station
[18:00] <kierank> depends on the pub i guess
[18:01] <Daemon404> yea
[18:02] <kierank> let's see what facebook has to say about bristol central pubs
[18:04] <kierank> bbl
[18:04] <Daemon404> k
[18:04] <Daemon404> poke me when ya get back.
[18:12] <Daemon404> an anytime return is cheaper than a scheduled one.... logic.
[18:15] <Daemon404> kierank, looks like i have to go weds... tesco deilvery is tomorrow.. lol
[18:15] <kierank> ok
[18:17] <ubitux> is Andrey Utkin on IRC?
[18:17] Action: Daemon404 books trains
[18:19] <Daemon404> man, what is it with british places and charging me extra if i want to use a credit card
[18:20] <av500> use her majesties huge coins
[18:20] <kierank> Daemon404: really?
[18:20] <kierank> you should be able to buy the ticket from whichever train company you please
[18:20] <kierank> http://www.moneysavingexpert.com/travel/cheap-train-tickets#find
[18:20] <Daemon404> im using thetrainline.com
[18:21] <Daemon404> anyway, never seen this practice in north america
[18:22] <kierank> well if you haven't finished already use one of the ones it suggests
[18:22] <kierank> because thetrainline is a ripoff
[18:22] <Daemon404> ok
[18:29] <ohsix> how much extra, .25-.50c is pretty normal at places that only do small transactions, to offset the 3% visa or whoever their merchant is gets
[18:29] <Daemon404> 3.5 pounds
[18:29] <ohsix> ouch
[18:30] <ohsix> on top of currency conversion?
[18:30] <Daemon404> yea
[18:30] <Daemon404> and yes.
[18:30] <ohsix> could come from a lot of places then
[18:30] <Daemon404> kierank, booked
[18:30] <Daemon404> i arrive wed, ~1:30pm (cheapest), open return.
[19:32] <kierank> Daemon404: ok
[19:33] <kierank> all my friends suggest wetherspoons
[19:34] <Daemon404> all right
[19:34] <kierank> the 8 or 9 bus and get off at the big university tower
[19:34] <kierank> (park street)/berkeley squre
[19:34] <overflow_0f8b> hellou
[19:34] <kierank> the bus should stop right outside the pub
[19:34] <Daemon404> ok
[19:34] Action: Daemon404 saves that on his phone
[19:37] <kierank> the university tower is called the wills memorial building
[19:37] <kierank> you won't miss it
[19:38] <Daemon404> aight
[19:59] <knoppix> Hello.I am not sure whether I should ask questions on developing libavfilter here or at #ffmpeg.
[19:59] <kierank> here
[20:01] <knoppix> O.k.....I had a look at http://wiki.multimedia.cx/index.php?title=FFmpeg_filter_HOWTO .However,it seems to be outdated.I am interested in learning how to add filters to libAVFilter.What should I do?
[20:02] <ubitux> audio or video filter?
[20:03] <knoppix> Video.
[20:04] <ubitux> what kind of processing is it? modifying the video, or deleting/adding frames?
[20:04] <knoppix> I am interested in both.
[20:06] <ubitux> you have various kind of filters, better look at simple ones
[20:06] <ubitux> vf_scale is "simple" for instance, as a content modifying example
[20:06] <ubitux> (changing props too)
[20:06] <ubitux> vf yadif/select/thumbnail do "selection"
[20:07] <ubitux> if you're looking for filering out frames
[20:07] <ubitux> you also have filters like vf_blackframe just to analyze content, and not modify it
[20:08] <Compn> knoppix : ask saste about filters :)
[20:08] <knoppix> I see...I will eventually understand it by studying the code,but I want documentation.
[20:08] <Compn> saste should update the wiki docs ;\
[20:08] <knoppix> Ohh...I see............thanks Compn.
[20:10] <knoppix> Hello saste.
[20:11] <ubitux> knoppix: also, asking specific questions about the issues you expect is a good way of learning about the internals
[20:11] <saste> work calls for more work (and no profit)... i should have never start to work on lavfi :/
[20:12] <ubitux> imo, more efficient that incomplete (or deprecated) doc :)
[20:12] <saste> knoppix: yes i'll eventually document it (or write a tutorial) but only $gods know when
[20:12] <ubitux> saste: hey, you already have nicolas & me to work on the filters to help you ;)
[20:12] <saste> ubitux: yep :)
[20:13] <ubitux> (or ask you for reviews)
[20:13] <ubitux> (and help)
[20:13] <knoppix_> O.k.........thank you.
[20:13] <saste> knoppix: better thing is to ask on ffmpeg-devel if you want to contribute, or ask on libav-user if you want help for using the API
[20:17] <saste> gottago now
[20:18] <ubitux> cya :)
[20:43] <CIA-17> ffmpeg: 03Reimar Döffinger 07master * r016c7bb762 10ffmpeg/libavformat/matroskadec.c: 
[20:43] <CIA-17> ffmpeg: matroskadec: remove H.264 special-case that disabled header parsing.
[20:43] <CIA-17> ffmpeg: The parser was fixed so this workaround should no longer
[20:43] <CIA-17> ffmpeg: be necessary.
[20:43] <CIA-17> ffmpeg: This allows using stream-copy to fix files with keyframes
[20:43] <CIA-17> ffmpeg: incorrectly marked.
[20:43] <CIA-17> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[20:44] <overflow_0f8b> took another a few hours to reverse GCC's logic and make it generate 2 cmovs :) though it still does mov edx,edx;mov ebx,ebx; ;>>>> but hey, it works
[20:44] <Daemon404> this sounds painful...
[20:46] <overflow_0f8b> this is probably because gcc is not able to realize ebx == ebx
[20:46] Action: Daemon404 does all his development with clang :D
[20:48] <overflow_0f8b> i should do a second pass on my asm output and remove mov eax, eax and similar lines
[20:52] <Daemon404> why are you pokign gcc-generated asm anyway?
[20:53] <overflow_0f8b> coz i like to synthesize C code instead of inline asm hacking
[20:54] <overflow_0f8b> it is possible to code in C using logic
[20:54] <Daemon404> not sure what you mean by synthesize
[20:54] <Daemon404> so "code in such a way that you minimize gcc retardisms"
[20:54] <overflow_0f8b> yes
[20:54] <overflow_0f8b> but i have hit a wall now with these mov eax,eax type things
[20:55] <overflow_0f8b> gcc does not realize pointer a == uintptr_t a
[20:55] <kierank> those are normal, no
[20:55] <overflow_0f8b> and used -O3
[20:55] <Daemon404> i think optimizing for a certain compiler is a tad silly,imho
[20:56] <Daemon404> in fact, youre optimizing for that specific version
[20:56] <overflow_0f8b> hoping 4.4.4 and newer generates using similatr logic...
[20:58] <CIA-17> ffmpeg: 03Nicolas George 07master * r560b224f53 10ffmpeg/ (5 files in 2 dirs): libswr: allow to set custom matrices.
[20:58] <CIA-17> ffmpeg: 03Nicolas George 07master * ra44b510d56 10ffmpeg/ (doc/filters.texi libavfilter/af_pan.c): af_pan: use libswr for rematrixing.
[20:58] <CIA-17> ffmpeg: 03Nicolas George 07master * ra297856b8c 10ffmpeg/libavfilter/af_pan.c: af_pan: reindent after last commit.
[21:23] <CIA-17> ffmpeg: 03Reimar Döffinger 07master * r6a3f1726af 10ffmpeg/ffmpeg.c: 
[21:23] <CIA-17> ffmpeg: Add missing av_free_packet.
[21:23] <CIA-17> ffmpeg: Fixes memory leak when encoding at least with mpegvideo using the
[21:23] <CIA-17> ffmpeg: new encode2 function.
[21:23] <CIA-17> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[22:52] <kierank> Daemon404: we could go here actually: http://en.wikipedia.org/wiki/Llandoger_Trow
[22:52] <kierank> but it's more of a food place
[22:53] <kierank> worth seeing apparently
[23:01] <pasteeater> interesting looking place.
[23:03] <pasteeater> better than the shithole across my street: http://en.wikipedia.org/wiki/Alaskan_Hotel_and_Bar
[23:05] <Daemon404> kierank, cool
[23:05] Action: Daemon404 must tend to his other half atm,so bbl!
[23:16] <overflow_0f8b> Daemon404<< i like this style, i just leave the dirty asm scheduling and logistic thing to C, and write a C-code that can be logically translated into an ASM code
[23:16] <overflow_0f8b> in most of the cases i can get away with this
[23:38] <CIA-17> ffmpeg: 03Nicolas George 07master * ra923b6b8f4 10ffmpeg/libavfilter/af_pan.c: 
[23:38] <CIA-17> ffmpeg: af_pan: move query_format before config_props.
[23:38] <CIA-17> ffmpeg: That is the logical order in which the functions are called,
[23:38] <CIA-17> ffmpeg: and there is no longer any need of forward declarations.
[23:38] <CIA-17> ffmpeg: 03Nicolas George 07master * reb9ba527e7 10ffmpeg/libavfilter/af_pan.c: 
[23:38] <CIA-17> ffmpeg: af_pan: remove the indirect function call.
[23:38] <CIA-17> ffmpeg: libswresample is used in all cases.
[00:00] --- Tue Feb 21 2012