[Ffmpeg-devel-irc] ffmpeg-devel.log.20130429

[burek]

[01:39] <llogan> michaelni: can you check the logs for a reason why robert at videorx.com is getting 550?
[01:40] <michaelni> llogan, i already fixed it i think
[01:41] <llogan> what was the problem?
[01:41] <llogan> more blacklist misbehaving?
[01:42] <michaelni> yes, but it should be permanently fixed now
[01:42] <llogan> did you reply to his message?
[01:42] <michaelni> yes 
[01:42] <michaelni> my mutt says "r"
[01:42] <michaelni> must have fogoteen to cc owner
[01:43] <llogan> no problem. thanks for looking into it. i was travelling for 10 days, and then i leave again tomorrow for 3 more.
[03:28] <cone-737> ffmpeg.git 03Michael Niedermayer 07master:b691bc424f53: copy_packet_data: fix memleak
[10:54] <durandal_1707> michaelni: both interlace,tinterlace, il and maybe others show green bootom line when height is multiple of 2 but not 4 with yuv420p (subsamples height)
[11:02] <nevcairiel> interlaced 4:2:0 kinda requires that he height is mod 4
[11:02] <nevcairiel> the*
[11:03] <durandal_1707> Stream #0:0: Video: vp8, yuv420p, 854x362, SAR 1:1 DAR 427:181, 
[11:50] <BuxiNess> hello, in that case : http://fate.ffmpeg.org/report.cgi?time=20130428054135&slot=x86_64-archlinux-gcc-valgrindundef memory poisoninh is enabled or disabled ?
[11:51] <ubitux> disabled
[11:51] <BuxiNess> thks
[11:51] <ubitux> (michaelni: thx for the side_data fix!)
[11:51] <durandal_1707> ubitux: http://www.hometheaterhifi.com/technical-articles-and-editorials/technical-articles-and-editorials/the-chroma-upsampling-error-and-the-420-interlaced-chroma-problem.html still relevant for us?
[11:51] <durandal_1707> i want to remove ilpack, its asserts here
[11:52] <ubitux> durandal_1707: tl;dr :(
[11:52] <ubitux> BuxiNess: --enable-memory-poisoning is the default for all fate instances, so i had to disable it by adding the flag at the end
[11:53] <durandal_1707> ubitux: well it mentions verticall low pass filter, it think we have that
[11:54] <ubitux> okay :p
[12:06] <ubitux> btw, any comment on dctdnoiz?
[12:07] <ubitux> and on an unrelated note, i may submit a lut 3d filter in the next days
[12:08] <durandal_1707> what it could do?
[12:10] <ubitux> color mapping
[12:10] <ubitux> you have table of 17x17x17 or 32x32x32, and you use them to get a new color
[12:11] <durandal_1707> so why is 3d lut better than 1d? lut we already have?
[12:11] <ubitux> see http://www.lightillusion.com/luts.html
[12:12] <ubitux> it seems post prod studios are using such things
[12:12] <ubitux> anyway, i'm working on it
[12:13] <durandal_1707> than do it with 16 bit colorspace too
[12:13] <ubitux> yep
[12:33] <cone-334> ffmpeg.git 03Michael Niedermayer 07master:e92862e75f31: avfilter/vf_tinterlace: fix handling of not so even sizes
[12:42] <durandal_1707> really? previously you needed -strict -2 anyway
[12:42] <durandal_1707> ubitux: http://www.gemmantics.com/spectrograms-in-c-with-libav-ffmpeg-libpng-and-sphinx/
[12:42] <michaelni> BuxiNess, see https://ffmpeg.org/trac/ffmpeg/ticket/2510
[12:43] <ubitux> durandal_1707: do you want to send a message to the guy
[12:43] <ubitux> ?
[12:44] <durandal_1707> i need twitter to send him message?
[12:45] <ubitux> i don't have twitter :(
[12:45] <durandal_1707> i mush finish avectorscope
[12:48] <BuxiNess> michaelni, I'm on it
[12:51] <durandal_1707> ubitux: but his outputs have bigger width
[12:52] <ubitux> bigger width?
[12:52] <durandal_1707> output is "streched"
[12:53] <durandal_1707> like it use smaller window for drawing pixels
[12:59] <ubitux> durandal_1707: why didn't you add rgb48 and rgba64 to ff_fill_rgba_map() in colorchannelmixer?
[12:59] <durandal_1707> because that function is part of ugly drawutils
[13:00] <durandal_1707> and i did not wanted to touch it
[13:00] <ubitux> :/
[13:00] <ubitux> it's a trivial change
[13:01] <durandal_1707> but drawutils does not support 16 bit at all, so its inconsistent
[13:01] <ubitux> ok..
[13:02] <BuxiNess> for  https://ffmpeg.org/trac/ffmpeg/ticket/2510 its an expected problem ( Non Dcinema files)
[13:08] <zimbatm> michaelni: do you mind taking a look at https://ffmpeg.org/trac/ffmpeg/ticket/2509 ?
[13:08] <zimbatm> it's a regression on the moov header sizes that you introduced in fc09bf57a60d4c4a6d339b204b3282337067c06d
[13:10] <zimbatm> just let me know if it's too fuzzy, i had trouble identifying the problem initially
[16:29] <durandal_1707> burek: what you mean with that scale video filter ?
[16:31] <burek> what do you mean what do i mean? :D
[16:32] <burek> i described it in detail there
[16:32] <burek> it's hard not to understand (imho)
[16:32] <durandal_1707> i don't get it
[16:32] <burek> what part?
[16:33] <durandal_1707> do you mean user set max allowed output size, while keeping aspect ratio?
[16:40] <durandal_1707> michaelni: af_pan gives nonsense warnings
[16:45] <ubitux> burek: you can use min() and max() in the expressions
[16:52] <burek> durandal_1707 yes
[16:52] <burek> ubitux can you provide an example cmd line?
[16:54] <ubitux> dunno if it works but maybe w='if(gt(w,h),640,-1)':h='if(gt(h,w),480,-1)'
[16:56] <burek> ubitux, how is that keeping aspect ratio?
[16:59] <burek> oh i think i see what you mean
[17:04] <burek> i tried this: -vf scale='if(gt(w,h),640,-1)':'if(gt(w,h),-1,480)'
[17:04] <burek> [Parsed_scale_0 @ 0x9d08c60] Invalid size 'if(gt(w'
[17:04] <burek> :/
[17:05] <ubitux> escape
[17:05] <ubitux> add "" around the whole thing
[17:12] <burek> like this: -vf scale="'if(gt(w,h),640,-1)':'if(gt(w,h),-1,480)'"
[17:12] <burek> ?
[17:13] <burek> [Parsed_scale_0 @ 0x996cc80] [Eval @ 0xffd7aac8] Undefined constant or missing '(' in 'w,h),640,-1)'
[17:13] <ubitux> maybe iw and ih instead
[17:14] <burek> oh i see
[17:14] <burek> -vf scale="'if(gt(iw,ih),640,-1)':'if(gt(iw,ih),-1,480)'"
[17:14] <burek> that appears to work :) great! :) thanks a lot! :)
[17:18] <burek> btw, should i put this into feature suggestion on bug tracker: http://ffmpeg.gusari.org/viewtopic.php?f=13&t=911 (something about quick sync)?
[17:19] <durandal_1707> put whatever you want
[19:15] <ubitux> saste: i'm doing the first part of the fate-run.sh merge
[19:15] <ubitux> but i'll likely need your help later
[20:00] <saste> ubitux, i'll have a look at it later
[20:03] <cone-334> ffmpeg.git 03Clément BSsch 07master:c703f7bdbd8f: fate: move more tests from lavfi-regression.sh to filter-video.mak.
[20:03] <cone-334> ffmpeg.git 03Clément BSsch 07master:80bcbed15747: fate: restore our settings for drawtext/fade/unsharp.
[20:22] <cone-334> ffmpeg.git 03Clément BSsch 07master:9a0e40ec8c0c: fate: move colormatrix tests to the new system.
[20:32] <cone-334> ffmpeg.git 03Carl Eugen Hoyos 07master:00a61ac7559b: Recognize j2k as file suffix for the image2 muxer.
[20:32] <cone-334> ffmpeg.git 03Michael Niedermayer 07master:affd5018ac35: Merge remote-tracking branch 'cehoyos/master'
[20:43] <ubitux> durandal_1707: michaelni is right, showspectrum + pad might not work
[20:44] <durandal_1707> how?
[20:45] <ubitux> pad could increase linesizes
[20:45] <ubitux> though, it's a init time here
[20:46] <durandal_1707> but this is internal frame....
[20:46] <ubitux> but i think it's better to use width anyway, generally speaking
[20:48] <durandal_1707> so just replacing linesize, with w * 3?
[20:48] <ubitux> yes
[20:52] <durandal_1707> ubitux: do you know that libquvi switchhed to AGPL?
[20:54] <ubitux> no, didn't follow
[20:54] <ubitux> do i need to do something in the configure?
[20:57] <durandal_1707> add agpl switch?
[20:57] <durandal_1707> isn't it more restrictive than gpl3?
[20:58] <JEEBsv> yes
[20:58] <JEEBsv> basically it was made for web apps and such
[20:58] <JEEBsv> so that you had to give the source even to the 'users' of your service
[20:58] <JEEBsv> I really don't know why the hell the quvi guy thought it was a good idea to make his non-web-app app AGPL
[21:00] <ubitux> durandal_1707: what was the old licence?
[21:01] <ubitux> lgpl no?
[21:01] <durandal_1707> LGPLv2.1
[21:06] <cehoyos> The user who opened ticket 2512 thought it is a good idea to add somebody to CC
[21:06] <durandal_1707> lol
[21:08] <ubitux> :D
[21:09] <durandal_1707> burek: that example in ticked have to much "'", only 2 are needed, while it have 4
[21:14] <durandal_1707> no direct path for fade
[21:15] <cehoyos> It is the same Google employee who wanted to add Ronald to CC on ticket 2364
[21:15] <cehoyos> (unsuccessfully)
[21:15] <cone-334> ffmpeg.git 03Clément BSsch 07master:22a86492ecd1: fate: move alphaextract and alphamerge tests to the new system.
[21:16] <durandal_1707> cehoyos: you have list of Google employers?
[21:17] <cehoyos> No, but I think google.com email adresses are an indication.
[21:20] <gnafu> cehoyos: Yeah, that's usually a safe bet.
[21:22] <cehoyos> nevcairiel: I fear that the user wants to report that --extra-cflags=-Werror fails compilation and that is not going to be fixed afaict
[21:22] <durandal_1707> so, nobody writing planeextract, or should it be named like that, or something similar?
[21:23] <nevcairiel> cehoyos: he wants to get rid of all the deprecation warnings in his build, so he wants to manually remove soem deprecated parts of the code by setting the deprecation define, and that breaks compilation
[21:24] <nevcairiel> taking his last comment in mind, all he wants is someone to be aware of this :p
[21:25] <nevcairiel> be happy he reports a potential future problem? :)
[21:25] <cehoyos> I would be much, much happier if he reported current problems.
[21:26] <durandal_1707> or sent patch
[21:26] <cehoyos> (It is somehow hard to believe that the only wish Chromium has wrt FFmpeg is that it compiles with -Werror)
[21:27] <nevcairiel> different projects have different priorities
[21:27] <Plorkyeran> they have already fixed higher priority things like being able to compile it with msvc
[21:27] <nevcairiel> i bet its not the only wish, but its one =p
[21:27] <nevcairiel> i dont think msvc even honors those deprecation flags =P
[21:28] <nevcairiel> but maybe there is a special option to turn that on
[21:28] Action: nevcairiel never checked
[21:28] <cehoyos> If that is their highest priority, I would love to learn about the "wishes"
[21:28] <Plorkyeran> it has a different deprecation attribute syntax
[21:31] <Plorkyeran> deprecating things that ffmpeg itself still depends on it kind of silly
[21:32] <cone-334> ffmpeg.git 03Paul B Mahol 07master:32d464121959: lavfi/showspectrum: fix crash in case of negative linesize
[21:32] <cone-334> ffmpeg.git 03Paul B Mahol 07master:c645659988af: lavfi/showwaves: fix crash in case of negative linesize
[21:37] <ubitux> durandal_1707: outlink->w * 3?
[21:37] <ubitux> ah no, sorry, it's not rgb
[21:37] <durandal_1707> :/
[21:46] <durandal_1707> so no vfr is possible in lavfi?
[21:49] <cone-334> ffmpeg.git 03Clément BSsch 07master:6ea013fb62c9: fate: remove '#' in filter parameters.
[21:49] <cone-334> ffmpeg.git 03Clément BSsch 07master:3ae44df18f57: fate: move lavd filters tests to the new system.
[21:57] <ubitux> i think i'll need saste for the pixfmt thing
[21:57] <ubitux> it seems we have several improvement to the pixfmt tests...
[21:58] <durandal11707> like what?
[21:59] <ubitux> it seems we filter more pix fmts
[21:59] <durandal11707> and?
[21:59] <ubitux> dunno, the function is relatively different
[21:59] <ubitux> i'm going to check every test one by one
[21:59] <ubitux> here we go.
[22:11] <cone-334> ffmpeg.git 03Diego Biurrun 07master:20784aa6781a: x86: hpeldsp: Remove unused macro definitions
[22:11] <cone-334> ffmpeg.git 03Michael Niedermayer 07master:16b2472d209d: Merge remote-tracking branch 'qatar/master'
[22:19] <nevcairiel> ubitux: some of the changes seem to have broken msvc fate
[22:20] <ubitux> mmh
[22:20] <nevcairiel> http://fate.ffmpeg.org/report.cgi?time=20130429195243&slot=x86_32-msvc10-windows-native
[22:21] <nevcairiel> specifically overlay and alphaextract/merge
[22:21] <ubitux> i'll have a look in a moment
[22:22] <ubitux> i guess it's a sws flags thing again
[22:22] <nevcairiel> all the sws inline assembly doesnt run on msvc, so its pure C more or less
[22:25] <cone-334> ffmpeg.git 03Clément BSsch 07master:15d6436a2a0c: fate: move a few simple pixfmts tests to the new system.
[22:25] <cone-334> ffmpeg.git 03Clément BSsch 07master:9a5924f86254: fate: add filters dependencies to pixfmts tests.
[22:28] <ubitux> overlay_rgb has the flag though..
[22:29] <cone-334> ffmpeg.git 03Clément BSsch 07master:6b222ea9637a: fate: add bitexact flags to alphaextract/alphamerge tests.
[22:29] <durandal11707> ????
[22:30] <durandal11707> that tests should not do any conversion...
[22:30] <ubitux> which one?
[22:30] <durandal11707> extract/merge
[22:31] <ubitux> there is a format request
[22:31] <ubitux> so input is rescaled in case of rgb
[22:31] <ubitux> (pgmyuv source)
[22:31] <nevcairiel> it converts from the yuv420p input to bgra somewhere
[22:31] <durandal11707> sick
[22:32] <nevcairiel> and there is the second box finished :p
[22:41] <ubitux> i don't understand why overlay_rgb fails
[22:41] <durandal11707> michaelni: nut format does not show duration any more here
[22:41] <ubitux> :(
[22:42] <nevcairiel> ubitux: the output images look fine, i think, not quite sure how its supposed to look =P
[22:42] <ubitux> nevcairiel: does it change something if you do that: 
[22:42] <ubitux> -fate-filter-overlay_rgb: CMD = framecrc -c:v pgmyuv -i $(SRC) -filter_complex_script $(SRC_PATH)/tests/filtergraphs/overlay_rgb
[22:42] <ubitux> +fate-filter-overlay_rgb: CMD = framecrc -c:v pgmyuv -i $(SRC) -sws_flags +accurate_rnd+bitexact -flags +bitexact -filter_complex_script $(SRC_PATH)/tests/filtergraphs/overlay_rgb
[22:42] <ubitux> ?
[22:42] <ubitux> (no change here, which is expected)
[22:43] <nevcairiel> where is that?
[22:43] <ubitux> tests/fate/filter-video.mak
[22:43] <ubitux> want a proper diff?
[22:43] <nevcairiel> this is fine
[22:43] <ubitux> http://b.pkh.me/0001-msvc-fix-test.patch
[22:43] <ubitux> ok :)
[22:44] <nevcairiel> no change
[22:45] <ubitux> :(
[22:45] <nevcairiel> if i take out the sws_flags from the filter complex script, the hashes change
[22:45] <nevcairiel> so it is applied, apparently
[22:46] <ubitux> fate-filter-alphaextract_alphamerge_rgb is fixed btw?
[22:46] <nevcairiel> want the resulting images from that test as png to compare? :p
[22:46] <ubitux> yeah why not :)
[22:47] <nevcairiel> alphaextract still fails
[22:47] <ubitux> T_T
[22:47] <michaelni> ubitux, did you look at make fate-list before and after todays commits ?
[22:48] <michaelni> i just run it through wc
[22:48] <nevcairiel> oh yeah, the number of tests run seem to have gone down by 6 or so
[22:48] <michaelni> and it seems there are 6 lines less
[22:48] <ubitux> i've removed/merged some tests
[22:48] <ubitux> some tests were present two times
[22:48] <ubitux> (drawbox, unsharp, ...)
[22:49] <nevcairiel> ubitux: http://files.1f0.de/test1.png
[22:49] <michaelni> ubitux, ahh then it should be ok
[22:49] <nevcairiel> this is the first frame of the overlay test
[22:49] <ubitux> nevcairiel: ok, thx
[22:52] <ubitux> http://b.pkh.me/diff-overlay.png
[22:52] <ubitux> (./ffmpeg -i test1.png -i local.png -filter_complex blend=all_mode=difference diff-overlay.png)
[22:52] <ubitux> :(
[22:52] <nevcairiel> this is odd
[22:53] <ubitux> http://b.pkh.me/local.png
[22:53] <ubitux> generated with: ./ffmpeg -nostats -cpuflags all -c:v pgmyuv -threads 1 -thread_type frame+slice -i /home/ubitux/src/ffmpeg/tests/vsynth1/%02d.pgm -filter_complex_script ./tests/filtergraphs/overlay_rgb -frames:v 1 local.png
[22:53] <nevcairiel> i can see the differences even with my eyes, though
[22:54] <ubitux> yeah, like a wrong shift or something :p
[22:54] <nevcairiel> yeah like a chroma shift
[22:54] <ubitux> i wonder why the issue didn't show up previously...
[22:55] <durandal11707> michaelni: if i suspend muxing before it finished, duration is displayed with ffprobe
[22:56] <ubitux> maybe because i removed the explicit fifo?
[22:57] <nevcairiel> still has to be a bug somewhere if it gives us different results on different systems, right?
[22:57] <ubitux> removing the fifo may have triggered some different code path
[22:57] <ubitux> typically with the get buffer thing 
[22:57] <ubitux> nevcairiel: http://b.pkh.me/0001-fifo.patch
[22:57] <ubitux> can you try if this changes anything?
[23:00] <michaelni> "<durandal_1707> michaelni: af_pan gives nonsense warnings", how can i reproduce it ?
[23:01] <nevcairiel> ubitux: no changes
[23:01] <ubitux> :(
[23:01] <durandal11707> michaelni: the '|' warning
[23:02] <durandal11707> ubitux: what new system does differently from older one?
[23:02] <ubitux> usage of -filter_complex_script instead of -vf maybe
[23:03] <ubitux> also, previous was just a final md5, but it should have mismatch with such change
[23:03] <durandal11707> michaelni: i don't get why that warning happens with: -af pan=stereo:
[23:03] <ubitux> there was the fifo but it's not this according the the last test
[23:11] <durandal11707> so how filter that extract planes from planar formats should be named? (extended alphaextract)
[23:17] <durandal11707> planesplit
[00:00] --- Tue Apr 30 2013