[Ffmpeg-devel-irc] ffmpeg-devel.log.20151216

[burek]

[00:00:53 CET] <ubitux> i need someone to explain wth is going on when i'm doing ffmpeg -f lavfi -i testsrc=d=5 -vf scale=1000x100 out.mkv
[00:01:20 CET] <ubitux> is my brain malfunctioning hard tonight or the scaling isn't doing at all what we expect?
[00:02:42 CET] <ubitux> force_original_aspect_ratio is supposed to be disabled
[00:06:12 CET] <ubitux> ah dammit sar
[00:06:18 CET] <ubitux> that's evil
[00:11:02 CET] <rcombs> I have force_original_aspect_ratio hacked to mean "force an output SAR as close to 1:1 as possible by scaling the width or height downward by the input SAR"
[00:12:09 CET] <ubitux> it doesn't actually apply all the time
[00:13:06 CET] <ubitux> like, if the input has no sar set (0/1) it works as expected
[00:13:32 CET] <ubitux> but this is just evil:
[00:13:34 CET] <ubitux> [Parsed_scale_0 @ 0x28d55e0] w:320 h:240 fmt:rgb24 sar:1/1 -> w:1000 h:100 fmt:yuv444p sar:2/15 flags:0x4
[00:13:53 CET] <ubitux> compared to, let's say:
[00:13:54 CET] <ubitux> [Parsed_scale_0 @ 0x3b1f560] w:1920 h:1080 fmt:yuv420p sar:0/1 -> w:1000 h:100 fmt:yuv420p sar:0/1 flags:0x4
[00:15:02 CET] <ubitux> anyway, adding setsar=0/1 after testsrc "fixes" it
[00:27:45 CET] <Compn> durandal_170 : i've posted it to trac, http://trac.mplayerhq.hu/ticket/2261#ticket
[00:27:51 CET] <Compn> maybe beastd will apply it :P
[00:27:54 CET] Action: Compn runs
[00:57:47 CET] <michaelni> ubitux, the use case is/was that data could be sent to swscale sooner while its stll in L2 cache and the source could even reuse memory for a slice not requiring a whole frame
[01:53:04 CET] <J_Darnley> :)  Best way to reduce the memory usage of a browser:  keep using it till it crashes.
[04:47:57 CET] <cone-196> ffmpeg 03Ganesh Ajjanagadde 07master:65877ab935ad: lavc: typo fix uncliped -> unclipped
[09:32:43 CET] <j-b> 'morning
[10:12:44 CET] <ubitux> michaelni any idea how i can trigger "exotic" (as in sliceh ` h) with the cli?
[10:15:47 CET] <michaelni> ubitux, possibly theres no way with just ffmpeg, maybe mplayer supports it
[10:17:40 CET] <michaelni> ubitux, but it should be rather easy to add support to libavfilter/vf_scale
[10:18:18 CET] <nevcairiel> i tried to use it in the past and it just blew up, but maybe I used it wrong, its not l ike its documented well, also it was years ago
[10:18:57 CET] <nevcairiel> (wanted to use it for slice threading scaling)
[10:25:20 CET] <michaelni> i dont think slice threading will work with most scalers, that would need some volunteer to improve the slicing code
[10:25:41 CET] <michaelni> the code was written before multi core cpus where common 
[11:07:25 CET] <cone-322> ffmpeg 03Stefano Sabatini 07master:5defa3a0b09c: doc/encoders/libopenh264: update documentation after recent changes
[11:07:25 CET] <cone-322> ffmpeg 03Stefano Sabatini 07master:6e891d51f464: lavc/libopenh264: apply minor options text consistency fixes
[13:07:36 CET] <cone-322> ffmpeg 03Paul B Mahol 07master:10b16aee1bde: avfilter/vf_stereo3d: get rid of ts_unit hack
[13:58:43 CET] <cone-322> ffmpeg 03Ganesh Ajjanagadde 07master:4fa6f09c2c02: lavu/eval: remove pow and exp2 for postfixes
[13:58:44 CET] <cone-322> ffmpeg 03Ganesh Ajjanagadde 07master:83a04f103d38: lavc: move exp2fi to ff_exp2fi in internal.h
[14:04:56 CET] <cone-322> ffmpeg 03Michael Niedermayer 07master:30d770ca4481: ffmpeg: Allow specifying the program number for created programs
[14:04:57 CET] <cone-322> ffmpeg 03Vicente Olivert Riera 07master:e8bc43fe4730: mips: i6400 core is not mips32r2, so disable mips32r2
[14:04:58 CET] <cone-322> ffmpeg 03Vicente Olivert Riera 07master:a2f2abc88920: mips: add mips64r2 architecture variant
[14:44:52 CET] <mateo`> michaelni: hello, i would like to setup a qemu-arm environment like you did for http://fate.ffmpeg.org/report.cgi?time=20151210051641&slot=aarch64-linux-qemu-ubuntu-gcc-4.8-n2.4 (but for armv7), would you have some documentation/ressources that explain how to setup the env sysroot ?
[15:03:29 CET] <cone-322> ffmpeg 03Paul B Mahol 07master:fa2c1eab9555: avfilter/af_sofalizer: use SIMD in compensate_volume()
[15:55:22 CET] <kayrus> hey. did someone successfully run ffmpeg qsv on top of linux?
[15:56:02 CET] <kayrus> for some reason it stucks with the lots of "sched_yield()                           = " in strace
[15:59:09 CET] <nevcairiel> qsv generally likes to crap out like that
[16:03:21 CET] <saste> kayrus, I got it working on windows, on linux for me it fails with MFX_VideoENCODE_GetVideoParam() or something
[16:03:29 CET] <saste> but decoding seems to work fine
[16:14:08 CET] <kayrus> saste: I need encoding
[16:16:41 CET] <kierank> so when is ffmpeg going to grow a pair and talk to these people repeatedly breaking qsv
[16:16:44 CET] <kierank> when it worked in libav
[16:20:38 CET] <kayrus> does qsv work in libav?
[16:27:10 CET] <kayrus> I tried to follow this instructions (except kernel patch, as I use 3.19 + trusty) https://ffmpeg.org/pipermail/ffmpeg-user/2015-October/028761.html
[16:27:17 CET] <kayrus> these*
[16:29:25 CET] <j-b> it works in libav
[16:29:30 CET] <j-b> because it's actually tested.
[16:46:30 CET] <nevcairiel> j-b: for certain definitions of works, mediasdk is unstable crap, independent of implementation
[16:51:12 CET] <kayrus> crap. same "sched_yield()                           = 0" with avconc
[16:51:15 CET] <kayrus> avconv
[16:51:31 CET] <nevcairiel> its a driver bug
[16:51:34 CET] <nevcairiel> complain to intel
[16:51:40 CET] <nevcairiel> their shit is broken, is all
[16:52:36 CET] <Mavrik> ok it might sound like a silly question, but why do I have to take time_base from a filter_graph input not output when retrieving a filtered frame?
[16:53:15 CET] <durandal_1707> both
[16:54:08 CET] <Mavrik> (I'm talking about buffersink_ctx->inputs[0]->time_base call - since I'm noticing that fps filter does something funny with pts)
[17:36:45 CET] <durandal_1707> Mavrik: it changes output fps and timebase
[17:56:31 CET] <michaelni> mateo`, i dont remember setting anything special up, its just the aarch binutils/gcc/cpp packages from ubuntu and some arm64-cross packages i see too i assume tehy are all from  ubuntu
[18:18:30 CET] <Daemon404> can anyone fill me on on this QT pallette thing
[18:18:38 CET] <Daemon404> is this somethign specific to only mplayer?
[18:20:40 CET] <nevcairiel> I just know that this guy is nuts
[18:26:53 CET] Action: Daemon404 should change his title to Official Debbie Downer
[19:06:18 CET] <cone-322> ffmpeg 03Stefano Sabatini 07master:91c2a33d01d8: ffprobe: use print_val to print the frame pkt_size value
[19:34:38 CET] <cone-322> ffmpeg 03Andreas Cadhalpun 07master:f61d44b74aaa: opus_silk: fix typo causing overflow in silk_stabilize_lsf
[20:26:24 CET] <kierank> oh well trolled that guy back
[20:28:35 CET] <llogan> that's the second "i won't use git" this week.
[20:29:14 CET] <llogan> i can understand that it is complicated, but they don't seem to ever try
[20:30:05 CET] <J_Darnley> We only just tolerate that from Carl
[20:30:16 CET] <J_Darnley> He atleast manages to send unified diffs
[20:30:23 CET] <llogan> i'm not sure how he makes his patches.
[20:30:50 CET] <J_Darnley> The proprietary thing Linux used years ago?  Proforce?
[20:40:46 CET] <llogan> i wonder why claws-mail decided to refetch everything just now (IMAP). waiting...
[20:40:59 CET] <Mavrik> Perforce?
[20:41:20 CET] <Daemon404> J_Darnley, no, bitkeeper
[20:41:29 CET] <Daemon404> perforce is enterprise stuff
[20:42:09 CET] <J_Darnley> Ah.  Close enough.
[20:45:04 CET] <Daemon404> [19:29] <@llogan> i can understand that it is complicated, but they don't seem to ever try <-- "i didnt even bother to see what git is so i just attached a tarball with patches"
[20:45:08 CET] <Daemon404> is how i read it
[20:55:44 CET] <durandal_170> cbsrobot_: I managed to get output from jnoisemeter, you need to use qjackctl to connect ports
[21:00:12 CET] <cbsrobot_> durandal_170: yeah - I got it working too
[21:01:31 CET] <durandal_170> and?
[21:02:07 CET] <llogan> "sorry, but your 4.5 MB attachment is too big for this mailing list"
[21:03:21 CET] <cbsrobot_> durandal_170: I wanted to confirm the output of the filter with a  sounddesigner
[21:04:08 CET] <cbsrobot_> but I need to find some time to go over some samples with him
[21:07:18 CET] <durandal_170> well output was linear, I changed it now
[21:08:22 CET] <durandal_170> so it should be same as from jnoisemeter when both itu buttons are pressed
[22:32:26 CET] <kierank> hmmm so how do I get patches from gmail into git am
[22:33:11 CET] <atomnuker> evolution does a nice job
[22:34:00 CET] <atomnuker> well, it's basically the only thing I found which actually exports mbox files
[22:35:32 CET] <J_Darnley> kierank: I you're using the web interface then I think you want to click the "view original" link
[22:36:16 CET] <J_Darnley> I don't know where that might have gone to in the new design
[22:45:31 CET] <kierank> atomnuker: have you got that file that doesn't play
[22:46:58 CET] <kierank> vc2hq file
[22:47:04 CET] <atomnuker> HELL.vc2?
[22:47:07 CET] <kierank> yeah
[22:47:10 CET] <atomnuker> yeah, give me a sec
[22:47:47 CET] <atomnuker> correction, an hour, it's 223 megs and I have 1 Mbps upload
[22:47:57 CET] <atomnuker> I'll just encode something quick
[22:48:34 CET] <llogan> atomnuker: welcome to the shitty ISP club.
[22:49:02 CET] <atomnuker> WPA2 was the death of fast & free internet
[22:49:55 CET] <atomnuker> correction, it was WPS flood prevention
[22:55:04 CET] <kierank> atomnuker: what patches are missing, just hq profile?
[22:55:19 CET] <atomnuker> 06/11 and that one
[22:56:46 CET] <kierank> ah rebase told me =p
[23:00:19 CET] <atomnuker> kierank: https://0x0.st/-Cp.vc2
[23:00:34 CET] <atomnuker> 5 frames, legall, 3 levels, 1280x720
[23:01:19 CET] <kierank> woah
[23:01:37 CET] <llogan> atomnuker: claws-mail can also "export (selected) to mbox file"
[23:06:21 CET] <kierank> atomnuker: can you send me a jpeg of what it should look like
[23:06:59 CET] <atomnuker> uploading the raw right now
[23:07:04 CET] <atomnuker> 12 megabytes
[23:07:47 CET] <Daemon404> sounds like BT internet
[23:09:00 CET] <atomnuker> correct
[23:09:03 CET] <atomnuker> modem
[23:10:50 CET] Action: JEEB just learned that he needs moar dakka with his wireless
[23:10:58 CET] <Daemon404> man
[23:11:08 CET] <Daemon404> i dont have teh energy to agrue with that suse guy
[23:11:11 CET] <Daemon404> argue*
[23:11:16 CET] <Daemon404> i dotn get his beef
[23:11:24 CET] <Daemon404> "we dont wrong thinsg elsewhere so lets do it here, for no gain" ?
[23:11:44 CET] <Mavrik> JEEB, 802.11ac ftw :)
[23:12:47 CET] <atomnuker> kierank: https://0x0.st/-CO.yuv
[23:13:04 CET] <kierank> what pixel format
[23:13:05 CET] <JEEB> Mavrik: I had to actually debug channels tonight as this computer's networking is getting unbearable. switched to another channel for now but think of grabbing http://geizhals.eu/linksys-wrt1900acs-wrt1900acs-eu-wrt1900acs-uk-a1336073.html
[23:13:21 CET] <Mavrik> Yeah, that's a beast.
[23:13:34 CET] <atomnuker> kierank: oh crap, I think I forgot to set the pix_fmt
[23:13:49 CET] <Mavrik> I have AC66U and I'm getting 400Mbit or so over wifi which is more than enough to do TV streaming tests
[23:13:50 CET] <atomnuker> so it's yuv420p 8bit
[23:13:56 CET] <kierank> can you output yuv422p10le so I can md5sum
[23:14:04 CET] <kierank> and gzip it please :)
[23:14:07 CET] <atomnuker> pointless, I need to reencode
[23:14:47 CET] <JEEB> Mavrik: it's funny how quickly this stuff can change, didn't have any issues like 2 months ago and now the 2.4GHz area is getting filled
[23:14:48 CET] <J_Darnley> If you're thinking that new linksys router will be as good or useful as the ancient wrt54gl, then be careful.  Last I heard the source wasn't being made available yet.
[23:14:59 CET] <kierank> atomnuker: ok let me know when you have a new file
[23:15:07 CET] <Mavrik> JEEB, in my case it was bunch of bluetooth and other crap on 2.4GHz not wifis
[23:15:17 CET] <Mavrik> Xbox controllers and etc., anything on that band is useless :/
[23:15:19 CET] <atomnuker> kierank: huh, it works now
[23:15:25 CET] <atomnuker> in ffmpeg with the last 2 patches
[23:15:34 CET] <JEEB> J_Darnley: there seems to be an open source wifi driver now for it IIRC
[23:15:45 CET] <kierank> atomnuker: huh
[23:15:49 CET] <kierank> my view looks bad
[23:16:14 CET] <J_Darnley> JEEB: typical.  Bad news gets spread around but good news or corrections don't.
[23:16:49 CET] <atomnuker> kierank: yes, the vc2hqencode is stupid
[23:16:51 CET] <Mavrik> It's not a good idea to install 3rd party stuff on these anyways.
[23:16:58 CET] <Mavrik> 802.11ac is really flakey when it comes to configuration.
[23:16:59 CET] <kierank> atomnuker: what did you do...
[23:17:02 CET] <atomnuker> one more reason why to always use y4m to handle raw video
[23:17:19 CET] <Mavrik> And both on my ASUS and on the Linksys I lost like 50% of throughput on non-original firmware. Not worth it.
[23:17:20 CET] <atomnuker> kierank: I outputted yuv422p10le and fed that into the encoder
[23:17:40 CET] <atomnuker> the output vc2 file decodes correctly
[23:18:07 CET] <kierank> I am going to push then
[23:19:16 CET] <atomnuker> yeah, adding a fate test and a sample would be nice
[23:19:21 CET] <JEEB> Mavrik: I'd only flash something else in case abs. required
[23:20:04 CET] <atomnuker> kierank: though the fate test and sample can wait for a proper encoder too
[23:21:10 CET] <JEEB> Mavrik: any experience with stuff like http://www.tp-link.fi/products/details/cat-15_Archer-VR200.html#specifications ?
[23:22:12 CET] <Mavrik> JEEB, not personally, but I've heard great things about C7
[23:22:31 CET] <Mavrik> JEEB, C7, Linksys WRT1900AC and ASUS AC68U seem to be cream of the crop right now in AC1900 land.
[23:22:39 CET] <Mavrik> There's also UniFi if you need more APs which has this awesome management.
[23:23:44 CET] <durandal_170> atomnuker: so you had time to test sofalizer?
[23:24:14 CET] <llogan> durandal_170: is there a sample in fate you recommend to give itur468 a try?
[23:25:54 CET] <kierank> hmmm why can't I push
[23:26:11 CET] <llogan> durandal_170: because whatever input i use gives no metadata </tard>
[23:27:50 CET] <durandal_170> it should always return something, what you used?
[23:28:39 CET] <cone-322> ffmpeg 03Andreas Cadhalpun 07merge2:HEAD: opus_silk: fix typo causing overflow in silk_stabilize_lsf
[23:28:45 CET] <durandal_170> michaelni: why kierank can't push?
[23:28:45 CET] <kierank> fuck
[23:29:41 CET] <llogan> durandal_170: ./ffmpeg -i ../fate-suite/audio-reference/chorusnoise_2ch_44kHz_s16.wav -filter_complex itur468 -f null -
[23:29:58 CET] <atomnuker> kierank: https://0x0.st/-C4.vc2 should work fine
[23:30:02 CET] <kierank> damn won't let me delete remote branch
[23:30:18 CET] <kierank> kieran at ubuntu:~/ffmpeg$ git push origin master
[23:30:18 CET] <kierank> Everything up-to-date
[23:30:27 CET] <kierank> I don't understand
[23:30:58 CET] <Mavrik> git push origin :branch_name ?
[23:31:03 CET] <Mavrik> To delete remote branch :)
[23:31:29 CET] <kierank> To git at source.ffmpeg.org:ffmpeg
[23:31:29 CET] <kierank>  ! [remote rejected] merge2 (hook declined)
[23:31:29 CET] <kierank> error: failed to push some refs to 'git at source.ffmpeg.org:ffmpeg'
[23:31:56 CET] <durandal_170> add adrawgraph-lavfi.itur458.1.noise:min--90:max-9
[23:32:14 CET] <jamrial> kierank: ask j-b to delete it
[23:32:47 CET] <durandal_170> llogan ' like that
[23:32:48 CET] <kierank> I don't understand why push origin master dones't work
[23:32:58 CET] <atomnuker> kierank: git pull --rebase?
[23:33:18 CET] <kierank> kieran at ubuntu:~/ffmpeg$ git pull --rebase origin master
[23:33:18 CET] <kierank> From source.ffmpeg.org:ffmpeg
[23:33:18 CET] <kierank>  * branch            master     -> FETCH_HEAD
[23:33:18 CET] <kierank> Current branch merge2 is up to date.
[23:33:19 CET] <durandal_170> push branch:master
[23:33:34 CET] <jamrial> the hooks prevent committers to remove branches or tags to avoid breaking the public repo
[23:33:46 CET] <jamrial> you need to ask j-b to do it
[23:33:53 CET] <kierank> asked in #videolan
[23:34:34 CET] <kierank> git push merge2:master?
[23:34:47 CET] <kierank> oh
[23:34:48 CET] <kierank> origin
[23:34:57 CET] <cone-322> ffmpeg 03Kieran Kunhya 07master:037960363249: diracdec: Add 10-bits to pix_fmt table
[23:34:58 CET] <cone-322> ffmpeg 03Kieran Kunhya 07master:a349a10edf84: diracdec: Add support for HQ profile
[23:35:02 CET] <kierank> ok phew
[23:35:34 CET] <durandal_170> llogan: it have typos, im on phone
[23:40:27 CET] <llogan> durandal_170: ok, got it. i forgot about adrawgraph. i was expecting metadata showing in the console output.
[23:42:23 CET] <durandal_170> use ffprobe for that
[00:00:00 CET] --- Thu Dec 17 2015