[Ffmpeg-devel-irc] ffmpeg-devel.log.20131023

[burek]

[00:00] --- Wed Oct 23 2013
[00:35] <burek> is there any interest to implement some kind of a remote control for ffplay
[00:36] <burek> in a way, for example, to make it listen to a tcp connection and get the playback commands out of there
[00:36] <burek> rather than from the keyboard
[00:37] <llogan> burek: what's the use case?
[00:37] <burek> i assume you didn't read #ffmpeg but for example
[00:37] <burek> multiple players could be synced remotely
[00:37] <burek> to play the content in sync
[00:37] <burek> just by simple seeking
[00:38] <burek> and btw, vlc has got that feature with a lot of different types of controllers (telnet, http, rc, etc) and its quite cool and useful
[00:39] <llogan> ah. i didn't see #ffmpeg yet...
[00:39] <burek> it's a long conv
[00:39] <burek> in short, multiple players synced from 1 point to display the content in sync
[02:38] <Ruler2112> Hello.  I have a suggestion about the elimination of a restriction present in ffmpeg.
[02:40] <Ruler2112> I wrote a script many years ago to standardize digital video into a format that my TV can reliably play.  It uses several tools, ffmpeg and virtualdub among them.  Driving VirtualDub via command line, I set the ISFT metadata to be the name of my script.  (This way, I could tell right away which have been processed & which have not.)
[02:42] <Ruler2112> I'm currently re-working the script and eliminating dependencies.  I was hoping to do everything with ffmpeg, but the ISFT (encoder/encoded with) flag is locked to be Lavf####.  I would like to suggest that this field default to the string denoting the version of Lib AV Format, but allow it to be overridden by setting the command line parameter at the invocation of ffmpeg.
[02:47] <Ruler2112> I do have an alternate plan to change this metadata field after the file has been generated, but it would be much more streamlined to simply do it via ffmpeg.  (Plus, though I've nearly eliminated VirtualDub as a dependency, I'll need to add another dependency to change the metadata field.)  Having it locked to a set value when other tools allow it to be changed is just silly IMO.
[02:52] <cone-487> ffmpeg.git 03Michael Niedermayer 07master:60ea0a5236ce: avcodec/sgi: Fix dereferencing uninitialized pointer
[02:52] <Ruler2112> join #ffmpeg
[03:42] <BBB> ubitux: I think that's the same as reported ~1 week ago, it's sse2 instructions in emulated_edge_mc
[03:42] <BBB> I sent a patch for that but it wasn't applied yet
[03:42] <BBB> I haven't been able to look at it since then
[08:30] <wm4> so libswscale calls mprotect() to make mmap'ed memory executable for god knows what reasons
[08:31] <wm4> but it fails to check the mprotect() return value, so it'll just crash on "hardened" systems which do not allow self-modifying code
[08:41] <Zeranoe> Is there any way to compile FFmpeg without TLS (thread-local-storage)?
[09:31] <michaelni_> wm4, sws its used for the fast bilinear scaler which generates at runtime some optimized code, ill look at the return checks unless someone else is quicker
[09:32] <wm4> isn't fast bilinear said to be broken anyway?
[09:33] <michaelni_> last i tried it worked
[09:35] <michaelni_> Zeranoe, iam not sure atm where we use thread locals but --disable-pthreads or something like that would be the sledge hammer solution to avoid that
[09:35] <michaelni_> what problem do these thread locals cause ?
[09:37] <cone-490> ffmpeg.git 03Derek Buitenhuis 07master:3511d4fc9784: pthread: Avoid crashes/odd behavior caused by spurious wakeups
[09:37] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:d0f0ac426bdf: Merge commit '3511d4fc9784d5fbb024dce68ca7a0d7fdd74663'
[09:38] <Zeranoe> michaelni_: Just someone asking for a version without it, I'm sure everything is fine
[09:54] <cone-490> ffmpeg.git 03Derek Buitenhuis 07master:b1fcdc08ceb5: nut: Fix unchecked allocations
[09:54] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:078abccdd6dc: Merge commit 'b1fcdc08ceb5df69fac34aa0d57c56905d32b8b4'
[10:00] <cone-490> ffmpeg.git 03Alex Converse 07master:f457edb69794: aacdec: Use avpriv_report_missing_feature() instead of custom logging.
[10:00] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:e0f291822306: Merge commit 'f457edb69794c7e13b3630456d72f7942fff19f7'
[10:11] <cone-490> ffmpeg.git 03Alex Converse 07master:b3be41ca8252: aac: Add support for Enhanced AAC Low Delay (ER AAC ELD).
[10:11] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:fdb48225592c: Merge remote-tracking branch 'qatar/master'
[10:15] <durandal_1707> michaelni_: why you reverted my change to sgidec?
[10:18] <durandal_1707> actually you should revert my change instead....
[10:18] <durandal_1707> of adding another commit
[10:19] <durandal_1707> and this case looks to not be properly covered at all
[10:19] <durandal_1707> otherwise fate would be yellow
[10:20] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:eb4205cc8957: swscale/utils: check mprotect() return code
[10:22] <michaelni_> durandal_1707, i also realized that fate doesnt cover it but out encoder can generate files that trigger the issue
[10:22] <michaelni_> needs -coder 2 IIRC
[10:23] <durandal_1707> anyway that raw sgi should use planar rgb
[10:32] <durandal_1707> michaelni_: what happened to crash with unused pad i reported?
[10:40] <durandal_1707> michaelni_: something fisshy happened in aac merge
[10:41] <durandal_1707> i have big warning here
[10:41] <durandal_1707> lol it from libav
[10:47] <michaelni_> durandal_1707, "unused pad" ? which ticket or mail was that or how do i reproduce ?
[10:48] <durandal_1707> ffmpeg -f lavfi -i color -f lavfi -i color -filter_complex "[1]null[x],[0][1]overlay" -f null -
[10:48] <durandal_1707> do you get crash?
[10:51] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:72cf47b241c7: avcodec/aacdec: fix avpriv_report_missing_feature() arguments
[10:51] <durandal_1707> AAAAAAAAAAAAAAAAAAWW
[10:52] <durandal_1707> i had same commit
[10:52] <durandal_1707> next time i will not share my precious findings to anyone
[10:54] <michaelni> durandal_1707, yep crashes
[10:55] <durandal_1707> michaelni: lol you left 3 unfixed
[10:58] Action: durandal_1707 oh i shared it again
[10:58] <durandal_1707> michaelni: no, i'm not going to fix it as i removed my version
[11:09] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:435a730e2190: avcodec/aacdec: fix more avpriv_report_missing_feature() arguments
[11:24] <j-b> 'morning
[11:36] <EKRAM> Greetings to all people show makes and contributes  such a wonderful piece of software that makes wonder 
[11:36] <EKRAM> i have question regarding the Distributed Encoding of a single Videos using ffmpeg
[11:37] <EKRAM> so that i can use spread a big single file over different encoders and finally stitch them togather 
[11:37] <EKRAM> is it possible or would you give me a suggestions how can i achive it
[11:43] <j-b> mraulet: ever heard of hevc in avi and flv?
[11:43] <mraulet> no
[11:43] <mraulet> I am a crap mpeg guy :-)
[11:43] <j-b> mraulet: well, now, you do :)
[11:44] <mraulet> where?
[11:45] <j-b> uploading now
[11:48] <durandal_1707> someone just hacked ffmpeg
[11:52] <JEEB> note: the hevc-in-flv shit has NOTHING to do adobe
[11:52] <JEEB> it's a random chinese corp hacking gabest's dshow muxer
[11:52] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:b36257921e52: ffmpeg: Dont crash on unconnected output pads before an error is printed
[11:53] <JEEB> *to do with adobe
[11:54] <JEEB> and the avi samples are just random people taking the dshow flv demuxer's output and pushing it into some dshow avi muxer >_>
[11:55] Action: durandal_1707 bans random people
[11:56] Action: michaelni thinks 90% of avi is from random people pushing random things in it
[11:56] <JEEB> also as I noted elsewhere, surprised j-b didn't get a sample of hevc-in-asf yet :P that crap is already flowing through doom9 because someone decided it was a good idea to remux avi via dshow to asf
[11:56] <JEEB> thankfully, the flv/avi/asf crowd seems to be just a single person
[11:56] <JEEB> or so
[11:58] <durandal_1707> i cant get this: geq=r=X+128 + 30*sin(2*(PI/200)*((Y-50*T))):g=X + 128 + 30*sin(2*(PI/200)*((Y-50*T))):b=X + 128 + 30*sin(2*(PI/200)*((Y-50*T)))
[11:58] <durandal_1707> to be continuous, eg not change from 0 -> 255
[12:09] <Daemon404> [10:56]  * michaelni thinks 90% of avi is from random people pushing random things in it <-- has anyone managed to mux an avi stream withn an avi yet?
[12:10] <Daemon404> and do we support ASS/SSA in avi? :D
[12:10] <wm4> no
[12:10] <Daemon404> aww
[12:10] <Daemon404> because thats a real thing
[12:10] <Daemon404> for ages
[12:10] <wm4> I know, I found such a thing recently on the samples server
[12:11] <wm4> anyway, ffmpeg's avi demuxer generally behaves somewhat worse for incomplete files compared to mplayer's unmaintained crap
[12:11] <Daemon404> even with indexing?
[12:12] <wm4> well, mplayer requires you to pass -idx to make it seekable, but yes
[12:12] <michaelni> wm4 upload samples + open ticket
[12:13] <cone-490> ffmpeg.git 03Clément BSsch 07master:e01fba556213: movenc: Add an F4V muxer
[12:13] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:76424b2da850: Merge commit 'e01fba5562133ac56a4f44c6f214a80ee33cb388'
[12:20] <cone-490> ffmpeg.git 03Diego Biurrun 07master:f6633c55a3c0: avfilter: Fix typo in Loren's email address
[12:20] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:f4f8499c1973: Merge commit 'f6633c55a3c0e93a5b2bab6aa0692fb608f2a38d'
[13:12] <cone-490> ffmpeg.git 03Luca Barbato 07master:72072bf9de32: mpegvideo: Drop a faulty assert
[13:12] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:3d3a7f5034a2: Merge remote-tracking branch 'qatar/master'
[14:17] <BBB> pff
[14:17] <BBB> ubitux: I'll look again
[14:18] <BBB> do we have a sse/32bit x86 fate station
[14:18] <BBB> ?
[14:43] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:867e7bb4f101: fix order of operations in comments
[14:43] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:1e2ab9846076: avcodec/vc1dec: use logical operation instead of bitwise for twomv
[15:20] <durandal_1707> yes!, here goes ultimate filter
[15:46] <cone-490> ffmpeg.git 03Robert Krüger 07master:83922cccb207: avfilter/vf_scale: Correct vf_scale mappings from arguments to color range
[15:59] <burek> was there any attempt to implement the decoding of ARF files (cisco webex video) in ffmpeg?
[15:59] <durandal_1707> no
[16:02] <burek> ok, thanks
[16:20] <funman> michaelni: do you have a plan for removing CODEC_CAP_EXPERIMENTAL on hevc ? Just let a few weeks pass and see how it performs?
[16:20] <durandal_1707> once it is no longer experimental
[16:21] <funman> I'm asking how to help
[16:21] <funman> not when it will be done
[16:22] <durandal_1707> for help send patch
[16:25] <durandal_1707> burek: do you have sample?
[16:26] <michaelni> funman, i dont really have a plan, the removial time depends on the authors/developers. that is if you post a patch and noone objects i guess it can be removed. maybe ill wait till after 2.1 though with applying it in that case, not sure ...
[16:28] <cone-490> ffmpeg.git 03James Almer 07master:c994b42830cb: lavf/adtsenc: Add support for APE tags
[16:29] <j-b> APE tags on ADTS out? brrr
[16:30] <durandal_1707> users reports such bugs, and carl forgets to close them as invalid/wontfix
[16:31] <Daemon404> or users report valid bugs, and carl loses them as invalid/wontfix
[16:31] <Daemon404> or "full uncut output"
[16:32] <durandal_1707> admit you are lazy to provide "full uncut output"
[16:44] <saste> michaelni, ping on "lavu: add av_get_color_string() and use it in -colors option"
[16:45] <durandal_1707> you copy/pasted description from source code?
[16:48] <mraulet> michaelni, funman, smarter, I think we can go without CODEC_CAP_EXPERIMENTAL
[16:48] <mraulet> *go on
[16:54] <saste> any reason to prefer "avcodec:" to "lavc:"?
[16:54] <saste> (in commit messages)
[16:55] <durandal_1707> i think i already answered it
[16:56] <saste> durandal_1707, we do it because libav does it?
[17:00] <Compn> burek : i found arf/wrf samples the other day... should i put them on samples repo ?
[17:01] <michaelni> saste, patch should still be fine
[17:07] <durandal_1707> saste: we do it because you do not ;)
[17:08] <durandal_1707> so there are devs that do use lavf/lavc.. and others that use avformat/avcodec/... and they hate each other
[17:10] <saste> durandal_1707, ok let's start another fork ;)
[17:11] <ubitux> i switched to avformat/avcodec/..., i don't mind at all lavf/lavc/... 
[17:11] <ubitux> <@saste> any reason to prefer "avcodec:" to "lavc:"? // ease copy paste, according to michael
[17:12] <ubitux> don't remember the other arguments, maybe the fork following that convention
[17:13] <durandal_1707> saste: just pick name and i'm in
[17:13] <saste> uh? honestly I don't believe in such silly micro copy&paste optimization
[17:13] <saste> durandal_1707, ssmpeg?
[17:13] <durandal_1707> next
[17:13] <saste> ff2
[17:13] <saste> libmm
[17:14] <saste> ffnext
[17:14] <saste> fffork
[17:15] <saste> durandal_1707, what's the point of an option which is not implemented??
[17:16] <durandal_1707> i would already implement it if I do not need to answer such questions
[17:16] <saste> so please do, and i won't ask such answers
[17:23] <durandal_1707> saste: what about forking libavcodec/libavformat/libavfilter/libavutil/... each into separate project?
[17:23] <ubitux> start with libpostproc
[17:23] <ubitux> oh my bad, already done
[17:23] <ubitux> great success
[17:26] <durandal_1707> http://yro.slashdot.org/story/13/10/22/153259/call-yourself-a-hacker-lose-your-4th-amendment-rights?sbsrc=md
[17:28] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:3ce7c62e1b47: avformat/framehash: add software version to framecrc/md5
[17:28] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:094c40ca3787: avcodec/utils: use a default lock manager that uses a pthread mutex
[17:32] <Compn> wm4 : yes, i dont remember if vaapi got a vdpau module, but someone just gave up and now vdpau supports intel atom
[17:58] <durandal_1707> michaelni: i wonder why fieldorder does not accept vertically subsampled chroma
[18:01] <durandal_1707> i will just remove that limit
[18:11] <michaelni> durandal_1707, ok
[18:11] <durandal_1707> if you could test my patch on pcc/bigendian,etc it would be great
[18:11] <durandal_1707> because i suspect test is doing conversion with swscale
[18:12] Action: Daemon404 uses a cross-chain + qemu-ppc
[18:20] <michaelni> durandal_1707, which patch exactly?
[18:21] <durandal_1707> just sent
[18:21] <durandal_1707> on ml
[18:22] <durandal_1707> i hope you did not do something similar already....
[18:28] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:423ae31354a9: avutil/log: make default av_log callback thread safe
[18:28] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:05eb0f125c70: avutil/log: document log callback thread saftey requirements
[18:31] <durandal_1707> but i think for vertcially subsampled chroma filter should do nothing actually
[18:41] <michaelni> durandal_1707, yuv420 chroma is interlaced too, and patch seems to pass on mips-qemu
[18:44] <durandal_1707> well it appears it fails with disabled-asm
[18:44] <durandal_1707> but i'm more concerned about correctness of patch
[18:52] <durandal_1707> but similar is done with field filter...
[19:00] <cone-490> ffmpeg.git 03Diego Biurrun 07master:0d6d4a9e4a62: avstring-test: Mark pointer passed to av_free() as non-const
[19:00] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:23dec54c5dca: Merge commit '0d6d4a9e4a626be6ff55839eab4114aeac45ba88'
[19:02] <durandal_1707> michaelni: so for fieldorder fate failure are to solutions: add bitexact swscale flag or apply that patch (if patch makes sense)
[19:16] <Mista_D> Anyway to see "drop_frame_flag" via FFprobe? Need to distinct between drop and non-drop frame videos.
[19:21] <Mista_D> in -show_packets "flags" don't show anything like "drop_frame_flag=0 / closed_gop=1 / broken_link=0"...
[19:23] <mraulet> michaelni: you can remove CODEC_CAP_EXPERIMENTAL to HEVC
[19:29] <smarter> michaelni: I'm okay with removing CODEC_CAP_EXPERIMENTAL now, but I worry about potential security issues
[19:29] <smarter> I'm not sure how much fuzzing has been done
[19:33] <michaelni> the same applies to every other decoder, i dont think things would be more secure if we simply kept every new decoder 6month under CODEC_CAP_EXPERIMENTAL
[19:36] <cone-490> ffmpeg.git 03Paul B Mahol 07master:3a75c4e93b3b: fate: fix fieldorder test
[19:42] <cone-490> ffmpeg.git 03Luca Barbato 07master:1ce3ec24b35d: h263: K&R formatting cosmetics
[19:42] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:4c18c31438d0: Merge commit '1ce3ec24b35d9d549b12f17bd7fd391abaf3a6f7'
[19:51] <cone-490> ffmpeg.git 03Vittorio Giovara 07master:2c993e8b5eca: golomb: K&R formatting cosmetics
[19:51] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:02a31f6a7490: Merge remote-tracking branch 'qatar/master'
[19:52] <durandal_1707> K&R contest
[20:01] <michaelni> sounds like a C coding contest
[20:03] <durandal_1707> coding contest is when you write something out of nothing
[20:08] <michaelni> to quote wikipedia "The C Programming Language (sometimes referred to as K&R, after its authors' initials) is a well-known computer programming book written by Brian Kernighan and Dennis Ritchie ..."
[20:10] <michaelni> "... the latter of whom originally designed and implemented the language, as well as co-designed the Unix operating system with which development of the language was closely intertwined. ..."
[20:10] <Daemon404> coould also be a beard growing contest
[20:12] Action: michaelni just thinks using their initials for a specific whitespace formatting is somehow lame and degrading ...
[20:13] <wm4> "Libav developers degrade C inventors"
[20:13] <wm4> nice headlines
[20:14] <ubitux> poor ritchie
[20:14] <ubitux> having his name associated with cosmetics regularly
[20:15] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:92f649b411eb: avcodec/hevc: remove CODEC_CAP_EXPERIMENTAL
[20:58] <Compn> michaelni : make a hook that changes 'k&r' in commit log to 'Brian Kernighan and Dennis Ritchie'
[20:58] <Compn> ;D
[20:59] <Compn> wm4 : learn to troll better
[20:59] <wm4> Compn: ?
[21:00] <Daemon404> ubitux, costmetics are more of an RMS thing: http://farm5.staticflickr.com/4023/4535097784_1b40890d9a_o.jpg
[21:02] <Compn> wm4 : nm
[21:07] <durandal_1707> Daemon404: more!
[21:36] <llogan> Daemon404: heh. that could also be a boy-band/neckbeard album cover.
[21:36] <Daemon404> the backhair boys
[21:36] <llogan> RMS: Toe Cheese
[22:01] <ubitux> BBB: ping
[22:18] <cone-490> ffmpeg.git 03Michael Niedermayer 07master:fbdc98c758d5: avcodec/metasound_data: Fix inconsistency in tables
[00:00] --- Thu Oct 24 2013