[Ffmpeg-devel-irc] ffmpeg-devel.log.20170409

[burek]

[00:15:54 CEST] <BBB> ubitux: do you think its possible to re-run the helgrind machine also? (I believe you own that, right?) Im wondering if that improved also and/or if it detects other issues
[00:26:01 CEST] <michaelni> atomnuker, ive already tested the patchset a day or so ago but due to it changing all files generated its hard to say much with certainity, just that the files i looked at looked ok
[00:43:01 CEST] <atomnuker> michaelni: I'll commit the first patch to both master and the 3.3 branch, I'll leave the one which enables huffman by default only in master
[00:48:11 CEST] <michaelni> ok
[01:03:35 CEST] <cone-479> ffmpeg 03Rostislav Pehlivanov 07release/3.3:6c9574e49021: mjpegenc_common: check for codec ID before using avctx->priv_data
[01:03:36 CEST] <cone-479> ffmpeg 03Rostislav Pehlivanov 07release/3.3:7034009f625d: mjpegenc: disable huffman coding with AMV
[01:04:13 CEST] <cone-479> ffmpeg 03Rostislav Pehlivanov 07master:2c9be3882a03: mjpegenc_common: check for codec ID before using avctx->priv_data
[01:04:14 CEST] <cone-479> ffmpeg 03Rostislav Pehlivanov 07master:7f9b492d544c: mjpegenc: disable huffman coding with AMV
[01:04:15 CEST] <cone-479> ffmpeg 03Rostislav Pehlivanov 07master:45eeb1f785d6: mjpegenc: enable optimal huffman coding by default
[01:06:21 CEST] <durandal_1707> nobody understands frame threads?
[01:07:07 CEST] <JEEB> in what sense? I kind of implemented that in a thing few years back but it was an intra thing and all
[01:07:39 CEST] <JEEB> (and it's not like I did anything more than utilized what was already there as the framework)
[01:08:09 CEST] <durandal_1707> for intraonly decoders
[01:09:45 CEST] <JEEB> you pass each thread its own little data set and then make the decode function not depend on global things that could change?
[01:09:55 CEST] <JEEB> man, should look into that crap again at some point .-.
[01:10:12 CEST] <JEEB> barely have the time to test patches because of various shit
[01:11:51 CEST] <atomnuker> durandal_1707: does it pass fate (assuming we have tests for that)?
[01:12:28 CEST] <atomnuker> the patches look okay comparing to other intra-only threaded decoders, so it should be safe to enable if it works
[03:52:14 CEST] <cone-479> ffmpeg 03Michael Niedermayer 07master:9244b839b788: avcodec/dcadsp: Fix runtime error: signed integer overflow
[03:52:15 CEST] <cone-479> ffmpeg 03Michael Niedermayer 07master:23edd41a0d69: doc/examples/decode_video: Fix format string vulnerability
[03:52:16 CEST] <cone-479> ffmpeg 03Warblefly 07master:47ccefac211f: libavfilter/af_amix.c Increase sources from 32 to 1024
[09:44:50 CEST] <cone-342> ffmpeg 03Muhammad Faiz 07master:3408f4669427: avfilter/avf_showcqt: add attack option
[10:49:35 CEST] <cone-342> ffmpeg 03Paul B Mahol 07master:24d9b0c29c9d: avcodec/cllc: add support for frame threads
[10:51:07 CEST] <cone-342> ffmpeg 03Paul B Mahol 07master:20f7872d9903: avcodec/hqx: add support for frame threads
[14:25:30 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:4a1ef543983b: build: Drop support for old versions of libdc1394
[14:25:31 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:a329b822e17a: Merge commit '4a1ef543983b7480e2822f6ac281ba361d1f893d'
[14:56:15 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:3794062ab1a1: Remove Plan 9 support
[14:56:16 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:210678d3c517: Merge commit '3794062ab1a13442b06f6d76c54dce51ffa54697'
[14:57:44 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:1ea77aae927c: configure: Drop bogus xcb_event variables
[14:57:45 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:9986d014aa9b: Merge commit '1ea77aae927c7310034b1f75d4f1c2676fe641f2'
[14:58:25 CEST] <ubitux> anyone opposed to keeping --build-suffix?
[14:58:40 CEST] <ubitux> i believe it's useful as i actually have used it personally for a long period of time
[14:58:56 CEST] <BtbN> I have no idea what it is
[14:59:27 CEST] <ubitux> --build-suffix=btbn ’ libavcodec-btbn.so
[14:59:40 CEST] <ubitux> useful when dealing with multiple versions of libav*
[14:59:49 CEST] <ubitux> (forks or whatever)
[14:59:50 CEST] <BtbN> Yeah, if a distribution would decide to offer slotted builds, they'd kind of need that
[15:00:37 CEST] <RiCON> i know at least foobar2000 uses it 
[15:01:20 CEST] <RiCON> isn't it the same as rename?
[15:02:01 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:14bba44fb23c: build: Drop support for configuring library name suffixes
[15:02:02 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:110b010224a2: Merge commit '14bba44fb23c63bf82b2c15f5c754b2fed7dbea9'
[15:02:04 CEST] <ubitux> RiCON: no because it also affects .pc files
[15:03:34 CEST] <JEEB> yea, I have also used that at least once
[15:05:58 CEST] <ubitux> curious no one stopped him on libav side
[15:06:47 CEST] <RiCON> maybe "no known" users is true for libav vOv
[15:07:02 CEST] <RiCON> (that build suffix feature)
[15:07:30 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:6ef3360a38ef: configure: Add misssing qpeldsp dependency to mss2 decoder
[15:07:31 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:41a220803531: Merge commit '6ef3360a38efe027cb6ebf9785870048ca7eb38b'
[15:07:44 CEST] <BtbN> who even still uses libav in general? Most, if not all, distributions are back to ffmpeg.
[15:08:51 CEST] <BtbN> No project comes to mind that has it as default as well
[15:09:11 CEST] <RiCON> mpv still supports building with it
[15:09:18 CEST] <RiCON> barely
[15:09:40 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:39ecf0588fb2: webp: Convert to the new bitstream reader
[15:09:41 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:f5b7bd2a7c3f: imc: Convert to the new bitstream reader
[15:09:41 CEST] <BtbN> sure, a lot of stuff still supports it, but using it by default?
[15:09:42 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:c3defda0d80e: indeo: Convert to the new bitstream reader
[15:09:43 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:6fad5abcadd4: lagarith: Convert to the new bitstream reader
[15:09:44 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:cdc6727c3e0a: metasound: Convert to the new bitstream reader
[15:09:45 CEST] <cone-342> ffmpeg 03Alexandra Hájková 07master:6c916192f3d7: mimic: Convert to the new bitstream reader
[15:09:46 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:bc94b9534cbe: Merge commit '6c916192f3d7441f5896f6c0fe151874fcd91fe4'
[15:12:13 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:0fdc9f81a00f: build: Add missing hevc_ps dependency for QSV HEVC encoder
[15:12:14 CEST] <cone-342> ffmpeg 03Clément BSsch 07master:63bb3db9aac3: Merge commit '0fdc9f81a00f0f32eb93c324bad65d8014deb4dd'
[15:18:06 CEST] <ubitux> argl, still a bunch of build stuff to merge
[15:18:10 CEST] <ubitux> i don't like those
[15:18:19 CEST] <ubitux> almost as much as the space shuffling ones :(
[15:24:13 CEST] <ubitux> jamrial: would you mind submitting/pushing the svq1 patch you suggested the other day
[15:24:33 CEST] <ubitux> (zero init entries or something, to fix the valgrind issue)
[15:25:21 CEST] <nevcairiel> Diego refactored half of configure somewhere in the backlog
[15:26:42 CEST] <ubitux> 48 "configure" commits, 43 "build" commits
[15:26:45 CEST] <jamrial> ubitux: i don't recall such patch or suggestion :/
[15:27:23 CEST] <jamrial> i fixed some valgrind issues some time ago (that were supposedly false positives) by zero initializing stack in some cases, but it wasn't svq1
[15:29:31 CEST] <ubitux> 2017-01-16 15:43:03     jamrial_        ubitux: zero initializing both entries[6] on svq1dec.c "fixes" the svq1 failures
[15:30:22 CEST] <jamrial> ah, january, no wonder :p
[15:30:37 CEST] <jamrial> let me test that again
[15:46:03 CEST] <jamrial> ubitux: yeah, still fixes it, but it probably just silences a false possitive
[15:46:12 CEST] <jamrial> want me to push it?
[15:48:41 CEST] <ubitux> i'm not the svq1 maintainer :p
[15:48:50 CEST] <ubitux> i only care about valgrind being clean
[15:49:03 CEST] <ubitux> if it's a bug in valgrind, it should be reported there
[15:52:03 CEST] <jamrial> patch sent, then
[15:52:37 CEST] <jamrial> there's no svq1 maintainer listed, though
[15:58:24 CEST] <nevcairiel> ubitux: i use build-suffix as well, for my oss project and work, so thanks for keeping it :D
[15:58:47 CEST] <ubitux> :)
[15:58:52 CEST] <ubitux> so much no known user
[15:58:59 CEST] <nevcairiel> well i dont use libav
[15:59:00 CEST] <nevcairiel> so!
[15:59:17 CEST] <ubitux> ;)
[16:09:30 CEST] <jamrial> foobar 2k also uses it
[16:09:38 CEST] <jamrial> i'm surprised libav removed it
[16:10:39 CEST] <durandal_1707> perhaps handbrake does not need it
[16:20:43 CEST] <durandal_1707> what filter to write next?
[16:23:58 CEST] <ubitux> motion estimation framework
[16:29:19 CEST] <nevcairiel> BtbN: to answer your earlier question, handbrake uses it, for some reason. I'm not aware of anything else
[16:38:54 CEST] <durandal_1707> ubitux: ?
[16:39:49 CEST] <durandal_1707> ok to push one small patch? nobody is in mid-merge?
[16:39:54 CEST] <ubitux> durandal_1707: in order to do stuff like http://slowmovideo.granjow.net/
[16:40:00 CEST] <ubitux> durandal_1707: yeah, go ahead
[16:40:48 CEST] <durandal_1707> ubitux: there are already filter for this
[16:40:54 CEST] <ubitux> ah?
[16:41:01 CEST] <ubitux> which one?
[16:41:29 CEST] <cone-342> ffmpeg 03Paul B Mahol 07master:c20cc61890a7: avfilter/vf_midequalizer: add gray10 and gray12 support
[16:49:37 CEST] <jamrial> michaelni, nevcairiel: ok to push the rest of my hevc set?
[16:59:09 CEST] <durandal_1707> ubitux: minterpolate, mestimate
[17:23:06 CEST] <nevcairiel> jamrial: should be fine
[17:40:55 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:e1dc5358af0d: build: Create a component for MPEG audio header decoding
[17:40:56 CEST] <cone-342> ffmpeg 03James Almer 07master:521197ebecba: Merge commit 'e1dc5358af0dadef548114ec9cabba766b7c6260'
[17:42:04 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:660470cd62c5: build: Add missing audio_frame_queue dependency for libwavpack encoder
[17:42:05 CEST] <cone-342> ffmpeg 03James Almer 07master:e5ac47c06818: Merge commit '660470cd62c59f25cc52c731af6230e7bc1aba62'
[18:05:26 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:ffb9025f39b2: configure: Simplify MMAL check
[18:05:27 CEST] <cone-342> ffmpeg 03James Almer 07master:0905c679eb24: Merge commit 'ffb9025f39b274abfc4ad2b264205a44cc0238a4'
[18:07:06 CEST] <atomnuker> durandal_1707: didn't you submit that rangescope filter like a year ago?
[18:07:23 CEST] <atomnuker> oh wait that was the zoneplate filter
[18:08:03 CEST] <durandal_1707> this one is different
[18:10:06 CEST] <ubitux> durandal_1707: so minterpolate and mestimate are "done"? no improvement needed?
[18:10:17 CEST] <ubitux> i think fieldmatch and decimate need some love
[18:10:21 CEST] <ubitux> and probably many other filters
[18:11:22 CEST] <ubitux> my position wrt lavfi is that filters need to be more efficient, faster, and more advanced
[18:12:09 CEST] <atomnuker> frame threading isn't supported :/
[18:12:22 CEST] <ubitux> yeah well, that's actually relatively complex
[18:13:17 CEST] <ubitux> many filters should actually support frame threading though
[18:13:25 CEST] <ubitux> i wonder the impact on memory this has
[18:13:37 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:2a096440768b: configure: Separate package name and version requirements in helper functions
[18:13:38 CEST] <cone-342> ffmpeg 03James Almer 07master:adf9f0416fab: Merge commit '2a096440768b1086bb437939f827b8b7a5716bf7'
[18:13:42 CEST] <ubitux> i would guess it has quite a few implications
[18:31:10 CEST] <jamrial> do i apply 5d45fe7da9?
[18:31:29 CEST] <jamrial> seems a bit overkill
[18:32:18 CEST] <jamrial> it can be appended to ELIBS on a case by case basis, depending on the tool
[18:32:25 CEST] <jamrial> like it's currently done for uncoded_frame
[18:39:27 CEST] <cone-342> ffmpeg 03James Almer 07master:aed84ee4d1b0: avcodec/svq1: zero initialize entries array
[18:46:56 CEST] <nevcairiel> isnt extralibs the one you specify on the command line
[18:47:44 CEST] <nevcairiel> hm its apparently also used for all other library deps
[18:47:58 CEST] <jamrial> no, that's ELIBS
[18:48:10 CEST] <ubitux> jamrial: thx for the patch
[18:48:32 CEST] <ubitux> with the valgrind patch i sent, i can re-run the valgrind fate instances and get them green again
[18:48:57 CEST] <ubitux> (since using the svn valgrind will be possible without accrobatics)
[19:13:17 CEST] <cone-342> ffmpeg 03James Almer 07master:159ab4625bd3: avcodec/hevc_parse: check for parameter set decoding failure
[19:13:18 CEST] <cone-342> ffmpeg 03James Almer 07master:f1533979a228: avcodec/hevc_parse: allow setting apply_defdispwin when decoding SPS NAL units
[19:13:19 CEST] <cone-342> ffmpeg 03James Almer 07master:498e9e8e2a67: avcodec/hevc_parse: split build dependencies
[19:13:20 CEST] <cone-342> ffmpeg 03James Almer 07master:4588063f3ecd: avcodec/hevcdec: remove duplicate extradata parsing function
[19:18:14 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:5d45fe7da972: build: Add EXTRALIBS to TOOLS linker command
[19:18:15 CEST] <cone-342> ffmpeg 03James Almer 07master:a308324ef7e5: Merge commit '5d45fe7da972da528915fbdfe3dbf22eb2effd01'
[19:47:13 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:29d2e03ed96d: configure: Simplify and fix avfoundation indev handling
[19:47:14 CEST] <cone-342> ffmpeg 03James Almer 07master:fa1fe495fc82: Merge commit '29d2e03ed96d9f7b8cb2cc87dca0a63c2a6eff70'
[19:50:14 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:601f8dde13cc: configure: Move COMPONENT_LIST to the bottom of CONFIG_LIST
[19:50:15 CEST] <cone-342> ffmpeg 03James Almer 07master:6992c806e60d: Merge commit '601f8dde13ccd0e1993b7840a0304fa2cfe53432'
[19:54:35 CEST] <cone-342> ffmpeg 03Diego Biurrun 07master:3d6135eacf3b: configure: Simplify OMX check
[19:54:36 CEST] <cone-342> ffmpeg 03James Almer 07master:d8e5a4818b19: Merge commit '3d6135eacf3b6a82c3024620c6a28169960464a7'
[22:38:18 CEST] <cone-342> ffmpeg 03Michael Niedermayer 07master:724bb805ef8a: Revert "mjpegenc: disable huffman coding with AMV"
[22:38:19 CEST] <cone-342> ffmpeg 03Rostislav Pehlivanov 07master:03eb0515c126: mpegvideo_enc: disable optimized huffman coding with AMV or slice threads
[22:45:36 CEST] <atomnuker> michaelni: thanks, I missed to change that
[22:45:56 CEST] <michaelni> np
[22:46:04 CEST] <atomnuker> (you should probably port those 2 commits to release/3.3 since I commited that commit there as well)
[22:46:15 CEST] <michaelni> already did locally
[22:46:24 CEST] <atomnuker> ah cool
[00:00:00 CEST] --- Mon Apr 10 2017