[Ffmpeg-devel-irc] ffmpeg-devel.log.20170930

[burek]

[00:07:41 CEST] <JEEB> ok, posted what I see as a fix to check_pkt on the ML
[00:08:02 CEST] <JEEB> although to be honest at this point check_pkt is probably doing more damage than the assert0s added in 2014
[00:08:25 CEST] <JEEB> as the asserts were handled before movenc.c started poking the timestamps
[00:08:35 CEST] <JEEB> but check_pkt seems to be rather late in the chain
[00:49:32 CEST] <jamrial> ubitux, michaelni: big merge (moving cli to a separate folder). can you test to see if anything broke? https://github.com/jamrial/FFmpeg/commits/mergework
[00:49:37 CEST] <jamrial> or anyone, really
[00:51:46 CEST] <atomnuker> are you sure its a good idea to merge this?
[00:51:59 CEST] <atomnuker> that's a pretty big change
[00:52:55 CEST] <michaelni> atomnuker, +1
[00:54:35 CEST] <jamrial> we already merged the one for the build system. this finishes the root directory cleaning by moving also the cli source files to their own folder
[00:54:53 CEST] <jamrial> the compiled binaries are still compiled to the root, in any case
[00:55:44 CEST] <jamrial> not merging it will probably also be a pain with future merges
[00:59:07 CEST] <jamrial> we could maybe merge it after 3.4 is tagged if that's preferred? i assume that's happening soon
[00:59:15 CEST] <jamrial> it would look nice as something introduced in 4.0 alongside the major bump, i guess
[01:00:29 CEST] <atomnuker> jamrial: they are?
[01:00:42 CEST] <atomnuker> ok, if the binaries are still in root then its okay
[01:01:49 CEST] <jamrial> yeah, the point is cleaning the source tree a bit
[01:02:10 CEST] <jamrial> since ffmpeg is getting new .c files with every new hwaccel, it's getting cluttered
[01:02:27 CEST] <nevcairiel> those are on their way out tho
[01:03:00 CEST] <jamrial> not anytime soon i guess, seeing they are still in libav
[01:03:12 CEST] <nevcairiel> they already were greatly reduced
[01:03:24 CEST] <nevcairiel> we used to have dxva2, vdpau and vaapi, those are all gone
[03:48:00 CEST] <cone-596> ffmpeg 03Brian Matherly 07master:876dada0b58b: avcodec/hevc_sei: Support HEVC paired fields.
[03:48:00 CEST] <cone-596> ffmpeg 03Jan Ekström 07master:0275bdbd2c00: MAINTAINERS: add myself to the general developers list
[06:42:01 CEST] <cone-596> ffmpeg 03James Almer 07master:450cee522ec1: build: fix cleaning compiled unstripped examples
[17:13:21 CEST] <jamrial> ubitux, michaelni: ping for the latest merge on my github repo
[17:51:50 CEST] <cone-581> ffmpeg 03Ashish Singh 07master:148c8e88c43c: avfilter: add vmafmotion filter
[18:07:51 CEST] <ubitux> jamrial: will test now
[18:08:12 CEST] <jamrial> thanks
[18:10:05 CEST] <ubitux> so& we still have ffserver?
[18:10:38 CEST] <ubitux> jamrial: you want to keep AVPROGS instead of FFPROGS?
[18:14:01 CEST] <ubitux> seems to work fine so far
[18:14:28 CEST] <wm4> ffserver will stay with us forever
[18:14:45 CEST] <ubitux> fforever
[18:17:21 CEST] <jamrial> ubitux: avprogs is the current name. do you want me to change it?
[18:17:34 CEST] <jamrial> wm4: it'll be gone a month or so after the major bump
[18:17:34 CEST] <ubitux> maybe later
[18:17:52 CEST] <ubitux> jamrial: wrt FF_DEP_LIBS, no change forgotten?
[18:18:21 CEST] <ubitux> like, Makefile:%$(PROGSSUF)_g$(EXESUF): $(FF_DEP_LIBS)
[18:18:31 CEST] <ubitux> isn't this supposed to be in fftools/Makefile?
[18:19:46 CEST] <ubitux> same for the other _g thing
[18:19:48 CEST] <jamrial> i don't think so,
[18:19:54 CEST] <ubitux> ah wait, it also affects tools and examples etc
[18:19:58 CEST] <ubitux> my bad
[18:20:00 CEST] <jamrial> in the libav commit it stays in makefile
[18:20:01 CEST] <jamrial> yeah
[18:20:26 CEST] <ubitux> fate-source fails
[18:20:46 CEST] <jamrial> what's the error?
[18:21:05 CEST] <ubitux> +fftools/cmdutils.h
[18:21:07 CEST] <ubitux> +fftools/ffmpeg.h
[18:21:09 CEST] <ubitux> +fftools/ffserver_config.h
[18:21:11 CEST] <ubitux> missing this
[18:21:48 CEST] <jamrial> ah, nonstandard headers
[18:22:47 CEST] <jamrial> why weren't those reported before, though?
[18:22:59 CEST] <ubitux> probably because you should have FFTOOLS_CMDUTILS_H
[18:23:04 CEST] <ubitux> instead of CMDUTILS_H
[18:23:06 CEST] <ubitux> or sth like that
[18:23:08 CEST] <jamrial> ah right
[18:28:51 CEST] <ubitux> jamrial: no other comment from me
[18:28:55 CEST] <ubitux> seems to be working
[18:29:09 CEST] <jamrial> ok, thanks a lot
[18:33:18 CEST] <jamrial> i like that moving the files around doesn't affect git blame
[18:36:09 CEST] <jamrial> michaelni: any comments?
[18:40:36 CEST] <michaelni> jamrial, give me a moment, i ll take a look & need t post 3 unrelated patches i was working on
[18:45:21 CEST] <jamrial> michaelni: sure
[19:25:49 CEST] <Shiz> aa/w 44
[19:35:36 CEST] <ubitux> it's a real pita to make the vda wrapper work
[19:35:43 CEST] <ubitux> because the api depends on vda stuff
[19:35:54 CEST] <ubitux> like, i can't include vda.h because it relies on the native header
[19:36:15 CEST] <ubitux> and i can't do much about it since the exposed structure actually has vda fields
[19:36:32 CEST] <ubitux> so i'm abandonning the patch
[19:36:37 CEST] <ubitux> feel free to take over
[19:41:25 CEST] <wm4> lol
[19:41:35 CEST] <wm4> good job ffmpeg
[19:41:46 CEST] <wm4> can't even get rid of an API that Apple simply removed ages ago
[19:44:16 CEST] <ubitux> we should be extra careful when ppl submit stuff that touch the public api
[19:44:31 CEST] <ubitux> most of them have zero clue about symbols exposure
[19:48:15 CEST] <JEEB> wow, I think I was just mind-boggled by AVFMT_AVOID_NEG_TS_MAKE_ZERO actually doing rather funky stuff
[19:48:33 CEST] <JEEB> (or whatever is the default)
[19:49:23 CEST] <wm4> what is that even
[19:49:42 CEST] <JEEB> what -avoid_negative_ts controls
[19:49:47 CEST] <JEEB> in the cli
[19:49:53 CEST] <JEEB> it's an avformat option
[19:50:05 CEST] <JEEB> ok, it seems like the default in movenc is avoid_neg_ts_make_zero
[19:50:22 CEST] <JEEB> but the fun part is that I suddenly lose all of my timestamp offsetting
[19:50:31 CEST] <JEEB> even though it's waaay beyond zero
[19:50:48 CEST] <JEEB> think of `-itsoffset "+1506788648" -copyts -start_at_zero` in cli
[19:51:49 CEST] <JEEB> and while testing with cli -debug_ts shows me that the muxer is indeed fed with the initial dts of 1506788648 on the matching time base
[19:52:30 CEST] <JEEB> anyways, would have not expected avformat to suddenly "neuter" my offset
[19:52:35 CEST] <JEEB> even though it's way there positive
[19:54:05 CEST] <JEEB> instead I get an initial DTS=705717 for audio and DTS=834167 for video
[19:54:27 CEST] <jamrial> ubitux: apply it after the bump
[19:54:33 CEST] <jamrial> the symbols wont matter then
[19:55:05 CEST] <JEEB> anyways, I guess I'll whip up a test case and I guess someone can explain me why I am losing all of my offset
[19:57:41 CEST] <ubitux> jamrial: ok
[20:07:22 CEST] <michaelni> jamrial, "./configure && make -j12 alltools" fails with your tree (tools/ismindex.c:50:22: fatal error: cmdutils.h: No such file or directory)
[20:07:54 CEST] <jamrial> ah, tools/
[20:38:46 CEST] <cone-581> ffmpeg 03James Almer 07master:67057aaeb23d: avfilter/vmaf_motion: use correct header guards
[20:46:22 CEST] <cone-581> ffmpeg 03Carl Eugen Hoyos 07master:e9f9175db60a: lavf/bit: Fix the G.729 bit auto-detection.
[20:49:32 CEST] <jamrial> michaelni: alltools and fate-source should be fixed
[20:50:17 CEST] <jamrial> actually, it seems ismindex doesn't even need that header
[20:50:25 CEST] <jamrial> i don't know why it's including it
[00:00:00 CEST] --- Sun Oct  1 2017