[Ffmpeg-devel-irc] ffmpeg-devel.log.20180813

burek burek021 at gmail.com
Tue Aug 14 03:05:04 EEST 2018


[00:51:02 CEST] <Zeranoe> Anyone know how to debug FFmpeg with gdb when the stack is corrupted? "Backtrace stopped: previous frame identical to this frame (corrupt stack?)", I can't figure out why specifically the Windows 64-bit *shared* builds are immediately crashing
[00:51:56 CEST] <Zeranoe> It's something related to x265, and something related to FFmpeg. It only effects 64-bit shared linking, and x265 by itself seems to run fine.
[00:52:28 CEST] <Zeranoe> I've made like 50 different builds to try and figure out what's going on
[00:53:01 CEST] <BtbN> Try building with strong stack protectors
[00:57:29 CEST] <jamrial> also, it would be a good idea to bisect to find where it started failing. assuming it was an ffmpeg change and not some other change in your environment
[04:18:25 CEST] <Compn> [adpcm_ms @ 015101e0]Channel layout '4 channels (FL+FR+FC)' with 4 channels does not match specified number of channels 1: ignoring specified channel layout
[04:18:48 CEST] <Compn> Stream #0:0: Audio: adpcm_ms ([2][0][0][0] / 0x0002), 22050 Hz, stereo, s16, 176 kb/s
[04:18:55 CEST] <Compn> strange file !
[04:20:53 CEST] <Compn> now how to verify if it really has 4 channels or not
[04:20:56 CEST] <Compn> probably not
[09:14:17 CEST] <gagandeepsingh> kierank: thanks to durandal_1707, the intra only is not broken anymore and the ip frame thread is working,just gonna make sure sync is fine, now i am no longer getting seg fault
[09:19:44 CEST] <durandal_1707> gagandeepsingh: so? have you managed so solve it?
[09:20:16 CEST] <gagandeepsingh> intra only like you said was going wrong because on init thread copy
[09:20:26 CEST] <gagandeepsingh> i was not able to work out this bug,
[09:20:55 CEST] <gagandeepsingh> that is why i wanted someone to review the file because frame threading is something i am new at
[09:21:33 CEST] <kierank> I would put it on ml
[09:21:43 CEST] <kierank> Also note cfhd patches on ml
[09:21:50 CEST] <gagandeepsingh> working on syncing up ip samples, now that ffplay is not crashing
[09:22:01 CEST] <gagandeepsingh> will send serial ip decoding patch first
[09:22:06 CEST] <gagandeepsingh> then the frame thread part
[09:26:03 CEST] <kierank> gagandeepsingh: there are cfhd patches from others on the mailing list you need to review
[09:34:04 CEST] <gagandeepsingh> just saw the patches right now, will do that
[10:55:37 CEST] <cone-762> ffmpeg 03sanil 07master:551a029a181a: libavformat/dashdec: Support negative value of the @r attrbute of S in SegmentTimeline element.
[11:34:24 CEST] <durandal_1707> michaelni: do you know why imm4 decoder have artifacts? both coefficients and idct seems to be fine
[11:40:35 CEST] <gagandeepsingh> kierank: samples (the ones that i have) are working fine now with frame threading
[11:47:08 CEST] <j-b> durandal_1707: pong
[11:48:27 CEST] <durandal_1707> gagandeepsingh: cant you encode more samples? even with virtualdub2 under windows/wine?
[11:51:03 CEST] <durandal_1707> j-b: are you willing to sponsor IMM4 decoder addition to libavcodec? I will add relevant info to commit log(This work is sponsored by VideoLAN), there is IMM4 listed on vlc bounty page.
[11:51:20 CEST] <j-b> of course.
[11:51:31 CEST] <j-b> https://forum.videolan.org/viewtopic.php?t=41968
[11:53:00 CEST] <gagandeepsingh> durandal_1707: the cfhd doesn't encode samples in ip format in new apps
[11:53:21 CEST] <gagandeepsingh> it used to be an option in the old video editing softwares cineform made
[11:53:35 CEST] <durandal_1707> what? it encodes only bayer ones?
[11:53:46 CEST] <gagandeepsingh> i mean intra only
[11:54:00 CEST] <gagandeepsingh> it doesn't encode bayer as well
[11:54:35 CEST] <gagandeepsingh> david newman said to get more ip samples one would need to use the old apps but sadly they are no longer avaliable for download
[11:55:10 CEST] <gagandeepsingh> and i mean the options avaliable in adobe
[11:58:19 CEST] <durandal_1707> gagandeepsingh: i just encoded intra only samples with virtualdub2
[11:58:37 CEST] <durandal_1707> ffmpeg decodes them just fine
[12:01:15 CEST] <durandal_1707> and virtualdub2 uses cineform sdk
[12:04:19 CEST] <gagandeepsingh> ok, but i fail to understand what are we discussing?
[12:04:31 CEST] <gagandeepsingh> i don't know anything about virtualdub
[12:05:03 CEST] <gagandeepsingh> and intra only is fine but can it do 3d transform (ip files)?
[12:06:46 CEST] <durandal_1707> gagandeepsingh: apparently not yet (you could ask on relevant doom9 forum for feature, if native cineform code supports it), virtualdub2 is GUI program that can encode files
[12:07:40 CEST] <gagandeepsingh> ok
[19:34:26 CEST] <gagandeepsingh> kierank: do i send a seperate patch for serial ip sample, codebook 0 decompanding fix and frame threading?
[19:37:44 CEST] <kierank> gagandeepsingh: yes
[19:37:53 CEST] <kierank> gagandeepsingh: nice work
[19:38:11 CEST] <kierank> Sorry I have not been able to help you recently
[19:38:33 CEST] <gagandeepsingh> thanks, really just seeing how all the structs are made and passed around in this codebase has left me amazed
[19:39:23 CEST] <gagandeepsingh> i mean to make all things consistent all the functions and api has been tough at times and amazing
[19:39:38 CEST] <gagandeepsingh> going through them, i mean
[19:45:45 CEST] <gagandeepsingh> kierank: also i had to use certain hacks to make mountain file run in ffmpeg, they are not needed for the other samples, and i haven't encountered any trouble in using them with the other samples, so should i keep them?
[19:46:11 CEST] <kierank> Maybe yes, we will see in review
[19:46:19 CEST] <gagandeepsingh> ok
[19:46:55 CEST] <kierank> Mountain file was pain of my life when trying to reverse engineer cfhd
[19:47:23 CEST] <kierank> So pleased that it works
[19:47:55 CEST] <gagandeepsingh> it works, but the vlc part is giving trouble, the table 9, other than that it is working
[19:48:14 CEST] <gagandeepsingh> all the other samples that we have use table 18
[19:48:39 CEST] <gagandeepsingh> haven't been able to fix that part yet
[19:49:25 CEST] <gagandeepsingh> also, you would want to repair it's container with ffmpeg
[20:10:53 CEST] <gagandeepsingh> kierank: how do i send the samples that i have to ffmpeg
[22:10:10 CEST] <durandal_1707> this IMM4 codec is insane, if I copy video into another avi container without subtitle stream it would play with same artifacts as my decoder
[22:11:11 CEST] <durandal_1707> stupid mystery finally resolved
[22:13:13 CEST] <durandal_1707> actually i do  not need to disable subtitle stream (which is always empty)
[22:13:24 CEST] <durandal_1707> looks like avi demuxer issue
[22:13:58 CEST] <durandal_1707> it simply does not demux all packets for some reason
[22:15:07 CEST] <durandal_1707> AVI demuxer expert WANTED!
[23:30:50 CEST] <TD-Linux> jamrial, I noticed that ffmpeg -strict -2 writes "AV01" as its sample type but the spec says "av01". is this intentional?
[23:31:56 CEST] <jamrial> i noticed it, it only happens if you remux from ivf, afaik
[23:31:58 CEST] <jamrial> it's weirdness in libavformat/ffmpeg.c code
[23:32:20 CEST] <jamrial> it does a case insensitive check for the codec tag, and in that case av01 == AV01, so it just writes it as is
[23:35:06 CEST] <TD-Linux> oh. I suppose changing that to a case sensitive check in ffmpeg.c would be too likely to break other things?
[23:35:38 CEST] <jamrial> i have no idea, but i bet it would :p
[23:36:00 CEST] <TD-Linux> should I post a patch just to hack the mp4 muxer to convert AV01 to av01?
[23:36:40 CEST] Action: TD-Linux didn't realize that ffmpeg made fourccs ~= sample type
[23:36:42 CEST] <jamrial> sure
[00:00:00 CEST] --- Tue Aug 14 2018


More information about the Ffmpeg-devel-irc mailing list