[Ffmpeg-devel-irc] ffmpeg.log.20181221

[burek]

[00:04:59 CET] <Keshl> Just a little FYI: After setting VFR up for my weirdo specific use-case, I have /zero issues/ with seeking. I suspect it's because I used faststart so the entire MOV atom is available right away, and thus nothing ends up in "odd" spots. Regardless, everything's totally fine. The only "weird" bit is I can't seek to the middle of a long-duration frame, but I can jump to it's beginning or end easily enough, and I generally don't need the
[00:05:00 CET] <Keshl> audio from those portions.
[00:07:59 CET] <Keshl> Also, using mpdecimate, I have no dropped frames where I don't expect them or heavily desire them. Everything just worked out. o_o
[00:08:31 CET] Action: Keshl hugs all the developers. -É-.
[00:49:59 CET] <ntd> can ffmpeg be convinced to pull, say, a sep audio stream suing anything other than sep -i?
[00:53:43 CET] <pink_mist> -i is how you tell it to pull things
[00:53:57 CET] <pink_mist> your question really doesn't make much sense to me
[01:45:17 CET] <fengshaun> is it necessary to run frames through filter if I only want to re-encode?
[01:45:35 CET] <fengshaun> the transcoding example uses filters and I'm slowly trying to get rid of it
[01:49:57 CET] <fengshaun> if I remove filter graph from the transcoding example, I get "corrupted size vs prev_size"
[01:53:23 CET] <fengshaun> nevermind, it seems the segfault was coming from elsewhere
[05:53:13 CET] <fengshaun> I'm trying to transcode a file with ffmpeg api, so far I have completed decoding, but when encoding, I'm getting the following error from libx264: "Input picture width (1280) is greater than stride (0)"
[05:53:58 CET] <fengshaun> I've tried setting AVFrame.linesize[0 to 2] manually which results in segfault, and using avpicture_fill results in same segfault.
[05:54:03 CET] <fengshaun> not sure what to do at this point
[05:54:09 CET] <fengshaun> and I can't find any info on the web
[05:54:21 CET] <fengshaun> using h264 decoder and encoder
[05:54:26 CET] <fengshaun> there are no filters
[05:54:44 CET] <fengshaun> it seems adding a filter in between works fine like transcoding example
[05:54:53 CET] <fengshaun> but I'm trying to skip filtering
[06:06:57 CET] <fengshaun> av_image_fill_linesizes(frame->linesize, pixel_format, frame_width) also results in return value of 0 and a segfault at send_frame
[06:29:44 CET] <fengshaun> nvm again, looks like I was mis-using send/receive api
[06:29:53 CET] <fengshaun> calling them repeatedly until EAGAIN
[06:30:00 CET] <fengshaun> which messed up frame data
[06:31:29 CET] <fengshaun> looks like it's working!!
[13:26:04 CET] <th3_v0ice> How can I force the sending of PAT (Program Association Table) in MPEG-TS streams?
[13:26:18 CET] <JEEB> pretty sure it's done by default, since it's a required table
[13:26:27 CET] <JEEB> you would have to specifically disable sending PAT
[13:26:40 CET] <JEEB> how often PAT is sent is a separate discussion, for that you'd have to look into the mpeg-ts muxer
[13:55:52 CET] <th3_v0ice> JEEB, Well the control software is reporting that its not sent fast enough (>500ms). Will look into the MPEG-TS muxer. Thanks.
[14:17:43 CET] <th3_v0ice> It seems that these options are configurable with input parameters. -pcr_period -pat_period.
[17:25:24 CET] <tombb> hi all, can ffmpeg create mxf-atom files? (for avid environment)
[17:26:17 CET] <JEEB> MXF, yes
[17:26:31 CET] <JEEB> now what "MXF-Atom" means I have no idea
[17:26:54 CET] <tombb> there's mxf op1a, which includes several tracks, video+audio
[17:27:14 CET] <tombb> if u split it to different files, one per stream, they're atoms (as far as I understand)
[17:29:01 CET] <tombb> if you're familiar with bmxtranswrap.exe, that what it does, split op1a to atoms.. I'm not really sure how that works and whats the difference between an mxf with a single stream to mxf-atom
[17:31:15 CET] <^Neo> I'm finding some odd behavior on 4.0, I'm generating an HLS stream from a hardware device, using -use_wallclock_as_timestamps because I don't trust the hardware. I have '+append_list' in the hls flags. If I probe the resulting m3u8 and print the PTS's, and restart the HLS instance I see the PTS's jump by ~24 hours...
[19:26:43 CET] <funyun> what's the best way to run ffmpeg in windows in the most linuxy way possible?
[19:28:14 CET] <Rick627> just like linux
[19:28:18 CET] <Rick627> in the command line
[19:29:42 CET] <funyun> Rick627: okay thanks
[19:30:26 CET] <^Neo> when I run '-fflags +genpts' do I need to specify a '-r' argument with a framerate?
[00:00:00 CET] --- Sat Dec 22 2018