[Ffmpeg-devel-irc] ffmpeg-devel.log.20151113

[burek]

[00:02:51 CET] <michaelni> if you just remove it, things will stop working i guess, if you fix them you likely will end up with similar code then whats in the removed function, quite possible that code could be elsewhere of course
[00:04:09 CET] <nevcairiel> the plan seems to be to stop the muxer from making up its own timestamps and just requiring the caller to set proper timestamps, the new warning in compute_pkt_fields2 speaks to that quite clearly
[00:09:15 CET] <michaelni> that is possible, iam not sure its an improvment for the API though to require user apps to set all fields even when the missing ones could be computed
[00:10:29 CET] <michaelni> there are also some other things in there like timestamp debug av_logs and timestamp validity checks
[00:10:34 CET] <michaelni> and of course ugly hacks
[00:10:41 CET] <nevcairiel> the  validity checks got moved
[00:12:01 CET] <michaelni> i still see a "pkt->duration < 0" in there and also 2 return EINVAL
[00:12:15 CET] <nevcairiel> they didnt get removed from their old place
[00:12:43 CET] <nevcairiel> the new space is #if'ed out by the same deprecation guard that protects the old function, just in reverse
[00:13:05 CET] <nevcairiel> see prepare_input_packet
[00:13:40 CET] <nevcairiel> dunno why it wasnt just moved fully instead of this way, but I just merge =p
[00:14:14 CET] <nevcairiel> probably because the new code runs before compute_pkt_fields2 has a chance to generate timestamps
[00:14:22 CET] <nevcairiel> so it would break now, which is nnot wanted
[00:14:42 CET] <michaelni> the copied checks are different
[00:15:03 CET] <michaelni> from the existing ones
[00:17:33 CET] <nevcairiel> slightly simplified, and of course those that are just warnings that then set timestamps are not present, because invalid timestamps is not a desired usecase anymore
[00:18:41 CET] <nevcairiel> there seems to be some extra hackery for subtitles in the old code that didnt get moved over
[00:18:59 CET] <nevcairiel> that might be needed still
[00:19:27 CET] <michaelni> yes also the human readable timestamp printing code is lost
[00:20:16 CET] <michaelni> if libav moves code it might be best to move the matching code in ffmpeg in its place
[00:20:19 CET] <nevcairiel> that could be added back I suppose, it'll be 2 years until that code gets activated, anyway
[00:20:25 CET] <michaelni> and then compare
[00:20:50 CET] <michaelni> yes and by then it might differ even more
[00:21:05 CET] <michaelni> it would be better if this isnt duplicated as it is
[00:21:17 CET] <nevcairiel> for logging code its always a bit hard to judge, maybe that code was only useful to actually debug the timestamp generation, and not anything else
[00:23:41 CET] <michaelni> personally i think it would have made alot more sense to put the hacks and unwanted parts of  the function under ifdef and not duplicate anythng
[00:24:03 CET] <nevcairiel> dont tell me, i didnt write the patch :)
[00:24:20 CET] <michaelni> yes 
[00:24:39 CET] <nevcairiel> if a merged patch doesnt conflict, I'm usually happy!
[00:24:53 CET] <nevcairiel> if code gets actually moved, I do try to move the ffmepg code instead to make sure nothing is lost
[00:24:58 CET] <nevcairiel> but this is effectively new code
[13:27:31 CET] <devshark> hello
[13:29:08 CET] <devshark> anyone willing to help me out a bit with building ffmpeg for android? i found some code that supposedly builds using NDK r6b, but it's nowhere near whats promissed :(
[13:30:11 CET] <fritsch> https://trac.ffmpeg.org/wiki/CompilationGuide/Android
[13:32:59 CET] <devshark> thanks fritsch. I already tried some of them, but I guess I can try obucinac's as that one probably builds out-of-the-box. I was trying to get something newer than 2.4.2 if possible
[13:33:25 CET] <fritsch> you should ask in the user channel - as per topic
[13:38:05 CET] <devshark> sorry for being off-topic then; would the user channel be #ffmpeg or something else?
[13:46:38 CET] <nevcairiel> thats the one
[13:54:08 CET] <cone-164> ffmpeg 03Bryan Huh 07master:d07585f04add: avformat/async: Fix bug where async could not recover after seek to eof
[20:23:38 CET] <kierank> Actually, I think we already have the framework for a high-level
[20:23:38 CET] <kierank> optionally-asynchronous API: libavfilter. Of course it is far from ready for
[20:23:38 CET] <kierank> it, but the basic principles are sound.
[20:23:40 CET] <kierank> wm4: ^
[20:24:17 CET] <wm4> and I've replied to it
[20:25:26 CET] <kierank> oh 
[20:25:31 CET] Action: kierank can't read
[20:25:59 CET] <nevcairiel> I screamed out in pain
[20:26:57 CET] <wm4> now he wants details about my reservations
[20:27:19 CET] <wm4> writing a reply to this feels like having to write a paper
[22:38:58 CET] <Daemon404> isnt this 100% what i said would happen
[23:03:30 CET] <J_Darnley> Screw your crap OS and screw security.
[23:21:34 CET] <durandal_1707> lavfi rules
[00:00:00 CET] --- Sat Nov 14 2015