[Ffmpeg-devel-irc] ffmpeg-devel.log.20190213

burek burek021 at gmail.com
Thu Feb 14 03:05:04 EET 2019


[03:17:51 CET] <blue_misfit> hey-o! wondering if there's any particular reason some of the recent prores improvements didn't go out with 4.1.1 e.g. https://github.com/FFmpeg/FFmpeg/commit/e4788ae31b2e9af45d11f4bf4498c075dcc25a6c
[09:56:34 CET] <j-b> good moronings!
[13:07:36 CET] <cone-491> ffmpeg 03Jun Zhao 07master:ff5ed7abd6ee: avfilter/tests/integral: Correct the comment
[13:59:42 CET] <Shred00> i don't seem to be getting e-mail confirmation tokens trying to create an account on trac.  can anyone poke it and see if it's still working?
[14:00:27 CET] <JEEB> does anyone remember who's the admin for trac?
[14:18:11 CET] <cone-491> ffmpeg 03Reto Kromer 07master:bf78aa9ee9c4: doc/muxers: fix typo
[17:39:28 CET] <philipl_> BtbN: I've taken a look at the new video sdk headers. Does what the release notes says it does. I'll put together an update at some point this week
[17:44:44 CET] <BtbN> philipl_, nvidia has already sent them to the list
[17:51:34 CET] <philipl_> I see it now. How kind of them.
[17:54:05 CET] <philipl> BtbN: His cuviddec patch seems wrong. I had to introduce the new pixfmts to handle the 444P10/P12 content and he's trying to reuse the existing ones. I assume garbage on the screen as a result.
[17:55:04 CET] <philipl> or at least garbage if he does anything with software filters or encoders.
[17:58:40 CET] <philipl> Do you think I should resend without the new pixfmts and just treat everything as 444P16? And avoid that rabbit hole...
[19:16:47 CET] <blue_misfit> hey-o! wondering if there's any particular reason some of the recent prores improvements didn't go out with 4.1.1 e.g. https://github.com/FFmpeg/FFmpeg/commit/e4788ae31b2e9af45d11f4bf4498c075dcc25a6c
[19:17:43 CET] <JEEB> if it's a bug fix then it sounds like it could. most likely nobody just took notice of it
[19:18:21 CET] <blue_misfit> there were a bunch of improvements in the last couple months, one of the biggest of which is 12 bit decoding for 4444 / 4444 XQ
[19:18:32 CET] <JEEB> improvements are different from bug fixes
[19:18:42 CET] <blue_misfit> indeed
[19:18:46 CET] <JEEB> I wouldn't be surprised if improvements wouldn't go out into a frozen branch
[19:18:53 CET] <JEEB> but bug fixes probably should
[19:18:57 CET] <blue_misfit> I see
[19:19:19 CET] <blue_misfit> any sense of when the next release will be?
[19:20:06 CET] <BradleyS> that one does seem like a bug fix, michaelni ^ https://github.com/FFmpeg/FFmpeg/commit/e4788ae31b2e9af45d11f4bf4498c075dcc25a6c
[19:20:26 CET] <BradleyS> (consider for 4.1.x)
[19:20:53 CET] <BtbN> philipl, imagine if they made cuvid output what ffmpeg expects. The chaos.
[19:21:07 CET] <JEEB> :D
[19:23:24 CET] <blue_misfit> TBH this is what I really care about: https://github.com/FFmpeg/FFmpeg/commit/6a583261ea631e6da5593c4477e36fde9bc731ab
[19:23:36 CET] <blue_misfit> plus any related commits
[19:23:46 CET] <blue_misfit> thx again for these, kierank :)
[19:27:52 CET] <blue_misfit> just curious if there's known issues with 12 bit prores decoding or if it's generally considered safe to use
[19:31:59 CET] <michaelni> BradleyS, backported e4788ae31b2e9af45d11f4bf4498c075dcc25a6c to release/4.1 locally
[19:32:32 CET] <michaelni> blue_misfit, backports are for bugfixes not new features
[19:33:04 CET] <BradleyS> ty
[19:33:12 CET] <blue_misfit> understood, not expecting this to get backported :)
[19:34:28 CET] <blue_misfit> just curious whether anyone is aware of any known issues with this new feature since I'll probably be pinning my CI/CD to use the most recent commit
[19:34:41 CET] <BtbN> philipl, the patches are also not checking the version. I'd like to keep compat for older drivers.
[19:43:39 CET] <philipl> Yeah. So I'm going to refresh my old patches with the official SDK and resend.
[20:03:11 CET] <BtbN> philipl, the SDK update they sent is also in itself incomplete. That's so weird.
[20:03:30 CET] <BtbN> I copied over the headers from the official download, and it has more stuff
[20:03:43 CET] <JEEB> funky
[20:21:04 CET] <philipl> BtbN: yes. I've been comparing too.
[20:21:21 CET] <philipl> I would recreate from scratch
[20:23:24 CET] <BtbN> Nah, it was fine. Patch and ammendments are already merged.
[20:28:30 CET] <philipl> cheers
[23:08:21 CET] <blue_misfit> were there any breaking / non backwards compatible changes with 4.0 vs 3.4?
[23:08:44 CET] <JEEB> did the 0 != EOF change come up with that?
[23:08:55 CET] <JEEB> I think that's the largest change in the last year+
[23:09:05 CET] <JEEB> which changes how the API should be utilized
[23:09:20 CET] <blue_misfit> okay, but if I'm just using the CLI I should be okay, right?
[23:09:41 CET] <blue_misfit> (I will of course test my stuff, just wondering if there's known stuff to be aware of)
[23:10:04 CET] <JEEB> not that I know, cli in general hasn't changed too much unless someone accidentally broke something
[23:10:17 CET] <blue_misfit> I figured as much, thx!
[23:10:34 CET] <JEEB> because generally removing options needs to first warn, and only then remove
[23:10:41 CET] <blue_misfit> I'm sure you guys hear this a lot but this is an _AMAZING_ project. Thank you for so much hard work
[23:11:35 CET] Action: JEEB wonders on which level MMTP should be implemented
[23:12:01 CET] <JEEB> because the "files" that recorders do are literally the data that comes off of the receiver, which is just IP over broadcast
[23:12:10 CET] <JEEB> so suddenly your multimedia application has to be able to do IP
[23:20:13 CET] <BtbN> philipl, it looks like the nvdec hevc decoder also needs some work now. There are a bunch of new fields it needs to fill.
[23:34:09 CET] <philipl> BtbN: I have that done locally
[23:34:57 CET] <BtbN> It's also interesting that they introduced a version check macro for nvdec. Inside the header. So the macro can't be used, because it's not in the older headers.
[23:35:36 CET] <philipl> #ifndef? :-P
[23:48:40 CET] <J_Darnley> I need to stop reading the user list.  All it does is make me angry.
[23:49:13 CET] <J_Darnley> On a side note I cant believe it is nearly 9 years since I first proposed colored log messages.
[23:49:23 CET] <JEEB> :D
[23:53:13 CET] <feliwir_> Don't we have colored log messages?
[23:54:19 CET] <cone-532> ffmpeg 03Marton Balint 07master:dd6dd49f1d5e: avformat/mpegts: fix charset of type 0x11
[23:54:19 CET] <cone-532> ffmpeg 03Marton Balint 07master:b35843e3913d: avformat/mpegts: also convert strings without a specified encoding to UTF-8
[23:54:19 CET] <cone-532> ffmpeg 03Marton Balint 07master:dbfd042983ee: avformat/utils: parse some stream specifiers recursively
[23:54:45 CET] <J_Darnley> Yes but users still ignore them
[23:56:19 CET] <J_Darnley> Personally I think that is sometimes because too much is colored
[23:57:03 CET] <J_Darnley> Other times if you laser engraved it onto a bat and started hitting them it still wouldnt leave an impression
[23:58:21 CET] <BtbN> It'll probably be something like #if defined(CHECK_MACRO) && CHECK_MACRO(9, 0)
[23:58:33 CET] <BtbN> Normal if semantics do apply there, right?
[00:00:00 CET] --- Thu Feb 14 2019


More information about the Ffmpeg-devel-irc mailing list