[Ffmpeg-devel-irc] ffmpeg-devel.log.20170805

[burek]

[00:05:16 CEST] <michaelni> thardin, i didnt follow the discussion but why do you need a mandatory option ?
[00:11:26 CEST] <michaelni> AVOptions is a API to access C structs with introspection, enumeration, and much nicer forw/backw compatibility than direct struct access. It isnt the exclusive way to access (in general, it may be in specific cases), normal C style access can co-exist so its difficult for it to detect that a value has not been set except by a special default value
[00:13:13 CEST] <michaelni> can you explain why using a NOT_SET value as default is worse than a flag ?
[00:15:40 CEST] <michaelni> One could of course do both (flag and non valid default) and have the code check that both or neither occurs, not sure thats smethig you would prefer
[00:15:53 CEST] <thardin> michaelni: raw codec2 streams need -mode specified to make sense of
[00:16:23 CEST] <thardin> NOT_SET? didn't see that in there
[00:17:08 CEST] <thardin> ah like the -1 suggestion from BBB. not perfect, but it does work
[00:17:57 CEST] <thardin> with a flag you could make -h full a bit nicer
[00:18:19 CEST] <michaelni> ahh, yes, for help text it would make sense
[00:19:35 CEST] <thardin> a solution that doesn't require much logic is to just allow some extra help string
[00:19:44 CEST] <thardin> for each class
[00:19:52 CEST] <michaelni> btw i realize now we already have mandatory options, like width / height for raw video
[00:20:10 CEST] <thardin> so I could put in my codec2raw_demux_class  .extra_help = "option -mode is mandatory"
[00:21:07 CEST] <michaelni> adding a flag should be super easy if you want, it just would still need the default value invalid and wouldnt do anything besides afecting the help text
[00:21:43 CEST] <thardin> it would be nice not to have to expand the range of "legal" values
[00:22:06 CEST] <thardin> including checks for them. but that's more a "nice to have"
[00:27:26 CEST] <michaelni> you could add a av_opt_validate() that checks that all options are set (not invalid) in a struct. and then call that from the avfilter / avcodec / avformat core. I might be forgetting something but i think theres no existing code that gets executed from avoptions after all is set that could check easily
[00:28:19 CEST] <michaelni> av_opt_validate() could print a message / warning for not set mandatory options
[00:28:30 CEST] <michaelni> and return an error
[00:29:03 CEST] <thardin> right, just before init() is called or so
[01:54:22 CEST] <ZeroWalker> still having problems with swscale. I think it should be all fine, i get a packet and have the format context, which should have all the data i need. I basically just do av_image_fill_arrays and align to 1 (for safety), but still it fails. The destination is allocated with the resuled format size
[01:54:42 CEST] <ZeroWalker> http://paste.awesom.eu/xVkm
[02:07:02 CEST] <DHE> email to the list is still the way for patch submissions?
[02:07:22 CEST] <JEEB> yes
[02:13:44 CEST] <DHE> man I hope I don't botch this...
[02:32:32 CEST] <DHE> tfw a patch takes a feature away and didn't fix the documentation when it did it
[02:36:22 CEST] <ZeroWalker> No ideas, not missing anything obvious;o?
[02:38:34 CEST] <DHE> "tfw" is one of those meme things. but it did happen
[02:39:39 CEST] <DHE> intentionally presented as though the original patch was bad in both concept and execution, partially because I personally don't like taking that feature away.
[10:39:58 CEST] <cone-138> ffmpeg 03Paul B Mahol 07master:c47491041466: doc/filters.texi: add yet another laplacian edge detector
[17:29:43 CEST] <durandal_1707> what about if i write lavfi edit preview app in qt?
[17:33:03 CEST] <atomnuker> gtk or death
[17:33:23 CEST] Action: JEEB inserts a goject into atomnuker 
[17:33:26 CEST] <JEEB> *gobject
[17:33:36 CEST] <durandal_1707> death?
[17:34:37 CEST] <atomnuker> JEEB: gobject or not, it still looks the best, supports the most platforms and backends and won't kill you with dependencies
[17:34:58 CEST] <JEEB> have you looked at GTK+ text boxes in windows?
[17:35:14 CEST] <atomnuker> they can't be worse than wx or qt
[17:35:26 CEST] <JEEB> I'm not saying it won't work but non-*nix is very much a non-priority for GTK+
[17:35:36 CEST] <JEEB> so it really depends on one's priorities
[17:35:46 CEST] <JEEB> and yes, wxwidgets is awful, and Qt is also awful in a different way
[17:36:11 CEST] <JEEB> but so far for me Qt's awful emulation of native is the least awful. but that's just my experience and opinion
[17:36:37 CEST] <JEEB> gtk+ is very good on *nix and if you don't care about anything else it's a good selection since most DEs base on GTK+ or similar
[17:40:30 CEST] <durandal_1707> i will use qt
[17:50:15 CEST] <JEEB> yea, I didn't expect random comments on the internets to sway you :D
[18:25:52 CEST] <iive> just don't make ffmpeg depend on qt
[18:28:57 CEST] <nevcairiel> gtk apps look terrible pretty much anywhere, they just blend into the uglyness of full gtk desktops so people dont notice. at least wx uses native widgets on windows =p
[18:29:53 CEST] <BtbN> wx is just broken in general
[18:30:04 CEST] <BtbN> After seeing the dolphin guys rage about it, I wouldn't want to touch it
[18:30:13 CEST] <iive> link?
[18:30:30 CEST] <BtbN> dolphin-emulator. GameCube/Wii emu
[18:30:51 CEST] <JEEB> yea, Myrsloik also handles wx
[18:31:08 CEST] <JEEB> there's plenty of stuff about how awful its "write once, debug everywhere" ratio is
[18:31:45 CEST] <JEEB> Aegisub the subtitle editor went for wx back in the day which might have been a good idea then.
[18:36:54 CEST] <Compnn> i'm probably the only one that doesnt care about gtk / qt... :P
[18:49:15 CEST] <Blubberbub_> I like Tk. But i've only ever used it from Python, so...
[18:50:06 CEST] <JEEB> Compn: it's generally not that people care so vehemently about the framework itself, but as you gain experience in where they work and don't work you start getting an opinion. of course, if stuff just works for the use cases you need on the systems you care about - there really is no discussion about anything in that case.
[18:51:54 CEST] <Compn> i was just talking about the uglyness
[18:52:16 CEST] <Compn> which i understand if its not brushed metal and clear fonts... people have a fit
[19:04:11 CEST] <JEEB> I remember someone mentioning that the post/pre incrementing in loops is a C++ thing? was there a thing written about that somewhere?
[19:05:43 CEST] <JEEB> https://google.github.io/styleguide/cppguide.html#Preincrement_and_Predecrement
[19:05:46 CEST] <JEEB> ah
[21:10:22 CEST] <cone-093> ffmpeg 03Paul B Mahol 07master:2cc56741b1f4: avfilter: add floodfill filter
[23:24:36 CEST] <cone-093> ffmpeg 03James Cowgill 07master:013ec23cbe5d: swscale: fix gbrap16 alpha channel issues
[23:24:37 CEST] <cone-093> ffmpeg 03Steven Siloti 07master:949debd1d1df: avformat/utils: fix memory leak in avformat_free_context
[00:00:00 CEST] --- Sun Aug  6 2017