[Ffmpeg-devel-irc] ffmpeg-devel.log.20191031

[burek]

[06:42:57 CET] <filternet> Hi dudes, i brought you C madness: cOut() in C alpha test version  https://pastebin.com/miw35a1r
[07:38:02 CET] <cone-005> ffmpeg 03Limin Wang 07master:7c872df3e175: avformat/hlsenc: fix the av_dirname path isn't include separator in the end of string
[10:25:34 CET] <cone-005> ffmpeg 03Paul B Mahol 07master:86a42e954e8e: avfilter/af_afade: start crossfading only when first stream reached end
[10:25:47 CET] <j-b> hello
[10:30:35 CET] <durandal_1707> nobody's here
[10:32:40 CET] <j-b> durandal_1707: you are.
[10:47:58 CET] <cone-005> ffmpeg 03Paul B Mahol 07master:b414cff63040: avfilter/vf_median: add radiusV option
[11:31:55 CET] <JEEB> huh, someone actually asked for AC-4
[11:36:14 CET] <durandal_1707> nothing to see, move on.
[11:56:28 CET] <j-b> I do.
[13:43:15 CET] <JEEB> did we have XML writing code with attributes somewhere?
[13:43:26 CET] <JEEB> like <thing attr1 attr2>
[13:43:48 CET] <JEEB> mostly thinking how to handle this best with the spacing etc, and if there's existing solutions we already have
[13:45:11 CET] <thardin> ... there's an xml parser in the codebase now?
[13:45:25 CET] <thardin> ah writing
[13:45:29 CET] <thardin> yeah I think ffprobe has one?
[13:45:53 CET] <JEEB> and yes, the code base does utilize libxml2 already, but I'm 100% not sure if I want to touch that for just writing
[13:45:56 CET] <JEEB> ah yes
[13:46:07 CET] <JEEB> I got looking at dashenc and that wasn't nice
[14:41:35 CET] Action: JEEB has a crazy idea for using av dicts for attributes
[16:22:37 CET] <JEEB> pfft. at least I'm gaining experience in ASS splitting
[16:23:33 CET] <gnafu> {:-]
[16:25:35 CET] <JEEB> so basically since a single styled section is a <span></span> in TTML
[16:26:18 CET] <JEEB> I have to have an on-going style override buffer
[16:27:00 CET] <JEEB> then when I get the text callback, I write it out.
[16:27:24 CET] <JEEB> in case of cancel_overrides or end I reset it
[16:28:28 CET] <JEEB> although that will end up extra fun with <style overrides>TEXT<more style overrides>
[16:28:36 CET] <JEEB> (with TEXT at the end)
[16:30:20 CET] <JEEB> in which case I have to keep the current overrides still around (cached), quit the span, and update the cached style with the new override and be ready to write it out in case more text is received
[16:35:03 CET] <JEEB> ah no
[16:35:31 CET] <JEEB> I can actually have spans within spans
[16:38:20 CET] Action: JEEB checks if xml:id has to be unique within the document
[16:39:33 CET] <JEEB> > The values of all attributes of type ID (which includes all xml:id attributes) within a document are unique.
[16:59:26 CET] <Lynne> durandal_1707: does the new channel layout api make ambisonics easier and able to be properly described?
[17:08:55 CET] <durandal_1707> Lynne: yes, by adding new ambisonic channel order enum
[17:10:10 CET] <durandal_1707> Lynne: i need to wait for some other developer with his channel layout patches to be posted around VDD days, because that certain other anonymous is actually paid to do it, unlike poor me
[17:11:15 CET] <Lynne> so you know of someone else who's is working on ambisonics or just waiting for a drive-by patch?
[17:11:50 CET] <Lynne> (the google and china people won't add the API because they're apparently uninterested in properly described ambisonics)
[17:12:43 CET] <durandal_1707> Lynne: i do not know anonymous guy at all,  i just know because someone else told me that, koda
[17:16:24 CET] <Lynne> huh, nice
[17:16:54 CET] <Lynne> how are you awake at such time btw? jetlag?
[17:18:35 CET] <durandal_1707> what time? I'm in UTC
[17:31:35 CET] <Lynne> oh, I thought you were in japan
[17:31:59 CET] <durandal_1707> haha, i'm still lost in Tokyo
[18:03:52 CET] <Lynne> exit google street view's VR mode.
[20:00:18 CET] <durandal_1707> is there a  way hot at runtime define functions in C?
[20:00:26 CET] <durandal_1707> s/hot/how
[20:01:01 CET] <JEEB> yes, having them in something you load up and then using known symbols to figure them out?
[20:01:09 CET] <JEEB> or what do you exactly mean?
[20:03:00 CET] <durandal_1707> i need to create N functions depending on user supplied number at runtime
[20:03:21 CET] <BtbN> That's not how C works, no
[20:13:23 CET] <kierank> durandal_1707: thats self modifying code like swscale
[20:13:29 CET] <kierank> But just template the user cases
[20:13:40 CET] <kierank> See how I did it for bit depth and intermediate for idct
[20:14:49 CET] <JEEB> -41
[20:16:20 CET] <Lynne> if you don't mind gpu stuff you can use the vulkan code to generate glsl
[20:40:45 CET] <durandal_1707> i can not use for loops in macro?
[20:41:28 CET] <JEEB> why not?
[20:41:56 CET] <JEEB> have the macro have f.ex. { } ?
[20:42:01 CET] <durandal_1707> i mean with for loop define bunch of functions?
[20:54:19 CET] <taliho> durandal: maybe use a linked list with function pointer as member? 
[20:56:20 CET] <durandal_1707> i need to define 32 * 4 functions, how to do it nice way ? without repeating #DEFINE_FUNCTION(0) - #DEFINE_FUNCTION(31)
[20:58:14 CET] <jamrial> macro that expands to four macros that each expand to four macros that etc
[22:48:08 CET] <BBB> what is the option again to not resize the output stream to a common resolution, but keep all images at their input's original resolution (even if they contain resizes)?
[23:04:46 CET] <Lynne> as in disable filter autoinsertion/reinit when image size changes? -reinit_filter 0
[23:12:23 CET] <Lynne> BBB: as in disable filter autoinsertion/reinit when image size changes? -reinit_filter 0
[23:15:20 CET] <BBB> ty
[00:00:00 CET] --- Fri Nov  1 2019