02:06 < Nick___> hello tee ppl !
02:08 < Nick___> anyone here!
02:08 < yemDX`> me
02:08 < Nick___> ooh good
02:09 < yemDX`> how are u
02:09 < Nick___> good and u?
02:11 < Nick___> have to go bb 4 now
02:11 < yemDX`> bye
11:00 < Naruto-Man> hi tee ppl
11:00 < EastBite> hi Naruto-Man 
11:00 < Naruto-Man> what mod do u like more EastBite
17:11 < KingteEGER> Hey is there aleready an issue on github for the sometimes buggy physics? 1. http://youtu.be/mZXZc9Ho2JQ 0:04 look at the two red player "Arcanios" and "m80116"  2. The same Bug wich manny players use un DDRace but thistime in Vanilla http://youtu.be/l34ijNuZTwg
17:11 < Nimda3> [YouTube] Title: [Show]Teeworlds physics | Rating: 5.00/5.00 | Views: 36
20:59 < heinrich5991> I'd like to try to outsource text rendering to a library in TW
21:00 <@minus> do eeet, fgt
21:01 < heinrich5991> my first attempt would be to send the actual string through the graphic 'pipeline'
21:02 < heinrich5991> and render it in the backend
21:02 < heinrich5991> that means
21:03 < heinrich5991> requesting text width -> text interface -> library
21:03 < heinrich5991> drawing text -> text interface -> graphics pipeline -> library rendering -> putting it onto the screen
21:03 < heinrich5991> (which are the two functions the text interface currently exposes)
21:05 < heinrich5991> matricks: does that sound sane?
22:13 <@matricks> heinrich5991: why are you doing all this?
22:16 <@matricks> heinrich5991: sound like you might just wanna replace parts of the text interface
22:22 < heinrich5991> matricks: I'm doing this because I want to try to get better text rendering into tw
22:22 < heinrich5991> or what do you mean by 'all this'?
22:26 < heinrich5991> replacing parts of the text interface – do you mean I should change the interface?
22:27 <@matricks> noo, just the internals
22:27 <@matricks> you shouldnt let a library draw dirwctly to the backbuffer
22:28 <@matricks> do you wanna render a complete phrase to a texture and render or?
22:29 < heinrich5991> I wanna render a complete line using the library
22:29 <@matricks> so, render to a line to a texture and cache it then
22:30 < heinrich5991> but the lines weren't cached before either
22:30 <@matricks> no, but the glyphs are
22:30 < heinrich5991> I guess the library does that too
22:31 < heinrich5991> I at least have to create a cache (?) object
22:31 <@matricks> so if you are going to pass a complete line to a library you need to cache complete lines
22:32 <@matricks> does thlibrary know about opengl?
22:32 < heinrich5991> no
22:32 <@matricks> then it doesn't cache it on the level that is needed
22:33 <@matricks> you wanna get better kerning or?
22:33 < heinrich5991> I wanna get RTL stuff
22:33 <@matricks> >_<
22:35 < heinrich5991> I knew you were gonna do this :D
22:35 <@matricks> render it normally and flip the screen as a last pass
22:37 <@matricks> and now you think i'm joking, but im not :)
22:37 < heinrich5991> you were not >_<ing because it's i10n?
22:38 <@matricks> I stand by qhen i say that trnaslations was the worat thing I ever done
22:39 <@matricks> (on mobile so spelling is a bit off)
22:41 < heinrich5991> yes, I know that you think that
22:41 < heinrich5991> I don't think that the RTL stuff is that easy though
22:42 < heinrich5991> fun starts when one starts mixing non-RTL and RTL stuff
22:42 < heinrich5991> which is why I'd like to rely on that lib for rendering
22:42 <@matricks> >_<
22:43 <@matricks> the textinterface need to be redone for it
22:43 <@matricks> I dont know about the alignment rules for it either
22:47 < heinrich5991> why does the text interface need a redo for this? it already gets passed a box (which is most of the time correct, but can be fixed)
22:48 <@matricks> to make line caching work I think it might be needed
22:49 < heinrich5991> so, you say one needs caching on opengl level?
22:50 <@matricks> yah
22:51 <@matricks> otherwise there will be a lot of data transfers to the gpu from cpu
22:52 < heinrich5991> okay
23:24 < heinrich5991> matricks: do you think it should rather be 'text interface -> library -> render to texture, cache -> graphic pipeline -> render to screen' or with 'text interface -> graphic pipeline -> library -> render to texture, cache -> render to screen'
23:25 < heinrich5991> the latter case would need a mutex for the library unfortunately
23:25 <@matricks> graphics should be last
23:26 <@matricks> it;s the one that does the quad rendering
23:26 <@matricks> text interface should take care of caching
23:27 < heinrich5991> how does the graphic stuff ensure that the following situation can't happen:
23:28 < heinrich5991> command "render texture A" to pipeline – unload texture A – "render texture A" is executed for the next frame?
23:28 < heinrich5991> e.g. when disconnecting
23:28 <@matricks> the one that does the rendering shouldn't do that
23:29 < heinrich5991> ok, that might be hard though, as you don't know when the final render of texture A happens
23:29 <@matricks> ?
23:31 < heinrich5991> suppose we're sending the command to render a tile to the screen
23:31 < heinrich5991> then the user clicks disconnect
23:31 < heinrich5991> now I'm assuming that we instantly unload the map's textures
23:32 <@matricks> might be, dunno actually
23:32 < heinrich5991> what happens if the graphic rendering hasn't rendered the tile yet?
23:32 <@matricks> disconnect etc should be latched and happen at a beginning of a frame
23:32 <@matricks> wait... doesn;t matter
23:33 <@matricks> everything is queued to the render thread
23:33 <@matricks> including deletes
23:33 <@matricks> so everything is sequenced
23:33 <@matricks> it will render the tiles, then delete the texture etc etc
23:33 < heinrich5991> ah ok
23:34 <@matricks> only the render thread is allowed to make gl calls
23:35 < heinrich5991> ok, that's a clear separation
23:35 <@matricks> yup
23:35 < heinrich5991> do you know off the head how the textures are passed from the main thread to the gl thread?
23:35 < heinrich5991> (i.e. how is it ensured that the data lives long enough)
23:35 <@matricks> by a buffer on the heap
23:35 <@matricks> and the render thread deletes it
23:36 < heinrich5991> ok that works
23:36 <@matricks> I think :P
23:37 <@matricks> https://github.com/teeworlds/teeworlds/blob/master/src/engine/client/graphics_threaded.h
23:37 <@matricks> all the commands are listed there
23:38 <@matricks> line 210, texture create
23:47 < heinrich5991> matricks: is it viable to do this one allocation per different text line rendered?
23:49 <@matricks> heinrich5991: dunno
23:49 <@matricks> might trash the gpu mem
23:49 <@matricks> dunno how much data it will be