2014-05-07 03:46:01 +0000 <osa1> (I wish we had an option for Cabal for just type checking)
2014-05-07 03:46:03 +0000 <luite> right, you can use the --ghc-options switch for cabal
2014-05-07 03:52:55 +0000 <osa1> great, I added a -ffast-build flag and now I have 75 secs installation time
2014-05-07 03:55:35 +0000 <osa1> why ghcjs --make first generates native object file before generating JS files? is that compiled file used in some part of JS backend?
2014-05-07 03:56:00 +0000 <carter> osa1: hows TH work in ghcjs? :)
2014-05-07 03:57:22 +0000 <luite> yep TH, also you can use it as a GHC replacement with the native-executables flag
2014-05-07 03:57:44 +0000 <luite> (when building cabal packages, these things are also used for the setup executable)
2014-05-07 03:58:19 +0000 <osa1> yeah thanks that makes sense. I'll check the source to see how is generated .o used in GHCJS backend. I don't know what's inside a GHC-generated .o file right now.
2014-05-07 03:58:50 +0000 <luite> you can use --print-deps and --print-obj to get the info from the js_o fils
2014-05-07 03:59:01 +0000 <luite> the .o files are regular native object files for your platform
2014-05-07 04:00:22 +0000 <luite> ok boarding has started for my flight to sydney, see you in 4h or so, stopover there and then the second leg to brisbane for lambdajam
2014-05-07 04:01:18 +0000 <osa1> luite: thanks and have fun :)
2014-05-07 04:30:23 +0000diamium(~chatzilla@bioeng1099.bioeng.auckland.ac.nz) (Ping timeout: 240 seconds)
2014-05-07 04:50:53 +0000 <osa1> so doMake is called twice, one for compilation to native and one for JS but what flags are being changed between this two calls? I'm guessing that some callbacks are set differently but I can track the code to that point
2014-05-07 04:51:13 +0000 <osa1> (when ghcjs is called with --make)
2014-05-07 04:51:30 +0000 <carter> hes on a plane afaik :)
2014-05-07 04:51:54 +0000 <carter> osa1: btw are we getting a luajit backend for ghcjs this summer?
2014-05-07 04:52:03 +0000 <carter> as a weekend extra bonus thing?
2014-05-07 04:52:04 +0000 <carter> :)
2014-05-07 04:52:09 +0000carterkids
2014-05-07 04:52:46 +0000 <osa1> carter: haha probably not :P you'll need to do with purescript for that :)
2014-05-07 04:52:56 +0000 <carter> why not?
2014-05-07 04:53:49 +0000 <osa1> well if I have a lot of time after finishing the GSoC goals I may consider doing that but I don't expect that at this point
2014-05-07 04:54:06 +0000 <osa1> so it depends on how much time I'll have after GSoC work
2014-05-07 04:54:22 +0000 <carter> ok
2014-05-07 04:55:04 +0000 <osa1> carter: also, why do you need LuaJIT backend for GHC?
2014-05-07 04:55:13 +0000 <carter> osa1: to see if itd be fast :)
2014-05-07 04:55:20 +0000 <osa1> heh fair enough
2014-05-07 04:55:39 +0000 <carter> if i can beat torch7 with numerical-core via luajit via ghcjs
2014-05-07 04:55:42 +0000 <carter> that'd be kinda neat :)
2014-05-07 04:56:01 +0000 <carter> (you'd probably wanna figure out using ghcj's work)
2014-05-07 04:56:11 +0000 <carter> because then you'd have a real jit
2014-05-07 04:56:15 +0000 <carter> otoh, i don't care about jtting
2014-05-07 04:56:19 +0000 <carter> but it would be itneresting
2014-05-07 04:56:24 +0000 <osa1> btw I found what sets GHCJS callbacks for second call of doMake: Compiler.GhcjsPlatform.setGhcjsPlatform.