00:20:09 <stepcut> alpounet: we might need several variants
00:21:42 <stepcut> At some point in time we have  SourceInfo
00:21:53 <stepcut> we need to use that to generate a .tar.gz
00:22:18 <stepcut> we can do that by adding an sdist command to .Cabal
00:22:49 <stepcut> then we want to add that .tar.gz to LocalHackage copying it somewhere, extracting the .cabal file from it ?
00:22:56 <stepcut> or..
00:23:16 <stepcut> maybe we extend the SourceInfo type so that it also has a reference to the .tar.gz ?
00:24:29 <stepcut> i think we are going to need one function that just takes a .tar.gz and adds it to the repo by copying it into place
00:24:48 <alpounet> together with the package name and the cabal file, yes.
00:24:59 <stepcut> then we also need some other function that takes a SourceInfo and generates the .tar.gz and calls addPackage
00:25:37 <stepcut> we can always extract the .cabal from the .tar.gz if we don't have it around.. but it would be  more efficient if we don't have to
00:25:41 <stepcut> brb
00:25:54 <alpounet> yeah, since we're supposed to have it around to anyway
00:28:25 <stepcut> yes.. when you are getting the source from the Source fetcher
00:28:35 <stepcut> it is possible we may import source some other way in the future
00:28:39 <stepcut> but we can worry about it then
00:31:29 <stepcut> seems 'safer' to import the .tar.gz with out a .cabal so that you can't accidentally mix and match the wrong things.. but it shouldn't be a problem
00:48:35 <alpounet> stepcut, so i extract the package and get the .cabal file from there?
00:54:49 <stepcut> in theory
00:54:57 <stepcut> we can just pass it in with the .tar.gz for now though
00:55:25 <alpounet> yes
00:55:36 <alpounet> and i just take the cabal file out
00:55:54 <alpounet> (that's what hackage does, not sure we actually *need* this)
00:56:05 <alpounet> cabal-install does i guess anyway
00:56:32 <stepcut> we need the contents of the cabal to generate the index.tar
00:58:22 <alpounet> well yeah
00:58:30 <alpounet> through SourceInfo
16:45:14 <Lemmih> Found a bug in snap.
19:59:12 <stepcut> ooo
22:04:42 <stepcut> alpounet: looks like the haddock docs for some core packages like base, ghc-prim, etc, can not be built by downloaded the source from hackage.. so we need to do something to handle those packages
22:06:10 <alpounet> stepcut, what's the problem with doing so?
22:06:27 <stepcut> alpounet: they don't build or don't really exist
22:06:50 <stepcut> base, for example, can not actually be built using cabal.. only as part of the normal ghc build process
22:07:21 <stepcut> I am thinking/hoping, that we can some how extract the haddock docs from haskell platform or something
22:07:55 <alpounet> hmm
22:10:33 <stepcut> annoying.. but..
22:10:46 <stepcut> the other option would be to fix out  how to cross link to the versions on hackage
22:11:13 <stepcut> but then you don't have everything local which has tradeoffs
22:12:55 <alpounet> yeah
22:13:02 <alpounet> the base docs aren't even stored together with the others
22:13:05 <alpounet> that's pretty annoying
22:14:09 <alpounet> you don't really now exactly where they'll end up being
22:14:45 <stepcut> yeah
22:15:22 <stepcut> anyway, once that issue is dealt with I should have a pretty hacked up, but usuable solution for generating the api docs for the upcoming release
22:15:36 <stepcut> less hacked up than the old system though :)
22:15:49 <stepcut> and one with path forward to awesomeness
22:16:36 <alpounet> heh that's cool
22:16:47 <alpounet> really having all the cross-package doc referencing will be awesome
22:32:08 <alpounet> stepcut, i've been looking at this base doc issue, can't find anything useful :/
22:38:07 <stepcut> yeah
22:38:15 <stepcut> just going to have to hack something up
22:38:45 <stepcut> since we ultimately want to build against multiple different versions of haskell-platform, I think it is ok for us to try to yank the docs out of haskell platform
22:40:59 <alpounet> i guess so yes
23:11:07 <stepcut> don't have many options here :)
23:46:16 <alpounet> yes, ideally we will figure out something a bit more evolution-proof
23:46:21 <alpounet> but well
23:46:29 <alpounet> we'd have to talk to the HP people