On Thu, Aug 6, 2009 at 4:15 PM, Will Donnelly<will.donne...@gmail.com> wrote:
> 1. The 'yi.cabal' which I was building from doesn't actually define a 'Yi' > library for the custom configuration to build against. This looks like a > change you made. Well that was extraordinarily silly of me. :) For some reason, when my configs didn't fail to compile because of unknown modules (that is, the Yi modules), I assumed Dyre had fixed this from the old HConf model that requires two compilations for every update (one for the library, one for the binary). I suppose this will remain a necessity, as the config can't use the modules in Yi unless they're exposed through a library. > 2. When run in from the main 'yi' directory, Dyre doesn't know where to > instruct the compiler to look for uninstalled modules. It might be feasible > to parse the cabal file for the 'hs-source-dirs' option, but it might work > just as well to simply 'cd' to the source directory before running without > installing. I don't think we need to worry about this. Rather, just building the yi library should be sufficient; it'll always be available so the config can build using it. > 2a. When run from the 'src' directory, all of the 'Yi' modules seem to be > found properly, but it seems like some additional command-line options will > be required to make Yi rebuild properly. Also it has issues locating the > cabal-generated 'Paths_yi' file. Again, I think this is unnecessary work. I don't expect Dyre to be able to solve the double-compilation problem, and I feel this workaround is fairly awkward. > So in summary, reverting 'yi.cabal' back to its original version should be > sufficient to make it run properly when installed, and I am not entirely > sure how to make it work without installing. From reading over that bit of > the HConf source code, it looks like it simply didn't account for that > situation, and required installation to work. So from that standpoint, the > current behavior is at least *not* a regression. You're correct; the current behavior is the same as in HConf. For some reason I was being silly and thought you had found a solution to having to build the library. I've sent my patch against HEAD to the list. Your version of yi.cabal was a bit stale --- I noticed it was still requiring an older pointedlist than our current dependency --- but I merged the two in what seems to be working. Also, I like the idea of using the XDG-defined config locations, so I've reverted to using the default from Dyre, ~/.config/yi/, unless others complain. Now, recompilation works wonderfully, but arguments don't seem to be passed through to `yi' when running with a config file. I'll investigate and see if I can figure out the problem. Again, thanks for the help! Jeff Wheeler --~--~---------~--~----~------------~-------~--~----~ Yi development mailing list yi-devel@googlegroups.com http://groups.google.com/group/yi-devel -~----------~----~----~----~------~----~------~--~---
