Eric Bavier <ericbav...@openmailbox.org> skribis: > On Mon, 28 Mar 2016 18:12:05 +0200 > l...@gnu.org (Ludovic Courtès) wrote:
[...] >> It might be useful to honor a ‘cabal-package-name’ property in cases >> where guessing doesn’t work. That can always be added later, though. > > Sure. Let's save that for later. This scheme works for all packages > that we have so far, and seems like it should work for most others on > Hackage that I've seen. Sure. >> >> [...] >> >> > %elpa-updater >> > %cran-updater >> > %bioconductor-updater >> > + %hackage-updater >> >> Make it: >> >> ((guix import hackage) => %hackage-importer) >> >> to be on the safe side. > > I considered that at first, but it seems like that syntax is for > updaters that rely on guile-json (or other modules which might be > missing). I'd prefer not to unnecessarily confuse the situation. I thought (guix import hackage) would depend on guile-json via (guix import utils), but apparently no, so you’re right. >> If you confirm that ‘guix refresh -t hackage’ doesn’t crash, go for it! > > I have done so, and it seems to work quite well. There are a few > packages whose .cabal file fail to parse, so I included the > warning output to incite some future bug-fixing. Awesome! Thanks, Ludo’.
