Ludovic Courtès writes: > Hi Chris, > > Christopher Baines skribis: > >> Ludovic Courtès writes: >> >>> Good morning, Chris! >>> >>> Christopher Baines skribis: >>> >>>> * gnu/packages/web.scm (guix-data-service): New variable. >>> >>> Woohoo! \o/ >>> >>>> + (inputs >>>> + `(("guile" ,guile-2.2) >>> >>> Guile should be ‘native-inputs’ because we use it for its compiler >>> (which can act as a “cross-compiler” with ‘--target’). >> >> This makes sense, but the scripts will still refer to the guile used to >> build the package, so building for a different architecture won't work >> right? Does the package somehow need to take as input two Guiles, one >> for build time, and one for runtime? > > Yes, we probably need to have Guile both as input and native input. > Well don’t lose your hair on it anyway. I've just gone with making it a native input for now. >>>> + (native-search-paths >>>> + ;; guile-git requires this to be set >>>> + (list (search-path-specification >>>> + (variable "GIT_SSL_CAINFO") >>>> + (file-type 'regular) >>>> + (separator #f) >>>> + (files '("etc/ssl/certs/ca-certificates.crt"))))) >>> >>> It’s a bit of a workaround for >>> , isn’t it? :-) >>> >>> What about either removing it, or adding an “XXX” to the comment? >> >> I can't quite remember, I think I added that a long while ago when >> trying to get the Guix service working. I'm not sure it's a great >> approach. I'll add XXX to the comment though. > > What about removing it altogether and instead setting GIT_SSL_CAINFO in > the service itself? Turns out I was already setting GIT_SSL_CAINFO in the service, so I've removed the native-search-paths from the package definition. I've now pushed both patches, thanks for taking a look! Chris