Hi Mathieu, Mathieu Othacehe writes: > Hello Pierre, > >> While working on tree-sitter I realized our node packages didn't >> cross-compile [0] so I thought I'd give it a go, here's a patch! > > Nice! > >> I'm not sure there's any way to express this restriction in the package >> definition, it would be good to fail in a useful way, any thoughts? I >> suspect a mechanism for this isn't really worth the effort, I don't know >> of any other packages like this. > > Yeah that's a strange limitation, I don't think we have something > similar elsewhere. A restriction that compares the host and target > bitness is maybe not worth it. > >> - (string-append (assoc-ref inputs "python") >> + (string-append (assoc-ref (or native-inputs inputs) "python") > > You have several lines such as this one that are longer than the 78 > columns limit. I did my best to fix those by reworking the configure flags as such: --8<---------------cut here---------------start------------->8--- (let* ((prefix (assoc-ref outputs "out")) (xflags ,(if (%current-target-system) `'("--cross-compiling" ,(string-append "--dest-cpu=" (match (%current-target-system) ((? (cut string-prefix? "arm" <>)) "arm") ((? (cut string-prefix? "aarch64" <>)) "arm64") ((? (cut string-prefix? "i686" <>)) "ia32") ((? (cut string-prefix? "x86_64" <>)) "x64") ((? (cut string-prefix? "powerpc64" <>)) "ppc64") (_ "unsupported")))) ''())) (flags (cons (string-append "--prefix=" prefix) (append xflags configure-flags)))) --8<---------------cut here---------------end--------------->8--- It's still one column over, but I figured it was OK, I couldn't find another way to write it without sacrificing readability, hope that's OK :-). > Other than that, it looks cool, and you can go ahead. Thanks for the review! Pushed with 9f7c4f380fdd86d81c805b72e4d05e9e658d3dc2 and 5ee38c467298091e98fa12be45facdcc63a59a87. Pierre