Leo Famulari <l...@famulari.name> writes:
> So, a Go build system needs to somehow create a symlink union of the
> dependency graph, making the dependencies available at the correct paths
> relative to the root of the build environment. AFAICT, we can't figure
> these paths out programatically; they must be supplied by the packager
> in advance, for example:
>
> ------
> (arguments
> `(#:import-path "github.com/AudriusButkevicius/go-nat-pmp"))
> ------
>
> Petter's prototype creates the symlink union, but instead of using the
> #:import-path key, it instead duplicates the import path while listing a
> package's dependencies, like this:
>
> ------
> (inputs
> `(("github.com/AudriusButkevicius/go-nat-pmp"
> ,golang-github-com-audriusbutkevicius-go-nat-pmp)
> ------
[…]
>
> Are there any examples in the code base of accessing the inputs' keys? That
> seems like a better solution to me, but I'm not sure how to achieve it.
Instead of recording this information in the input labels (which I
consider rather inelegant) you could put the key in the properties
field. In the build phase of another package you can access it like
this:
(arguments
`(#:phases …
(lambda _
…
,(assoc-ref (package-properties the-package) 'the-key)
…
)…))
You could also access the arguments of another package with
“package-arguments”. Using the “properties” field isn’t pretty because
it is a free form alist.
You can use “find-tail” to jump to the keyword in “arguments” and then
pick the following value.
--
Ricardo
GPG: BCA6 89B6 3655 3801 C3C6 2150 197A 5888 235F ACAC
https://elephly.net
