| Message ID | b1b6504d725df23dc910cb04591a203979cdca7b.camel@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [bug#57704,core-updates] guix: packages: Remove #f from inputs when sanitizing. | expand |
| Context | Check | Description |
|---|---|---|
| cbaines/comparison | success | View comparision |
| cbaines/git-branch | success | View Git branch |
| cbaines/applying patch | success | View Laminar job |
| cbaines/issue | success | View issue |
On 09-09-2022 17:56, Liliana Marie Prikler wrote: > This makes it so that new-style inputs can be optional using regular Guile > patterns, e.g. (and (target-x86-64?) rust). Seems useful. > * guix/packages.scm (sanitize-inputs): Filter inputs by identity before adding > labels. Documentation is missing. > --- > Note that this patch was prepared using master, but since it affects the > package record, it needs to go to core-updates. I don' think there should > be a merge conflict here. It does affect the package record, but it doesn't cause any rebuilds, so master should be fine: * There aren't any current uses of #false: (use-modules (guix packages) (gnu packages)) (package (inherit (specification->package "hello")) (inputs (list #false))) ;; guix build -f [...] --> package ‘hello@2.12.1’ has an invalid input * In the absence of #false, the behaviour remains unchanged. * guix/packages.scm is not used by any derivation (except for "guix pull" and the guix package) As a test, I applied the patch and did ‘make && ./pre-inst-env guix build -n libreoffice’, and it turned out I already have it installed. Greetings, Maxime.
Hi Liliana, Liliana Marie Prikler <liliana.prikler@gmail.com> skribis: > This makes it so that new-style inputs can be optional using regular Guile > patterns, e.g. (and (target-x86-64?) rust). I’d rather avoid that and make sure input lists are just plain lists, remaining strict, and keeping the sanitize procedure simple (notably so it can be optimized in common cases). That means we have to live with idioms like: (append (list x y z) (if (target-x86-64?) (list rust) '())) The ‘openmpi’ package has sugar to make that more concise. Thoughts? Thanks, Ludo’.
Hi Liliana, Ludovic Courtès <ludo@gnu.org> writes: > Hi Liliana, > > Liliana Marie Prikler <liliana.prikler@gmail.com> skribis: > >> This makes it so that new-style inputs can be optional using regular Guile >> patterns, e.g. (and (target-x86-64?) rust). > > I’d rather avoid that and make sure input lists are just plain lists, > remaining strict, and keeping the sanitize procedure simple (notably so > it can be optimized in common cases). > > That means we have to live with idioms like: > > (append (list x y z) > (if (target-x86-64?) (list rust) '())) > > The ‘openmpi’ package has sugar to make that more concise. > > Thoughts? Any plans to revisit this, or should we close it?
diff --git a/guix/packages.scm b/guix/packages.scm index 94e464cd01..5bb2e81e18 100644 --- a/guix/packages.scm +++ b/guix/packages.scm @@ -430,11 +430,12 @@ (define %cuirass-supported-systems (define-inlinable (sanitize-inputs inputs) "Sanitize INPUTS by turning it into a list of name/package tuples if it's not already the case." - (cond ((null? inputs) inputs) - ((and (pair? (car inputs)) - (string? (caar inputs))) - inputs) - (else (map add-input-label inputs)))) + (let ((inputs (filter identity inputs))) + (cond ((null? inputs) inputs) + ((and (pair? (car inputs)) + (string? (caar inputs))) + inputs) + (else (map add-input-label inputs))))) (define-syntax current-location-vector (lambda (s)