[bug#52774] [PATCH] import: elpa: Also check NonGNU ELPA for updates.

[Liliana Marie Prikler]

Am Samstag, dem 25.12.2021 um 15:31 +0100 schrieb Xinglu Chen:
> Liliana schrieb am Samstag der 25. Dezember 2021 um 11:15 +01:
> 
> > Am Samstag, dem 25.12.2021 um 10:44 +0100 schrieb Xinglu Chen:
> > > Liliana schrieb am Freitag der 24. Dezember 2021 um 13:17 +01:
> > > 
> > > > Am Freitag, dem 24.12.2021 um 12:25 +0100 schrieb Xinglu Chen:
> > > > > * elpa.scm (latest-release): Determine the repository based
> > > > > on
> > > > > the URL of the source.
> > > > > (package-from-gnu.org?): Rename to ...
> > > > > (package-from-gnu.org-or-nongnu.org?): ...this.
> > > > > (%elpa-updater): Adjust accordingly.
> > > > > ---
> > > > >  guix/import/elpa.scm | 13 +++++++++----
> > > > >  1 file changed, 9 insertions(+), 4 deletions(-)
> > > > > 
> > > > > diff --git a/guix/import/elpa.scm b/guix/import/elpa.scm
> > > > > index edabb88b7a..038379e01b 100644
> > > > > --- a/guix/import/elpa.scm
> > > > > +++ b/guix/import/elpa.scm
> > > > > @@ -424,7 +424,11 @@ (define (guix-package->elpa-name
> > > > > package)
> > > > >  (define (latest-release package)
> > > > >    "Return an <upstream-release> for the latest release of
> > > > > PACKAGE."
> > > > >    (define name (guix-package->elpa-name package))
> > > > > -  (define repo 'gnu)
> > > > > +  (define repo
> > > > > +    (let ((url (origin-uri (package-source package))))
> > > > > +      (if (string-prefix? "https://elpa.nongnu.org"; url)
> > > > > +          'nongnu
> > > > > +          'gnu)))
> > > > >  
> > > > >    (match (elpa-package-info name repo)
> > > > >      (#f
> > > > > @@ -443,11 +447,12 @@ (define repo 'gnu)
> > > > >          (urls (list url))
> > > > >          (signature-urls (list (string-append url
> > > > > ".sig"))))))))
> > > > >  
> > > > > -(define package-from-gnu.org?
> > > > > +(define package-from-gnu.org-or-nongnu.org?
> > > > >    (url-predicate (lambda (url)
> > > > >                     (let ((uri (string->uri url)))
> > > > >                       (and uri
> > > > > -                          (string=? (uri-host uri)
> > > > > "elpa.gnu.org"))))))
> > > > > +                          (or (string=? (uri-host uri)
> > > > > "elpa.gnu.org")
> > > > > +                              (string=? (uri-host uri)
> > > > > "elpa.nongnu.org")))))))
> > > > >  
> > > > >  (define %elpa-updater
> > > > >    ;; The ELPA updater.  We restrict it to packages hosted on
> > > > > elpa.gnu.org
> > > > > @@ -455,7 +460,7 @@ (define %elpa-updater
> > > > >    (upstream-updater
> > > > >     (name 'elpa)
> > > > >     (description "Updater for ELPA packages")
> > > > > -   (pred package-from-gnu.org?)
> > > > > +   (pred package-from-gnu.org-or-nongnu.org?)
> > > > >     (latest latest-release)))
> > > > >  
> > > > >  (define elpa-guix-name (cut guix-name "emacs-" <>))
> > > > > 
> > > > > base-commit: e8c1562599c7ebee8b7b228237fb0d75d4472a61
> > > > Could we make it so that package-from-gnu.org? becomes elpa-
> > > > package? s.t. (member (elpa-package? pkg) '(gnu nongnu #f)) and
> > > > its
> > > > evaluation can be reused?
> > > 
> > > Wouldn’t it be enough with just '(gnu nongnu)?
> > > 
> > > We could make the ‘elpa-package?’ procedure memoized to avoid
> > > evaluating the same thing twice.
> > The reason for #f is so that packages that are neither from GNU nor
> > NonGNU ELPA are still excluded from the updater.
> 
> I am not sure I follow.  If ‘elpa-package?’ returned #f, the
> following would return #f.
> 
>   (member (elpa-package? pkg) '(gnu nongnu))
> 
> Meaning that PKG is not hosted on GNU or NonGNU ELPA, and thus,
> should not be checked for updates.
> 
> When including #f in the list, the value returned by the ‘member’
> expression would be '(#f), which means that PKG would be checked for
> updates, even though it shouldn’t.
I think you're misunderstanding what I'm saying, but that might be
because I worded it badly.  My suggestion was to make elpa-package? a
"predicate with meaning", which returns #f if PKG is not an ELPA
package, and a truthy value otherwise, said truthy value being the
symbol 'gnu or 'nongnu at the moment (which could be expanded to e.g.
include 'melpa or 'some-other-law-abiding-elpa in case the ELPA updater
is found to work for those as well).  Saying that it satisfies the
above member function was a shorthand for that.