Chris Marusich <cmmarus...@gmail.com> skribis:
> l...@gnu.org (Ludovic Courtès) writes:
>> (define-public guix guix-devel)
>> +(define-public guix-for-environment
>> + ;; We provide this pseudo-package just to allow people to run:
>> + ;; guix environment guix-devel
>> + ;; to reliably get the development environment for Guix (using 'guix' for
>> + ;; this purpose does not work when 'guix' points to the built-from-tarball
>> + ;; package rather than to the built-from-checkout package.)
>> + (package
>> + (inherit guix-devel)
>> + (name "guix-devel")
>> + (source #f) ;not meant to be built
>> + (supported-systems '())))
>> (define (source-file? file stat)
>> "Return true if FILE is likely a source file, false if it is a typical
>> generated file."
>> If that’s fine with you, could you incorporate it in your patch?
> Sounds good to me. I've attached the updated patch to this email and
> mentioned you as co-author. The new package fails to build with a
> cryptic error, but since it isn't meant to be built, that seems OK.
> What is the purpose of temporarily changing the "guix" variable to refer
> to the release package (as in commit 4420940f)? Because it usually
> refers to the development version, wouldn't it be better to make the
> "guix" variable ALWAYS refer to the development version? It isn't clear
> to me why we need to temporarily change this variable to refer to the
> release version.
Somehow this question turned out to be difficult to answer to me, which
contributed to the delay. ;-)
I think you’re right: in a way, this is a bit silly. This is a purist’s
approach (when we have the tarball, no need to depend on Autoconf et
al.), but it would work just fine if we simply added the ‘native-inputs’
currently on ‘guix-devel’ to ‘guix-0.11.0’ itself.
If there are no objections, I’ll just do that.
Thanks, and sorry for taking so long for such a simple answer!