[bug#57659] [PATCH v3] gnu: Add blesh.

[kiasoc5 via Guix-patches via <guix-patches@gnu.org>]

On Thu, Sep 15 2022, 05:26:30 PM +0200
Maxime Devos <maximedevos@telenet.be> wrote:

> On 14-09-2022 10:40, Christopher Baines wrote:
> > 
> > kiasoc5 <kiasoc5@disroot.org> writes:
> >   
> >> Tests pass now.
> >>
> >> * gnu/packages/bash.scm (blesh): New variable.
> >> ---
> >>   gnu/packages/bash.scm | 45
> >> +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 45
> >> insertions(+)  
> > 
> > Thanks. I've tweaked the name of one of the phaess
> > (use-LANG-for-tests rather than use-LC_ALL-for-tests) as that
> > seemed more correct. I also removed some inputs that were redundant
> > (coreutils is provided by the gnu-build-system, so is an implicit
> > input), and git-minimal didn't seem to be required.  
> 
> gnu-build-system only provides a _natively compiled_ coreutils (for 
> --system), not a cross-compiled coreutils (for --target).  As such,
> it is not redundant (unless it compile-time only (*), the build
> system isn't clear to me).

Readlink should be required for runtime, how about substituting
"PATH=/bin:/usr/bin readlink" -> "readlink" instead of the path to
readlink?

>  >+                        (add-after 'unpack 'make-readlink-work
>  >+                          (lambda _
>  >+                            (substitute* "ble.pp"
>  >+                              (("PATH=/bin:/usr/bin readlink")
>  >+                               (search-input-file %build-inputs
>  >+
>  >"/bin/readlink")))))  
> 
> You're mixing G-exp (new) and %build-inputs (old).  How about
> (search-input-file inputs "bin/readlink") or (which "readlink") 
> (depending on whether inputs or native-inputs is appropriate) instead?

If we keep coreutils as an input then it would be search-file-inputs.

--