Re: Order of manifest and overlapping binaries

[Kaelyn]

Hi,

------- Original Message -------
On Monday, October 23rd, 2023 at 6:18 AM, Greg Hogan <code@greghogan.com> wrote:

> 
> On Tue, May 16, 2023 at 4:55 PM Csepp raingloom@riseup.net wrote:
> 
> > Greg Hogan code@greghogan.com writes:
> > 
> > > I could not find documentation on this circumstance or how to resolve.
> > > Both 'parallel' and 'moreutils' produce a 'bin/parallel' and only one
> > > can go in the $GUIX_PROFILE.
> > > 
> > > Creating a container, the latter package overshadows the former
> > > package, as below. Unclear if this is consistent. In my manifest the
> > > former package overshadows the latter (I'd prefer to have parallel's
> > > parallel, but by default I have sorted the listing alphabetically). Is
> > > there a better way to fix this?
> > > 
> > > Greg
> > > 
> > > --8<---------------cut here---------------start------------->8---
> > > $ guix shell --container moreutils parallel which coreutils
> > > [env]$ readlink -f `which parallel`
> > > /gnu/store/xd9kbadmrrbpkjs9vl1v9rhgayfxwgbc-parallel-20230422/bin/parallel
> > > 
> > > guix shell --container parallel moreutils which coreutils
> > > [env]$ readlink -f $(which parallel)
> > > /gnu/store/60zdm9zm0nqm5d97vs30sf4plb2ib5p9-moreutils-0.67/bin/parallel
> > > --8<---------------cut here---------------end--------------->8---
> > > 
> > > This is operating from a recent guix pull:
> > > 
> > > --8<---------------cut here---------------start------------->8---
> > > $ guix describe
> > > Generation 44 May 11 2023 17:02:53 (current)
> > > guix d6f6b57
> > > repository URL: https://git.savannah.gnu.org/git/guix.git
> > > branch: master
> > > commit: d6f6b57766e95d2fa8af63d4460a2b303ca4d867
> > > --8<---------------cut here---------------end--------------->8---
> > 
> > You could create a package that just copies the contents of moreutils
> > to $output, but renames some files, then include the resulting package
> > in your manifest. If moreutils is not propagated from any other
> > package, then you don't even have to do an input rewrite.
> 
> 
> I'm still cutting my teeth on Scheme, and this looks like a simple
> error, but the following from my manifest results in the error below.
> The function accepts a package to inherit from and a list of files to
> rename by appending the package name. This works if I change to pass
> in a single string and create the list within the for-each argument.
> 
> (define (rename-files parent-package files)
> (package/inherit parent-package
> (arguments
> (substitute-keyword-arguments (package-arguments parent-package)
> ((#:phases phases #~%standard-phases)
> #~(modify-phases #$phases
> (add-after 'install 'rename-files
> (lambda* (#:key outputs #:allow-other-keys)
> (let ((out #$output) (name #$(package-name parent-package)))
> (for-each
> (lambda (file)
> (rename-file (string-append out "/" file)
> (string-append out "/" file "-" name)))
> #$files))))))))))
> 
> (define moreutils-decollide
> (rename-files moreutils (list "bin/parallel")))
> 
> --8<---------------cut here---------------start------------->8---
> 
> starting phase `rename-files' error: in phase 'rename-files': uncaught 
> exception: wrong-type-arg #f "Wrong type to apply: ~S" ("bin/parallel") 
> ("bin/parallel") phase` rename-files' failed after 0.0 seconds

This error is because using gexps adds an extra layer of expansion on top of 
normal scheme. The error is coming from the use of "#$files"... specifically in 
the subform "(for-each (lambda (file) ...) #$files)", #$files is replaced--in 
the usage example--with exactly ("bin/parallel"), resulting in:
   (for-each (lambda (file) ...) ("bin/parallel"))

Because it is a gexp, #$file is replaced with ("bin/parallel") and then the 
resulting form is evaluated on the builder. Since ("bin/parallel") now looks 
like a function call, it tries to treat it as one. The three main options that 
I know of are to
1) quote the argument when calling rename-files so that "list" is first in the 
literal list:
    (rename-files moreutils '(list "bin/parallel"))
2) quote the list within the gexp:
    (for-each (lambda (file) ...) '#$files)
3) build the list within the gexp:
    (for-each (lambda (file) ...) (list #@$files)

In my opinion the second option is probably the easiest and safest to work 
with. #1 and #3 both suffer from needing to specially craft the incoming 
argument to handle being evaluated twice. For #1, the argument has to be a list 
after being evaluated twice (the first evaluation is of the quote, the second 
occurs after the gexp was expanded and calls list with the string arguments). 
For #3, the expectation of a list is more explicit, but the argument has to 
evaluate to a list where all of the elements have to then evaluate to something 
meaningful (not too much of an issue for this case as strings evaluate to 
themselves). #2 should only require that the evaluated argument has a printable 
representation that can be read back in, which at least to me feels more 
natural to work with.

Hope my early morning explanation helps!

Cheers,
Kaelyn

> Backtrace:
> 9 (primitive-load "/gnu/store/qrj9l194a552vpg2234xx55k76j…")
> In guix/build/gnu-build-system.scm:
> 908:2 8 (gnu-build #:source _ #:outputs _ #:inputs _ #:phases . #)
> In ice-9/boot-9.scm:
> 1752:10 7 (with-exception-handler _ _ #:unwind? _ # _)
> In srfi/srfi-1.scm:
> 634:9 6 (for-each #<procedure 7ffff5b0b140 at guix/build/gnu-b…> …)
> 
> In ice-9/boot-9.scm:
> 1752:10 5 (with-exception-handler _ _ #:unwind? _ # )
> In guix/build/gnu-build-system.scm:
> 929:23 4 ()
> In ice-9/eval.scm:
> 159:9 3 (_ #(#(#(#<directory (guile-user) 7ffff77f7c80>) (#)) …))
> 
> 159:9 2 (_ _)
> In ice-9/boot-9.scm:
> 1685:16 1 (raise-exception _ #:continuable? _)
> 1685:16 0 (raise-exception _ #:continuable? _)
> 
> ice-9/boot-9.scm:1685:16: In procedure raise-exception:
> Wrong type to apply: "bin/parallel"
> --8<---------------cut here---------------end--------------->8---