bug#49827: Error message for missing synopsis in opam importer

[zimoun]

Hi,

I am back from holidays. :-)

On Wed, 11 Aug 2021 at 15:15, Alice BRENON <alice.brenon@ens-lyon.fr> wrote:

> Thanks for your answer Sarah. Simon, I don't know if you have been able
> to make any progress but I wanted to make sure you had seen the patch
> proposal I sent to let the opam importer work from more repositories
> than the few initially defined (opam's official and three for coq):
>
> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=49958

I have seen but not read in details.  I will do.

>> > The package iter is missing a "synopsis" field as can be seen on
>> > https://opam.ocaml.org/packages/iter/ , which when I tried
>> >
>> > guix import opam iter
>> >
>> > yielded the following backtrace:
>> >
>> > Backtrace:
>> >            8 (primitive-load
>> > "/home/alice/.config/guix/current/bin/g…") In guix/ui.scm:
>> >    2185:7  7 (run-guix . _)
>> >   2148:10  6 (run-guix-command _ . _)
>> > In guix/scripts/import.scm:
>> >    120:11  5 (guix-import . _)
>> > In guix/scripts/import/opam.scm:
>> >    104:23  4 (guix-import-opam . _)
>> > In guix/utils.scm:
>> >     752:8  3 (call-with-temporary-output-file _)
>> > In guix/import/opam.scm:
>> >    337:34  2 (_ _ _)
>> > In srfi/srfi-1.scm:
>> >    460:18  1 (fold #<procedure 7f3baca56fe0 at
>> > guix/import/opam.scm…> …) In guix/import/opam.scm:
>> >    193:15  0 (_ _ _)
>> >
>> > guix/import/opam.scm:193:15: Throw to key `match-error' with args
>> > `("match" "no matching pattern" string-pat)'.
>> >
>> >
>> > the final error is raised l.193 of guix/import/opam.scm because
>> > metadata-ref supports various types for a metadata field, but not
>> > the lack of it. As discussed with Maxime Devos on the IRC channel,
>> > it would be helpful to either allow the import of a package with a
>> > missing field (possibly filling it in the output scheme code for
>> > the imported package with some bad value requiring the user to fill
>> > it and causing any build to crash until replaced properly) or at
>> > least to handle that missing field with a more explicit error
>> > message than the above backtrace (something like "Can't import that
>> > package because it's missing such or such field").  

>From my understanding, there is 2 issues:

 - gentle handler for error
 - warn for incomplete metadata

With Jérémy (jeko), we have started to work time to time using
experimental pair-programming to fix the former.  Currently, each
importer uses its own error mechanism and obviously incoherence between
them happens; especially when ’--recursive’.  We are trying to unify
that.

Thanks for the report of this use case. :-)


Cheers,
simon