CI should fail if a package does not present a description or a synopsis

[mseri]

See the initial discussion with @pmetzger here: #12727

[mseri]

Luckily there are not many such packages at the moment:

packages/cohttp-async/cohttp-async.1.1.1/opam
packages/cohttp-lwt/cohttp-lwt.1.1.1/opam
packages/cohttp-lwt-jsoo/cohttp-lwt-jsoo.1.1.1/opam
packages/cohttp/cohttp.1.1.1/opam
packages/cohttp-lwt-unix/cohttp-lwt-unix.1.1.1/opam
packages/cohttp-mirage/cohttp-mirage.1.1.1/opam
packages/cohttp-top/cohttp-top.1.1.1/opam
packages/conf-bap-llvm/conf-bap-llvm.1/opam
packages/conf-bap-llvm/conf-bap-llvm.1.1/opam
packages/conf-bap-llvm/conf-bap-llvm.1.2/opam
packages/crlibm/crlibm.0.3/opam
packages/functoria-runtime/functoria-runtime.2.0.0/opam
packages/mirage-clock-freestanding/mirage-clock-freestanding.1.2.0/opam
packages/mirage-protocols-lwt/mirage-protocols-lwt.1.1.0/opam
packages/mirage-protocols-lwt/mirage-protocols-lwt.1.3.0/opam
packages/mirage-protocols/mirage-protocols.1.3.0/opam
packages/mirage-protocols-lwt/mirage-protocols-lwt.1.0.0/opam
packages/mirage-stack/mirage-stack.1.2.0/opam
packages/mirage-stack-lwt/mirage-stack-lwt.1.2.0/opam
packages/opam-publish/opam-publish.2.0.0/opam
packages/opam-publish/opam-publish.2.0.0~beta/opam
packages/qcheck-core/qcheck-core.0.9/opam
packages/qcheck-alcotest/qcheck-alcotest.0.9/opam
packages/qcheck/qcheck.0.9/opam
packages/qcheck-ounit/qcheck-ounit.0.9/opam
packages/relit-reason/relit-reason.0.0.1/opam

[mseri]

It should also complain if test is used instead of with-test:
#12732

[pmetzger]

I obviously agree this should be fixed. :)

[rjbou]

The test issue is more complicated because of variables handling. Some variables are opam defined but user can also defines its own. Plus, as variables can be undefined, and filter expressed according, we can't raise an error if {test} is present (while packager meant {with-test}), or if the test variable is null.

Note that the opam package upgrader rewrite test and doc variables into with-test and with-doc.

[c-cube]

A warning (in the linter) would still do the trick, right? if someone wants to define test nonetheless they would be able to, but for most users using {test} is probably a mistake. Also, it's the first time I hear of a package upgrader, which is why I was confused by the same problem.

[rjbou]

The two PR merged on master.

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove the stale label, or comment, or this will be closed in 15 days.

[pmetzger]

I'm not sure that closing this was a good idea, merely because it hadn't had much activity of late. The problem remains interesting to fix.