Align QCheck 0.9 and 0.10 versions

[jmid]

The qcheck* opam files allow mixing versions across releases such as

  - install   qcheck        0.10   [required by fadbadml]
  - install   qcheck-core   0.21.2 [required by qcheck]
  - install   qcheck-ounit  0.21.2 [required by qcheck]

which ends up triggering a fadbadml lower bound failure in #24702:

#=== ERROR while compiling fadbadml.0.1.2 =====================================#
# context              2.2.0~alpha3~dev | linux/x86_64 | ocaml-base-compiler.5.1.0 | file:///home/opam/opam-repository
# path                 ~/.opam/5.1/.opam-switch/build/fadbadml.0.1.2
# command              ~/.opam/opam-init/hooks/sandbox.sh build dune build -p fadbadml -j 71 @install @runtest
# exit-code            1
# env-file             ~/.opam/log/fadbadml-7-6ebe6b.env
# output-file          ~/.opam/log/fadbadml-7-6ebe6b.out
### output ###
# (cd _build/default && /home/opam/.opam/5.1/bin/ocamlc.opt -w -40 -g -bin-annot -I test/unit/.run_test.eobjs/byte -I /home/opam/.opam/5.1/lib/bytes -I /home/opam/.opam/5.1/lib/ocaml/unix -I /home/opam/.opam/5.1/lib/ounit2 -I /home/opam/.opam/5.1/lib/ounit2/advanced -I /home/opam/.opam/5.1/lib/qcheck -I /home/opam/.opam/5.1/lib/qcheck-core -I /home/opam/.opam/5.1/lib/qcheck-core/runner -I /home/opam/.opam/5.1/lib/qcheck-ounit -I /home/opam/.opam/5.1/lib/seq -I /home/opam/.opam/5.1/lib/stdlib-shims -I src/.fadbad.objs/byte -no-alias-deps -open Dune__exe -o test/unit/.run_test.eobjs/byte/dune__exe__Run_test.cmo -c -impl test/unit/run_test.ml)
# File "test/unit/run_test.ml", line 43, characters 16-39:
# 43 |       match res.QCheck.TestResult.state with
#                      ^^^^^^^^^^^^^^^^^^^^^^^
# Error: Unbound record field QCheck.TestResult.state

For all qcheck* releases from 0.11 and above their versions are aligned.
This PR extends that range to uniformly include 0.9 and 0.10 too.

[raphael-proust]

This MR is a strict improvement ("consider for merge").

But there are some interesting errors in the CI:

qcheck-alcotest.0.10
lower-bounds (failed: The type constructor Alcotest.test_case expects 0 argument(s),)

We could use the opportunity to fix those, or we can do it in a follow-up MR.

[jmid]

I agree Raphael!

That one happens when testing qcheck-alcotest with alcotest.0.5.0, so I suspect qcheck-alcotest is missing a proper alcotest lower bound...
There are several other failed: Unbound value ... errors indicating missing lower bounds in other packages using qcheck.

[jmid]

I noticed an alcotest lower bound for qcheck-alcotest.0.9 but none for 0.10. 9bcbc1a therefore adds one.
Versions 0.11-0.18.1 were unfortunately also missing lower bounds so c0fe758 adds them.

Finally I spotted multiple opam lint warnings. f5447d2 fixes the don't-finish-synopsis-with-a-dot one.

[jmid]

The latest commit gets the 0.9-0.17 packages linting cleanly

[jmid]

From qcheck.0.9 the package was split into a qcheck-core package - and a backward compatible qcheck package.
This unfortunately opens for weird situations found by the opam lower bound search such as this one https://opam.ci.ocaml.org/github/ocaml/opam-repository/commit/8607834eb613eb7bac64c9ccbcbb86dd1a0053fa/variant/compilers,4.14,qcheck-alcotest.0.9,revdeps,pratter.1.2.1
allowing for

  - install qcheck-alcotest    0.9
  - install qcheck-core        0.9
  - install qcheck             0.7

I guess this can be solved by either

• marking qcheck-core.0.9 and greater as conflicting with qcheck < "0.9"
• marking qcheck < "0.9" as conflicting with qcheck-core

I've gone for the latter in d545271.
While fiddling in these ancient opam files I spotted that many of them would not pass opam lint.
In 137b25f I thus added missing fields and removed the final synopsis dots.

[jmid]

#24734 fixes missing lower bounds for some of revdeps.

[jmid]

Rebased on master after #24734 has been merged

[jmid]

• base32.1.0.0 and cborl.0.1.0 both hosted on https://inqlab.net/git both failed to get sources with a 404
• lbvs_consent.{2.1.1,2.1.2} failures which should be fixed by Patch lbvs_consent 2.1.1-2.2.0 to support openbabel-3 #24677
• ppx_regexp.0.5.1 failed to get sources with a 403 - which I suspect is a temporary glitch

Finally, pratter.3.0.0 failed with Unbound value char_range because the test lower bound for qcheck should not be 0.9 as stated - and undetected - in #24746 but 0.12 as it was set to in #24734 for pratter.2.0.0. CC @gabrielhdt

I'm starting to suspect that this is a coincidence:
Is the reason that we catch this on changes to dependencies (here qcheck) and not on the original package (here pratter) that dependencies are installed with --with-test during a revdeps lower bound search?
Would you know @kit-ty-kate?

If so, extending the lower bound search to test dependencies would be a nice feature to add... 🙏 🙂

[mseri]

Thanks