Add: bos.0.2.1, fmt.0.8.10, rresult.0.7.0, topkg.1.0.4, topkg-care.1.0.4

[dbuenzli]

• Add: bos.0.2.1 home, doc, issues
  Basic OS interaction for OCaml
• Add: fmt.0.8.10 home, doc, issues
  OCaml Format pretty-printer combinators
• Add: rresult.0.7.0 home, doc, issues
  Result value combinators for OCaml
• Add: topkg.1.0.4 home, doc, issues
  The transitory OCaml software packager
• Add: topkg-care.1.0.4 home, doc, issues
  The transitory OCaml software packager

---

bos v0.2.1 2021-10-04 Zagreb

• Require OCaml >= 4.08.
• OS.Dir.create fix function result on existing files. It returned
  non-sensical results. The function now errors as it should
  be. Thanks to Léo Andrès for the report.
• OS.Dir.create fix function returning false instead of
  true when the directory is created with ~path:false.
  Thanks to Léo Andrès for the report and patch.
• OS.File.read support for reading character devices and named
  pipes. Thanks to Rizo Isrof for the patch.

---

fmt v0.8.10 2021-10-04 Zagreb

• Require OCaml >= 4.08. This drops the dependency on the
  stdlib-shims and seq packages.
• Add the [@@ocaml.deprecated] annotation to deprecated
  functions. Thanks to Antonin Décimo for the patch.

---

rresult v0.7.0 2021-10-04 Zagreb

• Require OCaml >= 4.08. This drops the dependency on the result
  compatibility package.
• Users are encouraged to move the the Stdlib.Result module
  available in OCaml 4.08.

---

topkg-care, topkg v1.0.4 2021-10-04 Zagreb

• Remove mentions of Result.result in the code base. We got the
  dependency indirectly through bos and the latest version of the
  latter no longer depends on it.

---

Use b0 cmd -- .opam.publish bos.0.2.1 fmt.0.8.10 rresult.0.7.0 topkg.1.0.4 topkg-care.1.0.4 to update the pull request.

[dbuenzli]

I did dread the removal of the result dependency from rresult :-) That's what we get for having imprecise dependencies.

What do we do, constrain on rresult or add the result package to the dependencies of these packages ?

[dbuenzli]

So I added the result constraints on those packages failing with Result.result. Thanks to @dra27 for providing this snipped to output the failing packages which may be useful for others in the future.

curl -Ls https://opam.ci.ocaml.org/github/ocaml/opam-repository/commit/b41da67f2ede580fbdfdad1b5f5fd140eb547205 | sed -e 's/>/\n/g' | sed -ne 's/^\([^ ]*\) (failed: Unbound.*/\1/p' | sort | uniq

This was then piped to opam-ed as follows:

sed 's/\(\([^\.]*\).*\)/-f packages\/\2\/\1\/opam/g' | xargs opam-ed -i 'append depends "result"'

(the formatting was altered on some packages).

A few packages were also failing on fmt because they are using warn-error (boooo :-), they got a constraint on fmt.

[dbuenzli]

So as @dra27 aptly remarked, adding a result dependency doesn't work since the packages' build systems are not aware of it. :-)

So given $(PKGLIST) as listed by @dra27's rune above the commit was generated with:

opam admin add constraint `rresult<0.7.0' 
$(PKGLIST) | sed 's/\(\([^\.]*\).*\)/packages\/\2\/\1\/opam/g' | git add --
git checkout packages

It would be nice for add-constraint to support specifying the packages on which they should apply directly see ocaml/opam#3077

[dbuenzli]

So given $(PKGLIST) as listed by @dra27's rune above the commit was generated with:
[...]

Just a note to my future self, given the way add-constraint works the above procedure missed the packages that got rresult through a transitive dep :-(. We really want to be able to specify the list of packages to affect in add-constraint.

[dbuenzli]

Ok the next round of results are in. I propose to fix remaining deps in a separate PR to be merged before this one (will do as soon as I get a bit time on this). I don't think it makes sense to run 36k jobs again, I'm not too happy about the vast amount of ressources all that is using.

[mseri]

This looks good, the only remaining issue due to dropping result is the dockerfile-cmd one, if I did not miss anything.
I think we only need to fix that one separately before merging this. We can deal with the lower bounds at lower pace (I have opened an issue on the opam repo to track them)

[dbuenzli]

Why did the CI results disappear ?

[dbuenzli]

It seems they were restarted ?! I don't think we need another round.

[mseri]

I agree, let's stop it while the CI is restarting! Thanks a lot