Re: Bug#1073983: transition: ocaml

[Stéphane Glondu <glondu@debian.org>]

Dear Emilio,

Le 12/08/2024 à 10:31, Emilio Pozuelo Monfort a écrit :
> > As expected, removal hints will be needed, which I've put in 3 
> > categories:
> >
> > # Independent broken packages
> >
> >    hol-light (#1073882)
> >    ocaml-ffmpeg (#1072440)
> >    ocaml-lo (#1075329)
> >    pa-ounit (#1073907)
> >    ppx-tools (#1078383)
> >    sks (#1073911)

What about those?

> > # libguestfs related packages (#1078470)
> >
> >    guestfs-tools
> >    libguestfs
> >    virt-v2v
> >    guestfs-tools
> >    ceilometer-instance-poller
> >    ironic-python-agent
> >    nbdkit
> >    virt-p2v
> >    qemu-web-desktop
> >    oz
> >    kworkflow
> >    virtnbdbackup
>
> I see the bug has a patch. I'd rather it gets fixed rather than removing 
> all of that, if possible.

Done.

> > # coq related packages (#1078252)
> >
> >    aac-tactics
> >    coq
> >    coq-bignums
> >    coq-corn
> >    coq-deriving
> >    coq-doc
> >    coq-dpdgraph
> >    coqeal
> >    coq-elpi
> >    coq-equations
> >    coq-ext-lib
> >    coq-extructures
> >    coq-gappa
> >    coq-hammer
> >    coq-hierarchy-builder
> >    coq-hott
> >    coq-interval
> >    coq-iris
> >    coq-libhyps
> >    coq-math-classes
> >    coq-menhirlib
> >    coq-mtac2
> >    coqprime
> >    coq-quickchick
> >    coq-record-update
> >    coq-reduction-effects
> >    coq-reglang
> >    coq-relation-algebra
> >    coq-serapi
> >    coq-simple-io
> >    coq-stdpp
> >    coquelicot
> >    coq-unicoq
> >    coq-unimath
> >    elpi
> >    flocq
> >    mathcomp-algebra-tactics
> >    mathcomp-analysis
> >    mathcomp-bigenough
> >    mathcomp-finmap
> >    mathcomp-multinomials
> >    mathcomp-real-closed
> >    mathcomp-zify
> >    ott
> >    paramcoq
> >    ssreflect
>
> Those should be removed in unstable by ftp-masters, then the removal 
> will be propagated to testing once the transition migrates.

My idea was to remove them now to ease the transition, and let them 
migrate back to testing later, automatically. Despite the big number of 
packages, this is a self-contained cluster.

See also #1078578.


I had a look a currently "REJECTED_PERMANENTLY" in excuses. There were a 
few autopkgtest failures that were relevant, which I fixed (last upload: 
why3 today). Howevever, many of them are related to the fact that i386 
and armhf are no longer native (ocaml 5 dropped support for native 
compilation on 32-bit architectures), hence the corresponding 
uninstallability or autopkgtest issues for coq-related packages are 
irrelevant. Is britney smart enough to cope with this?


Cheers,

--
Stéphane