Dear Sebastian,
Le 23/04/2023 à 11:36, Sebastian Ramacher a écrit :
ocaml 4.13.1-4 causes the ABI to change for at least why3. Do you expect
that the ABI of ther ocaml packages also changes? If so, we should
rebuild the ocaml world before the release to not get any surprises if a
ocaml package gets a stable update.
See also:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1030785
The ABI of ocaml-compiler-libs changed (only on native architectures),
with no visible changes in virtual packages, so anything using that is
potentially broken. I (thought I) binNMUed all affected packages (there
were a lot of them), but missed why3 for some reason.
IMHO, the cleanest way to fix the issue for sure is to change the OCaml
ABI advertised in the virtual package name. But that means an amount of
work similar to an OCaml transition. Do we really work this kind of move
at this stage of the freeze? I don't think so.
A pretty good approximation to checking that everything is fine is to
mass-rebuild everything (as Lucas Nussbaum does regularly), identify the
(few, I expect) packages that FTBFS, and binNMU them (+ maybe some of
their dependencies). I suspect why3 is special because it embeds modules
provided by ocaml in a plugin (dh_ocaml's --nodefined-map is suspicious
in this context) but this situation should be rare.
Cheers,
--
Stéphane