On 31/01/2024 5:34 pm, Edwin Torok wrote: > > >> On 31 Jan 2024, at 17:17, Andrew Cooper <andrew.coop...@citrix.com> >> wrote: >> >> On 31/01/2024 4:36 pm, Anthony PERARD wrote: >>> On Wed, Jan 31, 2024 at 10:42:49AM +0000, Edwin Török wrote: >>>> We tried bumping to 4.06.1 [1] previously, but OSSTest was holding us >>>> back. >>>> So bump to OCaml 4.05 instead, which should match the version on >>>> OSSTest? >>> Yes, it's looks that's the version osstest can currently use. >>> I've started an osstest flight with this patch series and your other >>> ocaml patch series, and so far osstest seems happy with it. The flight >>> isn't finished but all build jobs succeed, and a lot of the tests jobs >>> as well. >>> >>> So: >>> Acked-by: Anthony PERARD <anthony.per...@citrix.com> >> >> A question, while I think about it. >> >> I understand why we want patch 1. The 4.02 -> 4.03 bump is necessary to >> also compile with 5.0 >> >> But why this 4.03 -> 4.05 bump? There is no other change in this patch. > > > The oldest supported Debian has 4.05, and I can’t find a non-EOL > distro with 4.03 or 4.04 here: https://repology.org/project/ocaml/versions > I also have another series (that I haven’t sent out yet) which would > use Dune 1.x in an attempt to use Dune in a way that works on OSSTest, > and the oldest release I can test this on is Debian 10. > > We could keep the minimum at 4.03, but would anything in the CI > actually be able to test that?
Nah - that's a good enough reason to go to 4.05. However, the two patches ought to be folded together, with both parts of the justification given in one commit message. Otherwise to anyone doing git blame, you've entirely hidden the 5.0 build fix with something that just looks like 4.03->4.05 I can sort this out on commit. ~Andrew
