Package cudajit.0.5.0

[lukstafi]

cudajit.0.5.0

Bindings to the cuda and nvrtc libraries with a unified interface
Bindings to manually selected parts of lcuda and lnvrtc, with a few types and conversion functions to facilitate use.

---

• Homepage: https://github.com/lukstafi/ocaml-cudajit
• Source repo: git+https://github.com/lukstafi/ocaml-cudajit.git
• Bug tracker: https://github.com/lukstafi/ocaml-cudajit/issues

---

🐫 Pull-request generated by opam-publish v2.3.1

[lukstafi]

Passing CI test target: https://github.com/lukstafi/ocaml-cudajit/actions/runs/11112452563/job/30874592644
I also updated lower bounds.

[shonfeder]

Could you add a contact email here? (This is part of a tighter policy we are trying to ease in.)

[lukstafi]

Done (forgot about this).

[shonfeder]

See #26542 (comment) to explain the CI failures.

As an aside, and for the future, ye need a way to handle CI for this sort of package in the future, as it is not practical to add a x-ci-accept-failure for every platform. I'm opening a separate issue to track this.

[lukstafi]

See #26542 (comment) to explain the CI failures.

As an aside, and for the future, we need a way to handle CI for this sort of package in the future, as it is not practical to add a x-ci-accept-failure for every platform. I'm opening a separate issue to track this.

Note the failures are already attributable to conf-cuda, which would be the primary beneficiary of ocurrent/opam-repo-ci#371 .
So a separate feature request would be to have a more elaborate blame assignment. (I realize it's less trivial than just not blaming a package if its dependency failed to build.)

[lukstafi]

Now also passing CI tests for:

• OCaml 4.12: https://github.com/lukstafi/ocaml-cudajit/actions/runs/11161753741/job/31024963647
• OCaml 5.2: https://github.com/lukstafi/ocaml-cudajit/actions/runs/11161753741/job/31024964021

[shonfeder]

Nice! Thanks for the additional CI results and for the fix here. Your suggestion is also excellent, and I've referred that upstream.