Fix Rust Releases

[AkshatRai07]

Closes #555

In this PR we,

• Fix MacOS linking errors.
• Remove staticlib generation.
• Make eko versioning dynamic, hence changing the .whl name.

[Copilot]

Pull request overview

This PR adjusts the Rust C-API build/test setup to address macOS dynamic linking issues and to stop producing a staticlib artifact during releases.

Changes:

• Configure the C-API crate to build only as a cdylib (no staticlib).
• Add a macOS-specific post-link step in the C-API test runner to rewrite an absolute dylib install name to use @rpath.
• Update the local build-capi task command used to generate the C-API distribution.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
pyproject.toml	Updates the build-capi task command for producing the C-API dist.
crates/ekore_capi/tests/run_tests.sh	Adds a macOS-specific install-name rewrite to make test binaries find the dylib via @rpath.
crates/ekore_capi/Cargo.toml	Removes staticlib from the crate’s crate-type list to stop generating static libraries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[AkshatRai07]

Doubt: Should Rust, C-API, and Python API tests be run on all OS and Architecture combinations which we will use when publishing?

Suggestion: Please try changing the permission of PYPI_TOKEN, and run the entire Maturin workflow, only then merge this PR. Else we will have to wait for v0.15.6.

[AkshatRai07]

There was one more issue, if you see the artifacts of maturin.yml, the version of eko-rs is 0.0.0. I checked locally and changing the version to be dynamic worked.

[scarlehoff]

Rust, C-API, and Python API tests be run on all OS and Architecture combinations which we will use when publishing?

A lot of researchers do use macOS locally and all clusters are linux so the minimal set would be mac-arm and linux-x64

[scarlehoff]

This lgtm, but one question, was dynamic = ["version"] a necessary change for this to work?

(can always be rolled back I guess)

[AkshatRai07]

I tried changing bump-versions.py to make it change the version in pyproject.toml too, but that didn't work. Setting dynamic = ["version"] makes it use the version of the Cargo.toml which is already bumped by bump-versions.py. Also, it is just a one line change and it works so I guess it is the best way here.

[felixhekhorn]

Suggestion: Please try changing the permission of PYPI_TOKEN, and run the entire Maturin workflow, only then merge this PR. Else we will have to wait for v0.15.6.

@scarlehoff eko (and so the token I think) is owned by the "N3PDF" user on PyPI - to which I have no longer access, since it needs 2FA now (I have only the password). I hope you still have access? 🙈 if so, can you please take care of this?

[felixhekhorn]

Concerning the location of the CAPI files: is this triggered by skipping "Publish release" https://github.com/NNPDF/eko/actions/runs/29902592251/job/89195933659 and will get fixed once we have all artifacts (which were failing before)?

[felixhekhorn]

Since we are fixing #537 here: can I please add one more unrelated thing? After having said that, I'm comparing https://nnpdf.github.io/pineappl/docs/installation.html#building-the-capi-from-source to https://docs.rs/ekore_capi/0.15.5/ekore_capi/#building--consuming . I think in PineAPPL pointing out how to install Rust or how to install cargo-c or to check out the repo is too verbose (and to be expected from the user), but I like the ${prefix} handling there. Instead in EKO pointing out how to compile a C program is too verbose (and to be expected from the user).

I suggest to do the following:

• tell the user they need cargo-c (is the comment there about staticlib still relevant after 2c4160b? )
• give the cargo command with ${prefix}
• tell that files x,y,z have been created
• remind the user they may need to adjust their environment vars
• tell the user they now can just use pkg-config in the usual way

[felixhekhorn]

one more glitch from #537: the links here

eko/crates/ekore_capi/src/lib.rs

Line 1 in d41da2f

//! C-language interface for [`ekore`], the crate providing the anomalous dimensions and

and here

eko/crates/ekore_capi/src/lib.rs

Line 7 in d41da2f

//! computed quantities, and the [ekore docs](https://docs.rs/ekore/latest/ekore) for the

are effectively the same, but different in practice - I suggest to use always the former. The second version is also used below

eko/crates/ekore_capi/src/lib.rs

Line 62 in d41da2f

//! [ekore docs](https://docs.rs/ekore/latest/ekore).

[felixhekhorn]

Add one more glitch from #550 (sorry for spamming this with unrelated stuff 🙈 ):

eko/crates/doc-header.html

Line 63 in d41da2f

for (let el of shorts) replaceAbbrev(el);

needs an additional

let tables = document.getElementsByClassName("item-table");
for (let el of tables) replaceAbbrev(el);

to make the abbreviations also work, e.g., also on the start page function list (etc.)