Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.49%. Comparing base (be2444e) to head (b7eb0c2).
Report is 5 commits behind head on main.

@@           Coverage Diff           @@
##             main     #530   +/-   ##
=======================================
  Coverage   91.49%   91.49%           
=======================================
  Files         178      177    -1     
  Lines       23623    23621    -2     
=======================================
- Hits        21613    21612    -1     
+ Misses       2010     2009    -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Setting build.rustdocflags in the workspace root's Cargo configuration file breaks the book doctests workflow, because the path to katex-header.html doesn't resolve from media/book/tests. As a workaround, I added an override in the workflow step, such that the RUSTDOCFLAGS are empty when running the book doctests.

There's a similar issue with a similar workaround for the argmin-math tests (and anything that is not invoked with the workspace root as working directory).

Open point: is it preferable to set RUSTDOCFLAGS='--html-in-header ...' only during the actual docs build (instead of the global setting in .cargo/config.toml)?

Firstly, this is an absolutely magnificent PR, both in terms of work done and description! Thank you for the massive effort you have taken to make this work!

Do I understand it correctly that the main problem is that both docs.rs and local builds have to be configured/set up separately? If that is the case, I think I can live with it.

Open point: is it preferable to set RUSTDOCFLAGS='--html-in-header ...' only during the actual docs build (instead of the global setting in .cargo/config.toml)?

Would this fix the discrepancy between docs.rs and local builds, or would this avoid the work arounds for argmin-math and media/book/tests (and instead requires setting RUSTDOCSFLAGS for argmin-testfunctions)?

The failing clippy lints should be fixed in main.

Thanks for the feedback!

Do I understand it correctly that the main problem is that both docs.rs and local builds have to be configured/set up separately?

Yes, that's correct.

Would this fix the discrepancy between docs.rs and local builds, or would this avoid the work arounds for argmin-math and media/book/tests (and instead requires setting RUSTDOCSFLAGS for argmin-testfunctions)?

It would remove the workarounds for media/book/tests and argmin-math. Setting RUSTDOCFLAGS for the CI step would work and remove the new .cargo/config.toml and the mentioned workarounds:

diff --git a/.github/workflows/docs.yml b/.github/workflows/docs.yml
index bbd3a70c..affe0c1f 100644
--- a/.github/workflows/docs.yml
+++ b/.github/workflows/docs.yml
@@ -19,6 +19,8 @@ jobs:
           components: rustfmt,rust-src

       - name: Build documentation
+        env:
+          RUSTDOCFLAGS: "--html-in-header ./crates/argmin-testfunctions/katex-header.html"
         run: cargo doc --features "argmin-math/latest_all,argmin/full" --no-deps --workspace --exclude "example-*"

       - name: Forwarding

Caveats:

• building locally during development is more cumbersome (one also needs to set RUSTDOCFLAGS in the local environment)
• the HTML header is again applied to all the crates, not just to argmin-testfunctions. I'm not aware of a way to achieve the latter while keeping the single doc build command issued from the workspace root.

The failing clippy lints should be fixed in main.

Thanks! I rebased the branch on main.