Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (6814180) 94.38% compared to head (9c114bd) 94.25%.
Report is 4 commits behind head on master.

@@            Coverage Diff             @@
##           master     #874      +/-   ##
==========================================
- Coverage   94.38%   94.25%   -0.14%     
==========================================
  Files          48       48              
  Lines        6665     6666       +1     
==========================================
- Hits         6291     6283       -8     
- Misses        374      383       +9     

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

What's the advantage of actively bumping dev-dependencies? Why shouldn't we specify the minimum version we actually need, and let the cargo resolver do its job?

Just read our "Cargo.toml" and saw either is "1.0" when the latest version is "1.10.0". Maybe we should update it to "1" or "1.10" as dependency trees might have it in multiple versions.
About dev-dependencies, maybe it's not that important though?
EDIT: My bad, I forgot this interesting page.

Maybe we should update it to "1" or "1.10" as dependency trees might have it in multiple versions.

either = "1" is a shorthand for either >= "1" (well, technically, either = "^1.0.0"). Cargo will only resolve a single version per major-version train, so there's no risk here of applications getting compiled with both "1" and "1.10".

See tokio-rs/tokio#6335 (comment), which sets a similar policy for tokio-rs. Unless we have good reason to change our current policy, I'd like to stick with it.

There is nothing urgent about updating dev-dependencies.
And while I think dependabot could in general be helpful about semver-major updates for our dependencies, we don't expect either (our only dependency) to ever go to "2.0".
And either = "1.0" actually means either = ">=1.0.0, <2.0.0" (cf link) which I forgot.

Your help on improving CI is very much appreciated, thanks!
However, it seems that this (and #875) should be closed. I'd promptly reopen if needed.