Downstream testing for language pacakge managers, it's buildable but perhaps not fundable?

https://nesbitt.io/2026/03/01/downstream-testing.html

Downstream Testing

Most library maintainers have no way to test against their dependents before releasing.

Andrew Nesbitt
@andrewnez can I point cargo-semver-check? It is not *exactly* what you are talking about here, but I think it helps a lot with that problem and allows to push it back upstream with less stuff to run.
@Di4na if we had that in every other language, we’d be in a much better place
@andrewnez so funnily enough, iirc they mentionned they had a prototype for python and ruby iirc... but no funding to do all the work to productivize it

@andrewnez Excited to see https://dependabot.ecosyste.ms/. The Dependabot compatibility score used to be more accessible before GitHub took it over.

Any early indication of problems in a release is very freeing particularly as a solo maintainer when development time is limited.

Open Source Dependabot Pull Request Tracker | Ecosyste.ms: Dependabot

Track Dependabot pull requests across open source repositories. Discover security updates, dependency changes, and automation insights for package maintainers and security researchers.

@andrewnez there’s an interesting intersection with resolution policy here.

Because bundler/rubygems always try to resolve every dep to a single version, a feedback loop occurs where breakage upstream quickly attracts patches that mostly re-level the ecosystem.

But in systems allowing multiple versions to be resolved, that feedback loop is largely absent and so the spread of versions in use is more heavy tailed.

@andrewnez I also just remembered that in my time at Shopify I wanted to build a kind of progressive rollout system for new versions. Working from low-impact repos up to The Monolith. Integrated with the ShipIt deployment tool as well to get traffic control. Got laid off before I could pitch it properly 🤷
@jacques @andrewnez Oh, man, same. I worked in Resiliency and tried to get Dev Infra on board with this. I got cut before I could get far enough, though.
@andrewnez also a little relevant maybe is cargo-semver-checks, as a way of at least trying to make it easier for maintainers to catch semver-incompatible changes. Leverages (1) the ability to encode API invariants in the type system and (2) fairly strict semver praxis in the crates.io ecosystem.