You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[EPIC]: Fix incorrect ERESOLVE errors from peer dependencies #8647
I've been affected by a number of incorrect ERESOLVE errors related to peer dependencies, forcing me to resort to --force or --legacy-peer-deps in order to get npm install to run. Peer dependency resolution seems to be one of the buggiest parts of npm, even after @owlstronaut's excellent work on fixing peer flag propagation (#8579).
To try to address this, I've gathered several individual ERESOLVE bug reports and used Claude Sonnet 4.5 to replicate each one and create a patch:
Unfortunately, there's no overlap between these patches, apart from all of them applying to build-ideal-tree.js. That suggests to me that there's a more fundamental problem with how peer dependencies fit in to npm's ideal tree algorithm. (Some of the patches may be strict improvements on their own merits, but I'm not confident in that, so I've only submitted them as draft PRs.)
I think this would be a valuable problem for an intrepid developer to dig into. I'll continue to update this ticket with any information I discover.
I think this would be a valuable problem for an intrepid developer to dig into.
These are not very helpful as PRs. Typically when a PR is submitted the work has been done to make it ready to land. If this was something we had the time to dig into we would be doing it ourselves.
We are happy to work through specifics in a PR that a community member has already worked through and needs some last mile help. A lot of the recent PRs for dep flag calculations are a good example of this. #8645
I'm going to close those PRs for now. If tests end up passing and you've done the work to make them acceptable then we can go from there.
Activity
wraithgar commentedon Oct 6, 2025
These are not very helpful as PRs. Typically when a PR is submitted the work has been done to make it ready to land. If this was something we had the time to dig into we would be doing it ourselves.
We are happy to work through specifics in a PR that a community member has already worked through and needs some last mile help. A lot of the recent PRs for dep flag calculations are a good example of this. #8645
I'm going to close those PRs for now. If tests end up passing and you've done the work to make them acceptable then we can go from there.