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
I searched open requests and couldn't find a duplicate
What is the idea?
conda-build should allow users to provide a conda-lock lockfile so that it can create environments in a reproducible fashion when building (and optionally testing) a package.
Why is this needed?
Because conda-build does not currently support lockfiles, running the same conda-build command on the same revision of a codebase can succeed the first time you run it and fail the second time you run it, due to out-of-band changes (e.g. a transitive dependency publishing a new version). This makes conda-build users' CI jobs unreliable, and makes it harder to find out why a build has broken. If conda-build consumed lockfiles that were versioned in the codebase, it would enable the invariant that every CI job's pass/fail result is a pure function of the revision of the code that it ran against, and so if revision X passed and revision Y failed, the cause of the failure can be found somewhere in the diff between revision X and revision Y.
What should happen?
conda-build should allow users to provide lockfiles which it should use for every environment that it creates as part of the build, bypassing solver-based non-determinism.
Additional Context
No response
The text was updated successfully, but these errors were encountered:
Checklist
What is the idea?
conda-build
should allow users to provide a conda-lock lockfile so that it can create environments in a reproducible fashion when building (and optionally testing) a package.Why is this needed?
Because conda-build does not currently support lockfiles, running the same conda-build command on the same revision of a codebase can succeed the first time you run it and fail the second time you run it, due to out-of-band changes (e.g. a transitive dependency publishing a new version). This makes conda-build users' CI jobs unreliable, and makes it harder to find out why a build has broken. If conda-build consumed lockfiles that were versioned in the codebase, it would enable the invariant that every CI job's pass/fail result is a pure function of the revision of the code that it ran against, and so if revision X passed and revision Y failed, the cause of the failure can be found somewhere in the diff between revision X and revision Y.
What should happen?
conda-build should allow users to provide lockfiles which it should use for every environment that it creates as part of the build, bypassing solver-based non-determinism.
Additional Context
No response
The text was updated successfully, but these errors were encountered: