Repository navigation
Kernel: a dependency update needs no model review #32
Description
Part of the epic "Epic: nine gaps in the Kernel, found in review".
Why. A pull request of class dependency with green checks waits for a model review, and one is admitted. A dependency update is judged by the deterministic Lenses alone; people can still block it.
Done when
-
test_a_dependency_update_needs_no_model_reviewpasses with@pendingremoved. - One function says which model Lenses are required for an item's class. It is used where the Kernel waits for reviews and where it reads gating threads, with a test and a mutant for each place.
- A person's thread still gates a dependency update.
- No document says
not held yetbeside this test.
Needs. The pull request that carries the pinned tests has merged. The simulator's world is always of class maintainer, so it will not exercise this rule; the tests must.
Design. §4.2, "Which Lenses are required is data".
Activity
don-petry commentedon Oct 15, 2026
Working ruling (Solution Architect, delegated). The Maintainer may overturn it in review.
Any open thread from a model Lens gates a dependency update. Whether a model Lens is required decides only whether the Kernel waits for that Lens's review. It does not decide whether a Finding that Lens has already raised still counts.
Otherwise an edge case converges with an unanswered Finding while the ruleset blocks the merge (the deadlock in L-08):
pr-reviewopens a thread.A pure dependency update is never reviewed, so it never has such threads and loses nothing.
Criterion 2 is reworded: "One function says which model Lenses an item's class requires, and is used where the Kernel waits for reviews. A Finding a model Lens has raised gates whatever the class. Each has a test and a mutant."