1. Most time: waiting on the two background runs. nikaia test --both-settings --project tests/compiler and the it filter set with clippy ran side by side, several minutes in all. The fix was a few lines of code; finding and building it took about ten turns.
2. Looked up: lambda_names, to check that it hands back resolved names comparable with run_code. Also whether the callee of f(x) is visited as an Expr::Variable, which I assumed and the probe confirmed. The brief’s Locations covered everything else. scripts/where.nika -- kept_hulls found nothing (a field is not found, as builder-facts says), so the brief’s pointer to it for “the hull tests” led nowhere. The hull tests turned out to sit in tests/compiler, which the full run covered anyway.
3. Checks:
   • Useful: the probe on the old binary (both rows failed, as the brief said); one build cycle; the probe on the new binary; the tests/compiler run (required by the check_ change); the it modules whose tests mention Kept/hull, plus the guard set; clippy; pre-push-check.
   • Wasted: the where.nika call. Possibly also lock_floor:: and lock_doors::: I picked them by a broad grep for hull, and the change only touches lambdas that capture an fn local.
   • Missed: the commit was rejected, so the push did not land. I did not rebase, as the rules say.
4. One change: when the cause is traced to a single predicate, the brief could name the exact edit point and the test set in one line: “edit a_name_of_a_hull; touched tests are tests/compiler plus it function_types:: and shared::”. It could also drop the where.nika -- kept_hulls hint, which lists nothing for a field.

This site is open source. Improve this page.