René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"oracle-gaps: mutant copies resolve bare imports from --package, not from the original file (288 setup-errors)","articleBody":"Reported from an adopter's first session, at c58b0c2. Not yet reproduced here.\n\n**Problem.** 288 of 388 mutants were `setup-error`: \"Cannot find package 'fast-check' imported from tests/realtime-oracles/src/.oracle-gaps-5482-10-ordered-live-results.oracle.test.ts\". The baseline loads the same import from `tests/realtime-oracles/node_modules/fast-check`. The comment at `check.mjs:1007` says copies are placed beside the file \"so that its imports resolve as the file's do\". But the copy is virtual: the wrapper plugin serves it through `resolveId` and `load`. So its bare imports seem to resolve from `--package`, the repo root, which has no fast-check.\n\nTwo more problems follow from that:\n- setup errors count toward the headline gap count, which came out as 307 gaps;\n- they get the directive \"Run an owner test that loads this module\", which doesn't apply.\n\n**Proposal.**\n- Resolve a copy's imports from the original file's directory. Alternatively, refuse at baseline when the oracle file's imports don't resolve from `--package`.\n- Either way, keep setup errors out of the gap count. A harness failure stops the run and is never charged to a mutant (decision LA-v20-1, cleanup plan Phase 2 A2). Give them their own directive.\n\nThis issue was written by Isaac.\n","author":{"url":"https://github.com/KyleAMathews","@type":"Person","name":"KyleAMathews"},"datePublished":"2026-11-13T21:25:20Z","url":"https://github.com/TanStack/oracles/issues/68"}