René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"random seed: generated finds a greppable-stream counterexample on main too: stdout has 189 gap lines for 190 gaps (seed 333199054)","articleBody":"## Summary\n\nThe end-to-end gap oracle's random unit `random seed: generated` found a counterexample in CI: at checkpoint `cli`, the greppable stream law fails with \"the stdout has 189 gap lines for 190 gaps\". The base fails the same way, so this is not caused by PR #90. Filed as an issue under P4, not rerun away.\n\n- CI: run 43688703700 (head 082991d), job `Gap report end-to-end oracle (random seed, 3/3)`. The seed was 333199054, path `1:1:1:1`.\n- Replays, each run twice with the same seed and path: on `origin/main` (5726197), 1 gap line short both times; on PR head 7df4a50, the same.\n\n## Replay\n\n    cd packages/oracles && ORACLE_SEED=333199054 ORACLE_PATH=1:1:1:1 ORACLE_PROPERTY=oracle-gaps.report ORACLE_UNIT='random seed: generated' pnpm exec vitest --run --config oracle-gaps/vitest.config.mjs gap-report-oracle -t \"^oracle-gaps\\.report(?: | > )random seed: generated$\"\n\nThe counterexample history is printed in the failure message (`History: {...}`).\n\n## Two defects in the report\n\n1. **The printed Replay line does not reproduce this unit.** The framework prints `ORACLE_SEED=... ORACLE_PATH=... ORACLE_PROPERTY=... ORACLE_UNIT=... vitest run -t '...'`. In this package, a bare `vitest run` uses the default config and finds no file: all 173 tests are skipped. The command needs `--config oracle-gaps/vitest.config.mjs gap-report-oracle`. The framework should print the project's test command, or this package's replay should say so.\n2. **The stream lost one gap line.** The stdout has one fewer gap line than the gaps reported. The cause is not known. Related but not the same: #87 (Windows, one extra line), and #76 (an intermittent stop with results missing beside the stream log).\n\n## Cost\n\nThe unit spent 817 s in this failing run, mostly in shrinking, and the job took 14 minutes. That is over the 10-minute budget (P12). A counterexample shrinks through up to six histories, and each one runs the real CLI.\n\n## Authority\n\nP4 (register): a counterexample is filed as an issue with its replay line, and a rerun never hides it. Under the M phase-1 decision, a random counterexample that the base also fails is filed, not blocking. The determinism-as-main reading in PR #90 applies the same rule to determinism units. The greppable-stream law (maintainer request 2026-11-12) is the one that fails.\n","author":{"url":"https://github.com/KyleAMathews","@type":"Person","name":"KyleAMathews"},"datePublished":"2026-11-15T07:40:05Z","url":"https://github.com/TanStack/oracles/issues/102"}