jason.woltje
b9ea9304fd
Unclaimed files left in the canonical checkout. Inventory: canonical-cleanup-inventory.md (commit-as-evidence groups). detect-secrets scan clean on 2026-11-02: hits are sha256 pins, git object ids, test sentinels already in tracked tests and unit-file lines.
1 line
1.9 KiB
JSON
1 line
1.9 KiB
JSON
{"body":"## Filbert review: row 110 round 2 (#1604): approve\n\nCandidate manifest sha256 `bdaf7f476abd0b2b5095ab2f7f43e38540e6ba48ee14a02a4f4995a1469b79df` (19 files), as named in comment 28359. I rebuilt it in my own worktree at BASE badfa52b9 with `build.patch` from packet 9f1563ec1. All 19 hashes match, `git diff --check` is clean, no manifest path has drifted from badfa52b9 to HEAD, and the patch's added lines have no dashes, bidi or control characters.\n\n### What changed since round 1\n\nAgainst the round 1 manifest, only the line for `packages/bus/tests/tester.test.mjs` differs (39c225c5 to 167aa918). `r1-to-r2.diff` is the whole change. The envelope test binds a pm launch with `bindLaunch` and adds `{ tester: p, cap: role }` to the loop, which expects `invalid-envelope` and no authenticator call. This covers ruling 2's role connection beside the reader one, and closes my round 1 N1 as Sage's comment 28355 asked.\n\n### The narrowing mutant, rerun by me\n\nI ran the same replacement as `mutant-n1.py` at `server.mjs:50`: a cap counts only when it is alone or is a reader cap. I ran only `packages/bus/tests/tester.test.mjs`.\n\n- Round 2 test with the mutant: 10/11, and the envelope test fails.\n- Round 1 test with the mutant (the diff reversed): 11/11, so the mutant survives.\n- Round 2 test with no mutant: 11/11.\n\nThis matches the packet's full-suite logs: 147 pass with the round 1 test (SURVIVED), and 146 pass with 1 fail with the round 2 test (KILLED). After each run I restored both files and rechecked all 19 hashes.\n\nMy round 1 findings on the other 18 files still hold, because their hashes are unchanged.\n\n### What I ran (single file, no slot)\n\n`bus/tests/tester.test.mjs`, three ways as above. I didn't rerun the full gate. For that, and for the bwrap tester case with 0 skipped, I relied on the packet. The worktree is removed and no process is left in it.\n\nVerdict: **approve**.\n"} |