Found while implementing #15414 (PR #15506). Prose only, in a third file that card deliberately did not touch.
The stale claim
scripts/pm/dispatch-gates.mjs cites check-self-test-workflow-commands.mjs twice as the LIVE specimen for the negative direction of repoRootWalkSpelling — a walk seeded at a bounded subtree, therefore not a repo-root walk:
After PR #15506 that gate holds no walk at all — walkScripts, readdirSync and statSync are gone from it, and an own-source self-test case pins that they stay gone. Its population is imported from check-self-test-wired.mjs.
What is and is not broken
The outcome both sentences report is unchanged: the gate is still selected by none of the limbs, and the census is still five. The reason given is now false — it is not selected because it holds no walk, not because its walk is subtree-seeded.
Nothing reds. The self-test case beside the second comment runs on a hand-written fixture string:
repoRootWalkSpelling("const scriptsDir = join(ROOT, 'scripts');\nconst files = walkScripts(scriptsDir);") === nullwhich is data, not a read of the real file. So the fixture keeps passing and the prose keeps claiming a specimen that no longer exists.
Why this matters more than ordinary comment rot
The claim is explicitly framed as a measurement over live files ("the reading of its source, not a coincidence"). The next author who goes looking for the negative-direction specimen will open a file that has no walk in it and has to reconstruct why the sentence was ever true. The repair is either to re-point both sentences at a gate that really is subtree-seeded today, or to say plainly that the fixture is the specimen and no live file backs it.
Related: #15414, PR #15506.
Generated by Claude Code
Found while implementing #15414 (PR #15506). Prose only, in a third file that card deliberately did not touch.
The stale claim
scripts/pm/dispatch-gates.mjscitescheck-self-test-workflow-commands.mjstwice as the LIVE specimen for the negative direction ofrepoRootWalkSpelling— a walk seeded at a bounded subtree, therefore not a repo-root walk:REPO_ROOT_WALK_SPELLINGSheader: "check-self-test-workflow-commands.mjs— whose walk is seeded atscripts/, a bounded subtree — is selected by none of them, which is the reading of its source, not a coincidence."After PR #15506 that gate holds no walk at all —
walkScripts,readdirSyncandstatSyncare gone from it, and an own-source self-test case pins that they stay gone. Its population is imported fromcheck-self-test-wired.mjs.What is and is not broken
The outcome both sentences report is unchanged: the gate is still selected by none of the limbs, and the census is still five. The reason given is now false — it is not selected because it holds no walk, not because its walk is subtree-seeded.
Nothing reds. The self-test case beside the second comment runs on a hand-written fixture string:
which is data, not a read of the real file. So the fixture keeps passing and the prose keeps claiming a specimen that no longer exists.
Why this matters more than ordinary comment rot
The claim is explicitly framed as a measurement over live files ("the reading of its source, not a coincidence"). The next author who goes looking for the negative-direction specimen will open a file that has no walk in it and has to reconstruct why the sentence was ever true. The repair is either to re-point both sentences at a gate that really is subtree-seeded today, or to say plainly that the fixture is the specimen and no live file backs it.
Related: #15414, PR #15506.
Generated by Claude Code