chore: T01.g est close, sa trouvaille bloquante est devenue T01.o

2026-09-24 19:54:42 +02:00
parent e07453c205
commit 7a37e44ac0
2 changed files with 1 additions and 0 deletions
@@ -67,3 +67,4 @@ The polyfill's published surface has drifted from the SDK's: invented names, inv
 - T01.l - `ng-e2e-helpers`' peer range tightened to `>=0.1.2-alpha.13 <0.1.2`, MEASURED BY RESOLUTION: before, 4 versions admitted with `maxSatisfying` = the bare `0.1.2` of 2025-10-19; after, 3, and pnpm resolves `0.1.2-alpha.15`. The premise it was handed was half wrong and it said so: pnpm never actually RESOLVED to `0.1.2` because it prefers the `latest` dist-tag, which upstream happens to point at alpha.15 - so the trap was one dist-tag move from firing, and only resolution testing shows that either way. `playwright` 1.61.1 -> 1.63.0 (raises `engines.node` to `>=20`, drops `fsevents`), e2e `46 passed / 46` on the first run. Both `typecheck` scripts moved from `bunx tsc` to `pnpm exec tsc`, VERIFIED with `~/.bun/bin` off PATH and `command -v bun` empty. MEASURED about the optional peer: a consumer of the packed polyfill receives NO `@ng-org` package under pnpm, so the old `bun install` semantics survived the tool change - but THIS checkout does get the ORM, through `packages/polyfill/package.json:33`'s devDependency, so the runner's "install it in packages/polyfill" was a false lead and both texts now say the optional peer governs what a CONSUMER receives. `pnpm link` PERFORMED and it works: the polyfill resolves to the checkout and stays at ONE copy, so the single-instance hazard is untripped - but `@ng-org/web` and `@ng-org/orm` end up at TWO copies when the app also imports the SDK, pnpm warning by name, and `file:` gives one copy at the cost of repacking, which forfeits live editing. Left alone, reported: the polyfill's own four `"*"` peers admit ONLY the bare `0.1.2` of 16 published versions and EXCLUDE every alpha, because `*` matches no pre-release - inert while the polyfill is unpublished, certain to bite at first publish (now T01.n); `examples/notebook` has no `typecheck` script; `bun build` at `orm-live-read.ts:128` and `bun run` in `mint-wallet` are still bun-bound.
 - T01.m - THE COUNT WAS ELEVEN, not the three the split had named. Deleted from `open-repo.ts`: `held`, `syncState`, `inFlight`, `settleInFlight`, `registryEpoch` - the last two IMPLIED openness through a short-circuit and nobody had ever counted them, including six adversarial rounds of T01.j. Surviving as the one answer, in `surface/subscribe.ts` (NOT `emulated-verifier/`, which is what the task file wrongly said): `fanOuts`, `establishing`, `realUnsub`, `lastState`, `attempt` - and `lastState` turned out to have been a second barrier record all along, the correct one. THE INVARIANT, written at `open-repo.ts:100-128` and on `DocChannel`: "is this document open in this session" is answered in ONE place, `docChannel(nuri)`, and `open-repo` records only the bootstrap open it ASKED FOR. Teardown is now safe BY CONSTRUCTION with no cross-module notification added, because nothing outside records openness any more; `resetOpenedRepos` disowns rather than forgets, since forgetting yields the lax `"unknown"`. CONTRACT: both non-guarantees DELETED and the arrival guarantee unqualified, `1.0.0-dev.10`, guarantee and test in the same commit - and an INDEPENDENT review attacked that first and hardest and could not break it. `462 pass / 0 fail` six times, e2e `46 passed / 46`. My own spec was wrong twice and the agent said so instead of contorting to satisfy it: the module path, and failure B's acceptance test ("2 subscribes in-window"), which is UNACHIEVABLE because the fan-out correctly dedupes. Four adversarial passes found SIX real defects in the agent's own work, including two regressions and a test of its own that asserted the defect. THE LESSON WORTH KEEPING, from the last one: it had the right fix, could not make a test discriminate, and REVERTED it - correctly - but recorded "could not construct an ordering" as an OPEN QUESTION rather than a clean bill. That wording is what let a later pass finish it. The window is exactly ONE microtask hop, swept 0-6: at 0 the clear has not happened so delivery is legitimate, at 1 a forgotten barrier is delivered, at 2+ `lastState` is already null so nothing is queued. Its earlier fake put the rejection three hops behind the replay, so the replay always won - what it had wrong was the TEST, not the remedy. Also verified by the independent review, and worth knowing because the guarantee is right for a reason nobody had written down: `"unknown"` is unreachable where a barrier was expected NOT for the stated reason, but because `fetchReadCap` -> `downloadReadCap` -> `ensurePhysicalRepoOpen` populates the open record BEFORE `mustNotAttempt` is evaluated, so even the refused path leaves a trace. Follow-ups left open: a never-settling `doc_subscribe` is unrecoverable; `public-store.attempted` skips an open after `configure`; `forgetEverything` clears four maps never.
 - T01.n - the four `@ng-org` peer ranges fixed, VERIFIED BY RESOLUTION against the registry with the real node-semver engine rather than by reading. WORSE THAN THE TASK STATED: `@ng-org/orm`'s `"*"` admitted ZERO of its 22 published versions - that library publishes no bare `0.1.2` at all, so the range was UNSATISFIABLE, not merely wrong. The other three admitted exactly 1, the stale bare `0.1.2` of 2025, and excluded every alpha. After: web `>=0.1.2-alpha.14 <0.1.2` (2 admitted, maxSatisfying alpha.15), orm `>=0.1.2-alpha.21 <0.1.2` (2), shex-orm `>=0.1.2-alpha.9 <0.1.2` (2), alien-deepsignals `>=0.1.2-alpha.13 <0.1.2` (2). The floor was taken PER PACKAGE from its own devDependencies pin rather than copying the sibling package's `alpha.13` that the task offered as an example - that floor predates its own devDep bump and is a stale artifact, not a rule. `peerDependenciesMeta: optional` confirmed correct for all four and VERIFIED rather than assumed: every `@ng-org/*` reference in `packages/polyfill/src` is `import type` only, zero runtime imports, the application injecting the real functions through `bootstrap.ts`. `examples/notebook` gained the `typecheck` script it lacked. Fresh install ends with ONE `.pnpm` store entry per package and every consuming project's realpath matching. `462 pass / 0 fail`. It also caught an error of MINE: the task file I wrote carried a baseline of 449 while the measurement says 462 - written before T01.m added its tests and never refreshed. That is the second of my own specs to carry a wrong number, after T01.m's unachievable acceptance test. Left alone, reported: `ng-e2e-helpers`' own floor looks stale for the same reason (INFERRED, out of scope).
 - T01.g - the closing review REFUTED the parent's success criterion, which is what it was for. 8 findings over 53 published symbols (not the 25 the orchestrator's naive grep counted - it measured export STATEMENTS), 41 clean. THE BLOCKING ONE IS A RIGHTS LEAK, and it is this batch's whole failure mode applied to access control: `read-filter.ts` returned non-function members RAW on the written ground that "a plain property carries no items" - true of the CITED tree (orm alpha.18, a bare `DeepSignalSet`), FALSE of the INSTALLED alpha.21, whose `useShape` returns a `UseShapeResult` carrying every identity's items on the plain properties `data`, `promise` and `subscription`. A stranger read the owner's data through all three and nothing threw. It became T01.o. Three majors: the contract instructed an injection that does not resolve (`@ng-org/orm`'s `.` maps to `core.d.ts`, which has no `useShape`); four present-tense claims that the SDK returns a bare `DeepSignalSet` are false of alpha.21, TWO of them load-bearing provenance grounds - the `readUnion` failure mode again, invisible to `provenance.test.ts` by construction; and the only live gate for the published `useShape` injects a HAND-WRITTEN substitute, so no fake in the repository returned the shipped shape, which is exactly why every gate was green. Four minors became T01.p. THE LESSON FOR THE ORCHESTRATOR, not the code: ten adversarial passes and every re-run gate of this batch were blind to this, because the only artifact exercising that path was a fake that had never been checked against what ships. A green suite is worth what its fakes are faithful to, and nobody had asked.