Repository navigation
Commit b61eef6
* signals(tests): pin what L2 already gives of #3761 — a mainline mount over a foreign hold
PR #3761 (GabbeV) proposes that a mainline mount's derivations read the
screen first and publish before a foreign hold, then re-derive under it.
L2 rules the opposite (A29 creation-time form — born held; the
mount-reads-the-screen arm was built and reverted 2026-10-02, plan §27.2
N1), so the PR's semantics are not ported; its tests that encode them are
left for the maintainer's ruling.
This commit keeps the five cases from the PR's test file that follow from
standing rules, retitled in L2 vocabulary: a born-held mount disposed
before the landing (ruling A / O2), a mount over a never-committed source
(A19 exc. 1), a direct render effect over two foreign holds (A15 stale
reader), and two join-then-land shapes. RULES-INDEX.md regenerated for the
new citations.
Co-authored-by: Claude via Cursor <[email protected]>
Co-authored-by: Cursor <[email protected]>
* signals(spec): membership in a hold is the tick's — passTx is tick-scoped (ruled 2026-10-04)
Maintainer ruling on the side observation from #3761's review: two mainline
mounts in ONE tick that read two different foreign holds merge those holds
(the tick is one frame — ruling 1); mounts in separate ticks stay
independent. A15's "disjoint graphs keep independent transitions" is about
writes in different ticks.
Spec: a dated clarification paragraph under the hold model, after the
2026-10-03 re-pins. Pin: tests/tick-scoped-pass-transaction.test.ts, both
arms (same tick → merge, both reveal at the second release, isPending true
on both until then; separate ticks → the first reveals at its own release).
RULES-INDEX.md regenerated. No engine change.
Co-authored-by: Claude via Cursor <[email protected]>
Co-authored-by: Cursor <[email protected]>
* signals(spec): the tick is the work one flush settles — clarify the tick-scoped hold ruling
Maintainer clarification on the 2026-10-04 ruling: "same tick" means until
the flush. Two mainline mounts created before the same flush are one frame
and their foreign holds merge; a manual flush() between them ends the frame,
the second mount is a new tick and escapes the first's transaction, and the
holds stay independent. The tick is defined by the flush, not by a
microtask, await or timer boundary.
Spec paragraph and the pin's header, titles and comments reworded in those
terms; the arms are named by their shape (`mountA(); mountB(); flush()` →
merged; `mountA(); flush(); mountB(); flush()` → independent). No
behavioural change to the pin; no engine change.
Co-authored-by: Claude via Cursor <[email protected]>
Co-authored-by: Cursor <[email protected]>
---------
Co-authored-by: Claude via Cursor <[email protected]>
Co-authored-by: Cursor <[email protected]>
1 parent a036c22 commit b61eef6
4 files changed
Lines changed: 373 additions & 48 deletions
File tree
- packages/signals
- docs
- tests
0 commit comments