Repository navigation
2.0: a second write during a parked latest() hold asks the frame's question again #4400
Closed as not planned
Description
What happens
const m0 = createMemo(() => ask("m0", 1 + latest(s) + s()));With #4393 (R5), m0's first pass after setS(2) belongs to the frame and asks m0:5. When the flush parks, the seam re-derives m0 in the verdict lane (m0:4). At the landing, m0 adopts the frame's kept flight instead of asking again, so after one write the asks are m0:5, m0:4, m1:5, m1:6, each once.
A second write during the hold (s=3) still asks m0:5 twice, the same as before the adopt fix. The cause isn't traced yet. One likely path is that the second write's pass drops the kept flight (_x._parked), so the landing asks the frame's question again.
Repro: the verdict-memo-flight-no-reask shape in #4393, with a second setS(3) while the hold is up. Related: #4398.
Activity
ryansolid commentedon Oct 29, 2026
Not a bug. The repeated
m0:5is two different questions that happen to share a number under1 + latest(s) + s():With a distinct encoding the second write asks
m0:23 m0:22 m0:32 m1:33 m0:34 m1:35, and each question is asked once in both the pair and split shapes. #4393 pins this case inlatest-two-readers-one-ask-4393.test.ts. Closing.Generated by Claude Code
test(signals): the #4393/#4400 two-reader specs run on fake timers
fix(signals): don't retain the transaction a latest() read opened aft…