You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 544ad8a
Browse filesBrowse the repository at this point in the historyBrowse files
An uninitialized async memo whose re-run throws from a dependency now reaches the wrapping `<Errored>` (#4364). The dependency's error was staged for the commit (D-41), but a reader threw it at once: a render effect reading a memo born staged, or the memo itself through `latest()`. The error reached the boundary naming that source, and the boundary's next pass took the source as settled because its error was only staged. It dropped the source and showed the content, so `Loading` revealed an empty hole that never recovered. An error staged on its source now keeps the boundary waiting until the commit publishes it.
`isPending(s)` read through a memo (`<Show when={isPending(s)}>`, a compiled ternary's condition) no longer stays `true` after an action completes when `s` is also written outside the action (#4367). The outside write that matches the action's value ends the verdict, and the memo's pass leaves the verdict lane on a value equal to its committed one, so the equality gate notified nobody; the lane seam woke the reader that still showed the lane's `true` only when the lane's frame was not held, and the action held it. A pass outside any transaction's flush is the screen's: its readers now re-derive at the seam, as the inline `isPending(s)` does.
A `<Loading on>` boundary no longer shows its fallback when a held `on` change commits under an action's own `refresh()` (#4379). When an `on` change is held with an action (an input changed under its optimistic guess, #4336) and the content's new answer lands inside that hold a flush after the change, the boundary now folds the staged fallback back, as it already did when the content was ready at the change. The action's commit shows the content, and the quiet `refresh()` that follows keeps it on screen until it answers.
Copy file name to clipboardExpand all lines: packages/signals/docs/RULES-DERIVED.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -120,7 +120,7 @@ Whether a fallback is seen depends on whether the boundary is on screen. A fresh
120
120
121
121
A flight whose only observers are render effects behind a fallback holds no change — a memo under the boundary that no render effect reads was never an observer at all, and one rendered by an effect outside the boundary is that effect's read (F-3). In both orders: a render effect created under a fallback never counts, and a reader that counted while the boundary showed content stops counting the moment `on` sends it back to the fallback. Writes the hold kept for those readers alone commit at once (`Boundary: Loading... | Sum: 2` publishes as one frame); a reader outside the boundary still holds (`Sum` waits until `details` lands). The wait does not vanish; it moves onto the boundary: after the reset the fallback stays until every reader under it has settled — including readers whose flights started before the reset — then one coherent reveal.
122
122
123
-
**Collapses:**[A33](./SPEC-ASYNC-SEMANTICS.md#a33-a-fallback-caught-flight-holds-no-transaction-a-loading-reset-moves-the-hold-onto-the-boundary) (#3375, #3459), §23.1 item 1, recovery 2.2. **Status:** standing. The nested form — an outer `on`-reset boundary not waiting for a flight its nested `on`-reset boundary catches — is built (U-4, 2026-10-13: a reader a nested collecting boundary holds is that boundary's alone; `tests/fuzz-findings-l2.test.ts`, fuzzer finding F1). The re-arm over a hold is built (U-9, 2026-10-14): content an earlier hold staged under the boundary, or that reads a hold after the re-arm, is the boundary's — the fallback shows now and the content reveals at that hold's commit, which does not wait for it; no redraw loop (`tests/loading-fallback-in-flush-3540.test.ts`, the re-armed shapes; `tests/loading-on-frame-following-3540.test.ts` §3b, §3c).
123
+
**Collapses:** [A33](./SPEC-ASYNC-SEMANTICS.md#a33-a-fallback-caught-flight-holds-no-transaction-a-loading-reset-moves-the-hold-onto-the-boundary) (#3375, #3459), §23.1 item 1, recovery 2.2. **Status:** standing. The nested form — an outer `on`-reset boundary not waiting for a flight its nested `on`-reset boundary catches — is built (U-4, 2026-10-13: a reader a nested collecting boundary holds is that boundary's alone; `tests/fuzz-findings-l2.test.ts`, fuzzer finding F1). The re-arm over a hold is built (U-9, 2026-10-14): content an earlier hold staged under the boundary, or that reads a hold after the re-arm, is the boundary's — the fallback shows now and the content reveals at that hold's commit, which does not wait for it; no redraw loop (`tests/loading-fallback-in-flush-3540.test.ts`, the re-armed shapes; `tests/loading-on-frame-following-3540.test.ts` §3b, §3c). A fallback staged into the hold its readers land in folds back at any seam, not only the re-arm's own: readers still pending at the re-arm land into that hold a flush later, and the content is ready when it commits (#4379, `tests/loading-on-held-filter-refresh-4379.test.ts`).
124
124
125
125
### D-15. `on` is a dependency list; the swap lands with the frame of the change that caused it
0 commit comments