solidjs / solid / 40701965435

78%

main: 89%

push

github

fix(signals): refresh(memo) inside an action with an isPending reader settles (#4349) (#4356)

* fix(signals): refresh(memo) inside an action with an isPending reader settles (#4349)

An action's transaction lands beneath its own refresh(): nothing on screen
observed the re-ask's flight yet, so the landing commits the node in flight
(CONFIG_INPUTS_PUBLISHED). A verdict reader then probes the flight before
anything joined (a seam watcher) and observes it with its plain read: it
holds a frame of its own, as that frame's pending observer or its staged
derivation (the JSX ternary's condition memo). The lane seam re-derived the
watcher on the committed world every round; the re-run re-entered the same
transaction and parked it again until the dev guard threw "Potential
Infinite Loop Detected".

A reader a non-lane transaction holds that read none of its stagings
(neither REACTIVE_STAGED_READ nor REACTIVE_PROBE_UNANSWERED) has nothing to
re-derive: the seam leaves it to run at the landing. The same loop hit
`[isPending(x), x()]` in one render effect after an action's write.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015hPxMkC3jg3Uivh65gvaf2

* fix(signals): an action's affects() mark lasts until its node's own refetch lands (#4349)

A declared reload whose action ended before the re-ask landed released its
mark at the transaction's landing, in the action's own flush: an inline
`isPending(x) ? … : x()` render effect never showed `pending` while the
memo form (JSX's condition memo) did. A landing's mark on a node whose own
flight is up now stays to the seam after that flight lands, as an ambient
mark already did, so both forms show `pending`, then the landed value.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015hPxMkC3jg3Uivh65gvaf2

* perf(signals): laneSeams reads the lane through `l?._lane` (#4349)

The seam's three `l != null && l._lane... (continued)

790 of 1078 branches covered (73.28%)

Branch coverage included in aggregate %.

1127 of 1357 relevant lines covered (83.05%)

28.52 hits per line