solidjs / solid / 41719459783

78%

main: 89%

push

github

fix(signals): a guess into a second optimistic log shows beside another action's refresh (#4666) (#4669)

* fix(signals): a guess into a second optimistic log shows beside another action's refresh (#4666)

laneDirty seats a held pending reader in the lane before its pass, so
lane work reads the hold's flight as the screen (D-26, D-37) in either
read order (D-28). It only did so when the reader's hold was the lane's
own parent, which is true when the second guess goes into the same slot
(D-33 entangles the actions, #4637). A guess into a different
createOptimistic opens a lane under its own action, so the overlay held
by the first action's refresh(server) was not seated: its pass ran as
that hold's work, threw on the pending list and stayed hidden until
every refresh landed.

The seat now also takes a reader whose every pending flight is a
refresh() re-ask (_reask; with none, its own flight). A re-ask asks no
new question and holds nothing new (D-34 para 3), so the guess shows at
once (F-11). A reader pending on another hold's new question stays with
that hold (#4577).

laneDirty now runs insertSubs itself (every call site did so first), and
drops the l._parent check (l is always a lane, which always has a
parent). That pays for the new arm: 0 B on the core floors.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0117BoWNSSgtemsSYsaX2hXE

* fix(signals): a guess into a log whose lane has shown is not held by another action's refresh (#4666)

Residual shape the #4666 fuzz found: rows 1-4; close(4), assign(3), land
assign:3, then close(3), and #3 stayed shown until every refresh landed.

close(3)'s guess goes into close(4)'s lane, which has already shown. The
overlay, held by assign's refresh, is seated there by laneDirty, but the
lane is judged in the same seam with the overlay's stale pending, so it
holds once. The overlay's lane pass then writes the lane's staging for
its next reveal (`_pendingVal... (continued)

807 of 1108 branches covered (72.83%)

Branch coverage included in aggregate %.

1132 of 1370 relevant lines covered (82.63%)

30.18 hits per line