solidjs / solid / 40347492914

79%

main: 89%

push

github

perf(signals,solid): win back the bytes from the 2026-10-26 fixes, and more (#4295)

* perf(solid,signals): trim the bytes #4279 and #4281 added

laneRead (#4281): the D-28 verdict-lane exit shares the never-shown test
with the A15 arm it falls into, computed once inside one EFFECT_RENDER
block; `!guess && status & STATUS_PENDING` is hoisted to one local, and the
untracked arm falls through to the shared enterLane tail instead of
repeating it. Same evaluation order and side effects.

hydratedCreateErrorBoundary (#4279): the content's no-claim window runs
through hydrateWindow (no id, no scope, hydration already on, snapshot
root already marked: it only swaps the claim owner) while hydrating, and
plainly after (claims are read only while hydrating); the has/load checks
and the two plain-boundary returns merge into one fall-through.

Local, minified, this commit alone: page: live server components -73 B
(signals -35, solid -38); app: hydrating -38 B; signals: + isPending/latest -35 B.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018LY7F9czoRqdMEvk2QwPr1

* perf(signals): fold hook-only arms away in builds without the hook module

A call site guarded only by state the hook module sets (`fam.opt`, a
boundary's reveal slot or controller, a lane's marks, CONFIG_GUESS) kept
its dead arm in a build that never imports that module: the bundler folds
the never-assigned hook binding to `null` but cannot prove the guard
false, so `null.view(...)`-style calls shipped. Each site now checks the
hook first (truthy, as async.ts already does), so the arm folds:

- store optimistic arms (optHooks: store.ts, reconcile.ts, affects.ts)
- reveal-order arms in boundaries (revealHooks), the boundary flip's
  laneStageHook call
- recompute's laneOutcomeHook arm, land()'s releaseMarksHook call

The hooks are installed once at module evaluation and never cleared, and
the guarded state is only ever set by thos... (continued)

777 of 1070 branches covered (72.62%)

Branch coverage included in aggregate %.

5 of 5 new or added lines in 1 file covered. (100.0%)

1114 of 1358 relevant lines covered (82.03%)

28.02 hits per line