solidjs / solid / 43161618528

79%

main: 90%

push

github

fix: a re-guess of a shown guess reveals with its lane's passes (#4775) (#4779)

* fix: a re-guess of a shown guess reveals with its lane's passes (#4775)

A second guess in a lane that had shown wrote the screen's value at once,
before the passes it re-runs had been judged. When one of them went
pending, the lane held, but a reader re-run by an unrelated write read
the unrevealed guess beside readers still showing the first one.

The re-guess is restaged for the lane's next reveal, and the lane is not
judged at the seam that applied it (S-30 row 2', F-1, F-11). Flips the
F5b pin in fuzz-findings-l2.

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

* fix: keep a re-guess off the screen behind the guess layer (#4775)

The first cut put the restaged value in truth.ts latestOf, so the
until/refresh bundle (no optimism) grew 24 B brotli and went over its
cap. The screen's value of a shown guess that is re-asked is now kept by
the guess layer (guesses.ts), which lanes.ts reads through screenOf's
hook; the lane is marked with the seam's clock instead of a list. Truth
readers pay nothing, isPending-only bundles pay the hook calls.

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

* chore(signals): record _reguessed in mangle-props table

The build appends new private names to the table, so a stale table changed
turbo's input hash between the build and test steps and re-ran signals:build
while the example tests read its dist.

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

---------

Co-authored-by: Claude <noreply@anthropic.com>

818 of 1112 branches covered (73.56%)

Branch coverage included in aggregate %.

1143 of 1372 relevant lines covered (83.31%)

30.35 hits per line