solidjs / solid / 43192579174

79%

main: 90%

push

github

fix(signals): plain reads of a held value agree around a first latest()/isPending() (#4777) (#4790)

* fix(signals): a first probing pass that read a held value plainly starts over as a verdict reader (#4777)

D-25 / A10: the two channels of one node agree within a pass in either
read order. A memo whose earlier passes never probed read id() (served the
hold's staging, the flush adopting that hold), then latest(id), then id()
again (served the screen): before=1, after=0, and an async body started its
request with those arguments.

When the window opens and the join is still a plain adoption (the flush had
no transaction before this pass), the flush gives the hold back and the pass
re-runs as the verdict reader it is: both plain reads are the screen's. An
asker (CONFIG_ASYNC, S-51) and a pass born held are left as they were.

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

* fix(signals): restart the pass through a quiet loading-window park, not a core catch (#4777)

The restart no longer needs its own arm in recompute's catch: the window
parks the pass as a loading window's unready read (no visible pending) and
latches REACTIVE_MISSED_WAKE, so the pass's own tail runs it again in this
flush. Apps without latest/isPending pay nothing: core floor and
until/refresh are back to base.

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

* fix(signals): a #4777 restart gives back every hold the pass joined

The restart dropped only the flush's adopted hold. A second hold the pass
merged in stayed entangled with the first and never settled on its own.
Merges made since the adoption are now recorded and undone in reverse.

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

* chore(size): raise isPending/latest and compiled live... (continued)

818 of 1112 branches covered (73.56%)

Branch coverage included in aggregate %.

1143 of 1372 relevant lines covered (83.31%)

30.18 hits per line