solidjs / solid / 41669134881

79%

main: 89%

push

github

fix(signals): a memo reading through a linked re-based lane no longer loops at the seam (#4643) (#4658)

* fix(signals): a memo reading through a linked re-based lane no longer loops at the seam (#4643)

A memo seated in one lane whose pass read through a lane linked to it,
re-based over a parked frame's answer (`over`, #4525), read that frame's
staging there (#4536). The seam's `stagedReaders` loop re-derived it on
the committed world: the re-run re-entered the linked lane, read the
staging again and parked again, every round (Potential Infinite Loop
Detected). The #4536 carve-out in `laneSeams` checked only the memo's own
lane; it now checks the lane's link group (one reveal unit).

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

* fix(signals): lane work re-based over a hold joins nothing; a push after a removal guess composes on it (#4643)

Lane work re-based over a hold's answer (`over`) read that staging and,
as a computed, joined the flush into the hold (`joinPass`). D-37: lane
work never makes a flush the hold's. Joined, the seam re-derived it into
a loop; a plain write beside it (the inbox filter) was held while the
lane showed it and retracted at the next seam (F-7, Hold); a guess
beside the held answer revealed it through a linked lane, torn against
the count (F-1). It now joins nothing, as the lane's render effects.

The writer's view compared a length guess with the source instead of
the clone, so a removal guess's `undefined` slot extended the draft and
a later push wrote after a hole (#4642's hole; the same line as #4656).

laneSeams' link-group carve-out compares with `==` (both sides are
Transaction | null), for size.

Pins every frame of the inbox shapes (the issue's, with a declined
delete), a toggle beside a held poll answer, and pushes after removals.

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

807 of 1108 branches covered (72.83%)

Branch coverage included in aggregate %.

1132 of 1370 relevant lines covered (82.63%)

30.13 hits per line