fix(scripts): correct bash-3.2 comment's stated mechanism in bump-objectui.sh

[yinlianghui]

Fixes#12222

What

scripts/bump-objectui.sh's bash-3.2 comment explaining why the ref-collection loop uses
if [[ -n "$x" ]]; then …; fi rather than a trailing [[ … ]] && … attributed the trap to
the empty-list case. Measured on bash 5.2.21, that does not reproduce: an all-empty read
leaves the loop body unexecuted and the while exits 0. The real trap is a non-empty
read whose last line fails the test, and only once that loop is the last command of a
function — the &&-list's false status becomes the loop's exit status, becomes the
function's return, and set -e kills the caller on that.

The if form itself is unchanged — this PR touches only the comment's stated reason, not
behavior.

Scope

Declared file surface: the one comment in scripts/bump-objectui.sh (lines 341-343 → the
corrected block). Nothing else changed:

• The loop's local x='' note is untouched (a separate, correct set -u point).
• No code changed — the if form stays exactly as written.
• PR fix(scripts): bump-objectui.sh runs on bash 3.2 — the pin-reachability warning is macOS-reachable again #12142's body (where the wrong claim was first stated) is immutable history and out
  of scope per the card.

Verification — the measured mechanism

Constructed all four combinations under bash --version 5.2.21 (set -euo pipefail, loop
body [[ -n "$x" ]] && arr+=(...)):

case	input	loop position	exit status
1	empty	last command of function	0
2	empty	NOT last command of function	0
3	non-empty, last line fails test	last command of function	1 (killed)
4	non-empty, last line fails test	NOT last command of function	0

Only case 3 dies — confirming the corrected comment's claim and refuting the original
empty-list attribution (cases 1 and 2 both exit 0 regardless of loop position).

Gates run (final commit a4dc82450d)

node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack at the final commit
names 8 matched local families for this diff (no STALE TREE warning); all run via the
shared verify lock, all exit 0:

• pnpm check:agent-test-spelling — 0
• pnpm check:cli-command-ids — 0 (✓ check-cli-command-ids: 276 command-id literal(s) … 1 baselined violation(s) — pre-existing baseline entry, unrelated to this diff)
• pnpm check:cross-package-test-inputs — 0 (All 117 self-test cases passed.)
• pnpm check:entry-guard — 0 (✓ check:entry-guard: 167 scripts/ file(s) …)
• pnpm check:parse-guard — 0
• pnpm check:pnpm-filter-targets — 0 (✓ check:pnpm-filter-targets: 136/173 …)
• node scripts/check-ci-filter-parity.mjs — 0 (OK: all 96 declared cross-package glob(s) …)
• node scripts/check-cross-package-test-inputs.mjs — 0

Also ran pnpm check:objectui-changeset (the R7 leg this file touches, per the claim
comment's flagged hazard) — exit 0, self-test line ✓ objectui-range --self-test: all checks passed, R7 line reads ✓ #12071 R7 bump-objectui.sh names no bash 4+/5 construct (mapfile, ${x^^}, declare -A, &>>, EPOCH*). The correction is on full-line comments only
(no trailing comment on a code line), so R7's ^\s*# exemption applies cleanly.

No changeset — this PR publishes nothing (a shell comment correction) — skip-changeset
label applied via the read→union→add flow (REST surface closed from this dev seat, #12123).

Generated by Claude Code

---

Generated by Claude Code