test(codegen): restore the unary-pos assertions lost in the #8657 squash

[proggeramlug]

main is red; this makes it green. Three
expr::computed_store_rooting_tests cases fail 4/4 on main and 0/3
with this change.

What happened

#8650 changed unary + on a non-numeric operand to emit js_dynamic_pos
instead of js_number_coerce (expr/unary.rs, UnaryOp::Pos), and updated the
three tests that assert on that helper. When I squash-merged it as part of
#8657, the emission change survived and the test edit did not — so the
assertions kept naming a helper the compiler no longer emits at that site.

Bisected:

commit	failures
c2da03439	0/3
06e1ab349 (#8657)	4/4
b5359b4de (#8664)	3/3
2382a9f15 (current main)	4/4

Each of #8657's five PRs is 0/3 alone, and all five re-merged together are also
0/3 — only the landed squash fails, which is what isolates it to the merge
rather than to any PR.

My error, recorded

I saw these same three tests fail 3/12 while validating #8657 and judged them
"likely pre-existing codegen tier nondeterminism," landing with a note. That was
wrong twice over: the failure was real and caused by my merge, and the
"nondeterminism" framing sent me looking at load sensitivity instead of at the
diff. #8658 was filed on that mistaken premise and should be re-scoped — the
intermittency I measured then was masking a deterministic breakage, not
describing one.

The lesson is narrow and worth keeping: a squash that drops one hunk of a PR
while keeping another produces a tree that no reviewer approved, and no
individual-PR check can see it. The only thing that catches it is testing the
merged result — which I did, and then explained away.

No product behaviour changes here; the tests now name the helper the compiler
actually emits.

Summary by CodeRabbit

• Bug Fixes

  • Restored coverage for unary + operations on non-numeric values.
  • Corrected assertions for masked indexes, right-hand-side expressions, and straight-line execution scenarios.

• Documentation

  • Added a changelog entry documenting the restored test coverage.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: d3f1e7a8-37c3-4f5e-994c-04294eceb8d1

📥 Commits

Reviewing files that changed from the base of the PR and between 2382a9f and 60e24cc.

📒 Files selected for processing (2)

• changelog.d/8658-restore-lost-unary-pos-assertions.md
• crates/perry-codegen/src/expr/computed_store_rooting_tests.rs

---

📝 Walkthrough

Walkthrough

The change restores three computed store rooting test assertions to match js_dynamic_pos emission for unary + on non-numeric operands. A changelog entry documents the correction. Product behavior is unchanged.

Changes

Unary plus assertion updates

Layer / File(s)	Summary
Update coercion-witness assertions crates/perry-codegen/src/expr/computed_store_rooting_tests.rs, changelog.d/8658-restore-lost-unary-pos-assertions.md	Three assertions now expect js_dynamic_pos for masked index, masked RHS, and straight-line region tests. The changelog records the restored assertions.

Estimated code review effort: 1 (Trivial) | ~2 minutes

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Create stacked PR
• Commit on current branch

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/8658-lost-test-edit

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}