fix(class): delete prototype method removes its keys_array field (#5441)

[proggeramlug]

Summary

Fixes #5441 — the ~760 generated language/{statements,expressions}/class test262 cases that regressed to Uncaught exception: m descriptor should be configurable (language parity 94.5% → 89.3%).

Root cause

The regression came from #5395. That PR made delete C.prototype.<member> on a class-declaration prototype mark the vtable key deleted and early-return, so hasOwnProperty / getOwnPropertyDescriptor would agree the member is gone.

That is correct for instance accessors (get x()) — they live only in the class vtable, with no keys_array entry to remove (the cases #5395 actually tested).

Instance methods are different: install_class_decl_prototype_method_fields plants each method as a real keys_array data field on the prototype object (writable:true, enumerable:false, configurable:true). For a method the early return marked the vtable key deleted but left the real field in place, so hasOwnProperty kept finding it via own_key_present. test262's verifyProperty configurable probe (delete obj[name] then assert the key absent) therefore failed with m descriptor should be configurable.

This is why isolated getOwnPropertyDescriptor looked fine but the full verifyProperty path (used by the flags: [generated] class tests) failed.

Fix

Drop the early return 1 so the class-prototype delete falls through to the existing keys_array scan:

• accessors → scan is a vacuous success (no entry), key stays marked deleted — unchanged behavior;
• methods → scan removes the real field, and the key is still marked deleted — now invisible to both reflective paths.

One-file change in crates/perry-runtime/src/object/delete_rest.rs.

Validation (test262 radar, node v26)

dir	parity	descriptor should be configurable failures
language/statements/class	94.0%	0
language/expressions/class	95.6%	0

• The specific generated repro (new-no-sc-line-method-rs-privatename-identifier-alt.js, which also checks writable:true) now passes.
• Accessor definition/{getters,setters}-prop-desc tests (fix(class): super-in-static resolution + accessor .name reflection (#5345) #5395's targets) still pass — no regression.
• cargo test -p perry-runtime: 1063 passed; the 2 failures are pre-existing/environmental (the known stream test-isolation flake and a cwd-dependent URL test that trips inside a worktree path), not related to this change.

Class-tail tracker: #5345.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Fixed incomplete property deletion behavior where certain class methods and accessors were not being fully removed from objects, improving consistency in deletion operations.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 0325224f-a749-421a-944a-6773cf9acdd3

📥 Commits

Reviewing files that changed from the base of the PR and between ffb7ca9 and a8b3b4e.

📒 Files selected for processing (1)

• crates/perry-runtime/src/object/delete_rest.rs

---

📝 Walkthrough

Walkthrough

In js_object_delete_field, the class-declaration prototype branch previously returned early after marking a vtable member key as deleted. The early return is removed so the function falls through to the keys-array scan, ensuring instance method entries backed by keys_array are also removed.

Changes

Class Prototype Field Deletion Fix

Layer / File(s)	Summary
Remove early return in class vtable member deletion crates/perry-runtime/src/object/delete_rest.rs	For non-constructor class accessor and method members, class_mark_key_deleted is called as before, but the early success return is removed so the function continues to the keys_array scan and removes the backing entry for instance methods. Accessor-only members continue to rely solely on the deleted-key mark.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

• PerryTS/perry#5395: Directly related — this PR modifies the same js_object_delete_field class-declaration prototype branch in delete_rest.rs; the regression introduced there (early return preventing keys_array cleanup) is what this PR fixes.

Poem

🐇 A return too soon, the keys stayed in place,
The descriptor cried: "I'm not configurable, no grace!"
But now I fall through, past vtable and mark,
The keys-array swept clean, no method left dark.
~760 tests cheer from dawn until dusk — ✨

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: fixing a class method deletion issue where prototype methods weren't being properly removed from the keys_array.
Description check	✅ Passed	The description thoroughly explains the regression, root cause, fix, and validation results, covering all key aspects of the change with concrete test262 metrics.
Linked Issues check	✅ Passed	The PR fully addresses the linked issue #5441 by restoring correct class method property descriptors and eliminating the 752 'descriptor should be configurable' failures.
Out of Scope Changes check	✅ Passed	All changes are scoped to fixing the class method deletion issue; no out-of-scope modifications are present beyond the single-file fix to delete_rest.rs.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch worktree-fix-5441-class-descriptor

---

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