[xiehuanyi]

plist([1, 2, 3])[-1:] currently raises IndexError, although PList supports sequence slicing. The optimized suffix-slice path passes the negative start directly to _drop, instead of counting from the end and clipping it as a built-in list does. That path also accepts floating-point starts and rejects objects that implement __index__.

Normalize suffix-slice starts while retaining the existing shared tail. Stop dropping once the list is empty, so an oversized positive start does not keep traversing the empty-list sentinel.

Validation on Python 3.12:

• Regression cases before the fix: 19 failed, 16 passed.
• pytest tests/list_test.py: 73 passed, including comparisons with built-in list slicing and shared-tail checks.
• pytest: 571 passed, 102 skipped (optional C-extension cases are unavailable in this local environment).
• git diff --check: passed.

Implementation and regression tests prepared with Codex; the patch and test results were reviewed against the existing PList code.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[carlitogwapito]

One small error-path test to add alongside this: test-only commit.

It compares plist with a built-in list when __index__ itself raises. Both should preserve the same exception object, call the provider once, and leave the input unchanged. The new case passes here; the base raises a different TypeError before calling the provider. The built-in oracle and an ordinary shared-tail control pass on both versions.

Focused checks used Windows / CPython 3.14.4 with the pure-Python backend. The full suite, C-extension tests, and other interpreter/platform combinations weren't run.

Optional suggestion: keep this beside the successful __index__ and wrong-type cases, since a provider's own failure is a separate path from an invalid index value.

[xiehuanyi]

Thanks for the focused __index__ error-path test. Checking exception identity, a single invocation, and unchanged input is useful additional coverage. I have reviewed the proposed test, but have not applied the optional commit or run it locally.