[davidhewitt]

This PR fixes two cases of incorrect inheritance of tp_alloc or tp_free causing crashes with various #[pyclass] types:

• We weren't previously explicitly setting the allocator, except for #[pyclass(freelist)] and invalid base allocators might not allocate space for the Rust data (this happens to be the case for datetime.time, for example). This could be viewed as a bug in the base allocator but I think it doesn't hurt for us to be defensive and select PyType_GenericAlloc explicitly (this is the documented behaviour for heap types but not what happens in practice, see also documentation of tp_alloc and tp_free inheritance for heap types is incorrect python/cpython#160749).

• For #[pyclass(freelist)] struct Base, an inherited #[pyclass] struct Child would accidentally inherit Base deallocator and corrupt the base freelist with a Child object.

Whilst fixing these cases I decided to make TP_ALLOC and TP_FREE constants inside the PyClassImpl trait, which allows for the most concise generated code while setting the defaults & allowing the freelist override.

Although both are nasty crashes, I think they would be discovered during development so not CVE-worthy.

Bug discovery credit OpenAI (OutboundDisclosures@openai.com)

[codspeed]

Merging this PR will degrade performance by 17.55%

1 warning

⚠️ Base and head ran on different runner types

Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.

Learn more about Macro Runners

Open the report in CodSpeed to investigate

❌ 2 regressed benchmarks
✅ 141 untouched benchmarks
⏩ 6 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	into_bigint_big_negative	4.2 µs	5.4 µs	-22.2%
❌	into_biguint_big	4.2 µs	4.8 µs	-12.63%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing davidhewitt:explicit-tp-alloc (68ad4b3) with main (26f9ec8)}

Footnotes

1. 6 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[Icxolu]

LGTM