🐛 fix(select): raise MemoryError on selector OOM

[gaborbernat]

An allocation failure while compiling a selector set the same flag as a syntax error. Failing the third PyMem allocation of document.select("main > p[zq] a, :is(b, i, [zq]) :has(> a)") raised SelectorSyntaxError ("unexpected token") for a valid selector. The same failure inside a forgiving :is() or :where() list dropped the arm as if it were invalid and returned a result. A failed build of the document's element index fell back to a full walk and hid the failure.

The parser now records an allocation failure as an error with no reason, which every caller, css_to_xpath() and the rewriter included, raises as MemoryError, and a forgiving list stops on it, since Selectors 4 §16.1 ignores only selectors that fail to parse. select(), select_one() and find_all() build the index before they walk and raise MemoryError when the build fails; the same check inside the walk cost select 3,000 instructions a call.

libxml2 2.16.0 gives a failed XPath compile its own XPATH_MEMORY_ERROR, and lexbor 3.1.0's CSS parser stops with LXB_STATUS_ERROR_MEMORY_ALLOCATION instead of reporting an unexpected token. Their Python bindings lose that: lxml 7.0.0b1 wraps every failed compile in XPathSyntaxError, and selectolax 1.0.0 raises SelectolaxError("Can't parse CSS selector."). Servo's selectors 0.41.0 uses Rust's infallible allocation, which aborts the process.

[codspeed]

Merging this PR will degrade performance by 24.51%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

❌ 1 regressed benchmark
✅ 580 untouched benchmarks
⏩ 32 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	test_feature[shadow-slot-comments]	83.3 µs	110.4 µs	-24.51%

Tip

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

---

_{Comparing gaborbernat:fix/alloc-selector-compile (29f7ef8) with main (74c1b54)}

Footnotes

1. 32 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. ↩