gh-111089: PyUnicode_AsUTF8AndSize() sets size on error

[vstinner]

On error, PyUnicode_AsUTF8AndSize() now sets the size argument to 0, to avoid undefined value.

• Issue: [C API] Change PyUnicode_AsUTF8() to return NULL on embedded null characters #111089

---

📚 Documentation preview 📚: https://cpython-previews--111106.org.readthedocs.build/

[vstinner]

See also PR #111100: Add tests for failing PyUnicode_AsUTF8AndSize() with psize=NULL.

[serhiy-storchaka]

LGTM. See also #110865.

[serhiy-storchaka]

Suggested change

	On error, set an exception, set *size* to 0 (if it's not NULL) and return
	On error, set an exception, set *size* to ``0`` (if it's not ``NULL``) and return

[serhiy-storchaka]

Why set size to 0, and not to -1? There is larger chance to miss error. For example PyUnicode_FromStringAndSize(NULL, 0) and PyBytes_FromStringAndSize(NULL, 0) return empty string and empty bytes object, but if size is negative, they raise exception.

[vstinner]

Why set size to 0, and not to -1?

That's a very good idea! I updated my PR to set size of -1 on error.

I just picked 0 randomly. I forgot that size can be negative.

[vstinner]

Oh, tests failed with:

Testing of doctests in the sources finished, look at the results in build/doctest/output.txt.
python: ./Include/internal/pycore_typeobject.h:105: _PyType_GetModuleState: Assertion `et->ht_module' failed.
Aborted (core dumped)

[vstinner]

I failed to reproduce the Sphinx crash locally. And the test passed when run again.

I merged my PR, thanks for review Serhiy.