GeezerVison

Documentation build overview

📚 cpython-previews | 🛠️ Build #35771966 | 📁 Comparing 4e1d94f against main (e891a91)

  🔍 Preview build  

CIFuzz / cpython3 (memory):

INFO: 3184 files found in /github/workspace/cifuzz-corpus/fuzz_struct_unpack
UnicodeEncodeError: 'ascii' codec can't encode character '\udcce' in position 0: ordinal not in range(128)

@StanFromIreland: Do you know the status of CIFuzz in the 3.13 branch? Is it still used/relevant? The code looks outdated, I suppose that these errors can just be ignored. Maybe we should even disable these CI in the 3.13 branch?

According to Modules/_xxtestfuzz/fuzzer.c, fuzz_struct_unpack() calls struct.unpack(). I downloaded the artifact file: it contains two bytes, 0xce and 0x00. On Python 3.13, I get:

$ ./python
>>> struct.unpack(b'\xce\x00', b'')
UnicodeEncodeError: 'ascii' codec can't encode character '\udcce' in position 0: ordinal not in range(128)

In the main branch, all ValueError are ignored (UnicodeEncodeError is a subclass of ValueError):

    /* Ignore any ValueError, these are triggered by non-ASCII format. */
    if (unpacked == NULL && PyErr_ExceptionMatches(PyExc_ValueError)) {
        PyErr_Clear();
    }

So I suppose that Modules/_xxtestfuzz/fuzzer.c is outdated in the 3.13 branch which no only accepts bug fixes.

I need to discuss with the fuzzer team, but as I've proposed before, I think we should rip it out for branches that aren't main. CC @sethmlarson

I need to discuss with the fuzzer team, but as I've proposed before, I think we should rip it out for branches that aren't main. CC @sethmlarson

I would be fine with that.