[vstinner]

gh-161705: Prevent shadowing stdlib in gzip/tarfile/zipfile CLI (#161706)

Remove sys.path[0] in gzip, tarfile and zipfile command line interface to prevent shadowing standard library modules from extracted files.

• Add sys._path0 variable.
• pymain_sys_path_add_path0() sets sys._path0.
• Add runpy._remove_sys_path0().
• gzip, tarfile and zipfile main functions now call _remove_sys_path0() to prevent shadowing stdlib modules from extracted files.

(cherry picked from commit e891a91)

(cherry picked from commit 8aa5d65)

[read-the-docs-community]

Documentation build overview

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

  🔍 Preview build  

418 files changed · + 1 added · ± 369 modified · - 48 deleted

+ Added

• library/asyncio-policy.html

± Modified

• about.html
• bugs.html
• copyright.html
• glossary.html
• index.html
• license.html
• py-modindex.html
• builtins/constants.html
• builtins/exceptions.html
• builtins/functions.html
• and 359 more...

- Deleted

• improve-page-nojs.html
• improve-page.html
• _static/tachyon-example-flamegraph.html
• _static/tachyon-example-heatmap.html
• builtins/threadsafety.html
• builtins/time-complexity.html
• c-api/extension-modules.html
• c-api/interp-lifecycle.html
• c-api/lifecycle.html
• c-api/profiling.html
• and 38 more...

[vstinner]

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 The code looks outdated, I suppose that these errors can just be ignored.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So I suppose that Modules/_xxtestfuzz/fuzzer.c is outdated in the 3.13 branch which no only accepts bug fixes.

[StanFromIreland]

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

[vstinner]

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.