`_remote_debugging`: non-ASCII both filenames and function names garbled

[maurycy]

Bug report

Bug description:

Ran into this one accidentally (one of the perks of having name with Polish diacritic characters is doing QA like this all the time.)

Basically, any non-ASCII filename, qualnmame, module, task seems to return garbage strings (NUL etc.), and they're valid since PEP 3131.

I did not investigate it fully, but it's likely caused by hard-coding asciiobject_size:

cpython/Modules/_remote_debugging/object_reading.c

Lines 82 to 88 in ae6adc9

	size_t offset = (size_t)unwinder->debug_offsets.unicode_object.asciiobject_size;
	res = _Py_RemoteDebug_PagedReadRemoteMemory(&unwinder->handle, address + offset, len, buf);
	if (res < 0) {
	set_exception_cause(unwinder, PyExc_RuntimeError, "Failed to read string data from remote memory");
	goto err;
	}
	buf[len] = '\0';

PS. If you're wondering what is zażółć gęślą jaźń.

Reproduction

2026-05-01T11:19:14.212125000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) % cat /tmp/zażółćgęśląjaźń.py
def zażółć_gęślą_jaźń():
    s = 0
    for i in range(10_000_000):
        s += i * i
    return s

def główna():
    for _ in range(20):
        zażółć_gęślą_jaźń()

główna()
2026-05-01T11:19:14.489608000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) % sudo ./python.exe -m profiling.sampling run --collapsed -o /tmp/profile.folded /tmp/zażółćgęśląjaźń.py
Captured 7,135 samples in 7.14 seconds
Sample rate: 999.98 samples/sec
Error rate: 0.07
Collapsed stack output written to /tmp/profile.folded
2026-05-01T11:19:23.243613000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) % head -n 10 /tmp/profile.folded
tid:12268318;<frozen runpy>:_run_module_as_main:196;<frozen runpy>:_run_code:87;tmp:<module>:11;tmp::9;tmp:z:4 6250
tid:12268318;<frozen runpy>:_run_module_as_main:196;<frozen runpy>:_run_code:87;tmp:<module>:11;tmp::9;tmp:z:3 880
2026-05-01T11:19:28.606723000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) %

or:

2026-05-01T11:20:40.003826000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) % cat /tmp/zażółć3.py 
import _remote_debugging, os
def zażółć():
    u = _remote_debugging.RemoteUnwinder(os.getpid(), all_threads=True)
    for interp in u.get_stack_trace():
        for tid, _, frames in interp[1]:
            for f in frames:
                print(repr(f[0]), '|', repr(f[2]))
zażółć()
2026-05-01T11:20:44.723466000+0200 maurycy@gimel /Users/maurycy/src/github.com/maurycy/cpython (main c0e0640) % ./python.exe /tmp/zażółć3.py
'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00' | '\x00\x00\x00\x00\x00\x00'
'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00' | '<module>'

CPython versions tested on:

CPython main branch

Operating systems tested on:

macOS

Linked PRs

• gh-149214: Fix non ascii characters in remote debugging #149228

[maurycy]

cc @pablogsal As always, can do if you want

[pablogsal]

I will take a look probably this weekend but if you want to give it a go before happy to review your PR!

[maurycy]

@pablogsal Thank you for taking this over :-)

I started digging but haven't gotten too far yet, apart from wondering whether the existing unicode_obj is helpful or not.