Exception on unencodable (file)name #533
Labels
bug
Something isn't working
Comments
|
Fixed in 696d259b6031 (bb). Figuring out how to reproduce and test the problem was the biggest challenge! |
|
This fix was released as part of Coverage.py 4.3. |
nedbat
added a commit
that referenced
this issue
Jan 3, 2022
…nly on 3.6
This was the failure it was protecting against on Python 3.6:
```
___________________________ ProcessTest.test_lang_c ____________________________
[gw0] linux -- Python 3.6.15 /home/runner/work/coveragepy/coveragepy/.tox/py36/bin/python
self = <tests.test_process.ProcessTest object at 0x7fe57e63b198>
@pytest.mark.skipif(env.PYPY, reason="PyPy is unreliable with this test")
# Jython as of 2.7.1rc3 won't compile a filename that isn't utf-8.
@pytest.mark.skipif(env.JYTHON, reason="Jython can't handle this test")
def test_lang_c(self):
# LANG=C forces getfilesystemencoding on Linux to 'ascii', which causes
# failures with non-ascii file names. We don't want to make a real file
# with strange characters, though, because that gets the test runners
# tangled up. This will isolate the concerns to the coverage.py code.
# #533
self.make_file("weird_file.py", r"""
globs = {}
code = "a = 1\nb = 2\n"
exec(compile(code, "wut\xe9\xea\xeb\xec\x01\x02.py", 'exec'), globs)
print(globs['a'])
print(globs['b'])
""")
self.set_environ("LANG", "C")
out = self.run_command("coverage run weird_file.py")
> assert out == "1\n2\n"
E assert 'Traceback (m...ion by zero\n' == '1\n2\n'
E - 1
E - 2
E + Traceback (most recent call last):
E + File "/home/runner/work/coveragepy/coveragepy/coverage/files.py", line 149, in abs_file
E + path = os.path.realpath(path)
E + File "/opt/hostedtoolcache/Python/3.6.15/x64/lib/python3.6/posixpath.py", line 395, in realpath
E + path, ok = _joinrealpath(filename[:0], filename, {})
E + File "/opt/hostedtoolcache/Python/3.6.15/x64/lib/python3.6/posixpath.py", line 429, in _joinrealpath
E + if not islink(newpath):
E + File "/opt/hostedtoolcache/Python/3.6.15/x64/lib/python3.6/posixpath.py", line 171, in islink
E + st = os.lstat(path)
E + UnicodeEncodeError: 'ascii' codec can't encode characters in position 3-6: ordinal not in range(128)
```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Originally reported by Antoine Pitrou (Bitbucket: pitrou, GitHub: pitrou)
Some code generators (e.g. Template engines) can produce fake module names. Recently a test in the Tornado test suite has started failing with a non-ascii fake module name and a C locale:
https://travis-ci.org/tornadoweb/tornado/jobs/176023596#L1120
I would suggest that coverage.py catch UnicodeEncodeError at that point and consider the file as non-existing.
The text was updated successfully, but these errors were encountered: