https://github.com/python/cpython/commit/ff92d4b2b8c534e1357963d843cf4bd7c8cdfad7 commit: ff92d4b2b8c534e1357963d843cf4bd7c8cdfad7 branch: 3.15 author: Miss Islington (bot) <[email protected]> committer: serhiy-storchaka <[email protected]> date: 2026-10-10T12:38:26+03:00 summary:
[3.15] gh-156713: Use the filesystem encoding in nturl2path (GH-156717) (GH-157423) urllib.request.pathname2url() and url2pathname() use the filesystem encoding and error handler since gh-85168, but nturl2path, which implements them on Windows before 3.14, was left unchanged. Paths containing surrogate characters raised UnicodeEncodeError. (cherry picked from commit 01599e203e25610421bb5df413f32fa0848f3ade) Co-authored-by: Serhiy Storchaka <[email protected]> * gh-156713: Fix test_nturl2path for non-UTF-8 filesystem encodings Since nturl2path uses the filesystem encoding, the expected results of quoting non-ASCII characters depend on it. Use os_helper.FS_NONASCII and os_helper.TESTFN_UNDECODABLE, and skip the tests when there is no suitable character or byte sequence for the current encoding. Co-authored-by: Claude Opus 5 (1M context) <[email protected]> (cherry picked from commit 3e1b5d890440ca4ea8f63b16fa27b13ac2a882ef) --------- Co-authored-by: Serhiy Storchaka <[email protected]> Co-authored-by: Claude Opus 5 (1M context) <[email protected]> files: A Misc/NEWS.d/next/Library/2026-08-31-17-10-00.gh-issue-156713.Nt2URL.rst M Lib/nturl2path.py M Lib/test/test_nturl2path.py diff --git a/Lib/nturl2path.py b/Lib/nturl2path.py index 57c7858dff0b81..47c49ca2020c5c 100644 --- a/Lib/nturl2path.py +++ b/Lib/nturl2path.py @@ -22,7 +22,10 @@ def url2pathname(url): # ///C:/foo/bar/spam.foo # become # C:\foo\bar\spam.foo + import sys import urllib.parse + encoding = sys.getfilesystemencoding() + errors = sys.getfilesystemencodeerrors() if url[:3] == '///': # URL has an empty authority section, so the path begins on the third # character. @@ -40,7 +43,8 @@ def url2pathname(url): if url[1:2] == '|': # Older URLs use a pipe after a drive letter url = url[:1] + ':' + url[2:] - return urllib.parse.unquote(url.replace('/', '\\')) + return urllib.parse.unquote(url.replace('/', '\\'), + encoding=encoding, errors=errors) def pathname2url(p): """OS-specific conversion from a file system path to a relative URL @@ -50,7 +54,10 @@ def pathname2url(p): # becomes # ///C:/foo/bar/spam.foo import ntpath + import sys import urllib.parse + encoding = sys.getfilesystemencoding() + errors = sys.getfilesystemencodeerrors() # First, clean up some special forms. We are going to sacrifice # the additional information anyway p = p.replace('\\', '/') @@ -65,10 +72,11 @@ def pathname2url(p): # an authority section with a zero-length authority, and a path # section starting with a single slash. drive = f'///{drive}' - drive = urllib.parse.quote(drive, safe='/:') + drive = urllib.parse.quote(drive, encoding=encoding, errors=errors, + safe='/:') elif root: # Add explicitly empty authority to path beginning with one slash. root = f'//{root}' - tail = urllib.parse.quote(tail) + tail = urllib.parse.quote(tail, encoding=encoding, errors=errors) return drive + root + tail diff --git a/Lib/test/test_nturl2path.py b/Lib/test/test_nturl2path.py index a6a3422a0f75b2..56e47aef5fe4c5 100644 --- a/Lib/test/test_nturl2path.py +++ b/Lib/test/test_nturl2path.py @@ -1,5 +1,9 @@ +import os +import sys import unittest +import urllib.parse +from test.support import os_helper from test.support import warnings_helper @@ -34,7 +38,6 @@ def test_pathname2url(self): self.assertEqual(fn('C:\\a\\b.c\\'), '///C:/a/b.c/') self.assertEqual(fn('C:\\a\\\\b.c'), '///C:/a//b.c') self.assertEqual(fn('C:\\a\\b%#c'), '///C:/a/b%25%23c') - self.assertEqual(fn('C:\\a\\b\xe9'), '///C:/a/b%C3%A9') self.assertEqual(fn('C:\\foo\\bar\\spam.foo'), "///C:/foo/bar/spam.foo") # NTFS alternate data streams self.assertEqual(fn('C:\\foo:bar'), '///C:/foo%3Abar') @@ -45,7 +48,7 @@ def test_pathname2url(self): self.assertEqual(fn("\\\\\\folder\\test\\"), '///folder/test/') self.assertEqual(fn('\\\\some\\share\\'), '//some/share/') self.assertEqual(fn('\\\\some\\share\\a\\b.c'), '//some/share/a/b.c') - self.assertEqual(fn('\\\\some\\share\\a\\b%#c\xe9'), '//some/share/a/b%25%23c%C3%A9') + self.assertEqual(fn('\\\\some\\share\\a\\b%#c'), '//some/share/a/b%25%23c') # Alternate path separator self.assertEqual(fn('C:/a/b.c'), '///C:/a/b.c') self.assertEqual(fn('//some/share/a/b.c'), '//some/share/a/b.c') @@ -58,6 +61,29 @@ def test_pathname2url(self): for url in urls: self.assertEqual(fn(nturl2path.url2pathname(url)), url) + @unittest.skipUnless(os_helper.FS_NONASCII, 'need os_helper.FS_NONASCII') + def test_pathname2url_nonascii(self): + encoding = sys.getfilesystemencoding() + errors = sys.getfilesystemencodeerrors() + char = os_helper.FS_NONASCII + quoted = urllib.parse.quote(char, encoding=encoding, errors=errors) + self.assertEqual(nturl2path.pathname2url(f'C:\\a\\b{char}'), + '///C:/a/b' + quoted) + self.assertEqual(nturl2path.pathname2url(f'\\\\some\\share\\a\\b{char}'), + '//some/share/a/b' + quoted) + + @unittest.skipUnless(os_helper.TESTFN_UNDECODABLE, + 'need os_helper.TESTFN_UNDECODABLE') + def test_pathname2url_surrogates(self): + # gh-156713: the filesystem encoding and error handler are used, + # so that paths containing surrogate characters can be converted. + encoding = sys.getfilesystemencoding() + errors = sys.getfilesystemencodeerrors() + path = os.fsdecode(os_helper.TESTFN_UNDECODABLE) + url = urllib.parse.quote(path, encoding=encoding, errors=errors) + self.assertEqual(nturl2path.pathname2url('C:\\' + path), + '///C:/' + url) + def test_url2pathname(self): fn = nturl2path.url2pathname self.assertEqual(fn('/'), '\\') @@ -103,5 +129,18 @@ def test_url2pathname(self): self.assertEqual(fn(nturl2path.pathname2url(path)), path) + @unittest.skipUnless(os_helper.TESTFN_UNDECODABLE, + 'need os_helper.TESTFN_UNDECODABLE') + def test_url2pathname_surrogates(self): + # gh-156713: the filesystem encoding and error handler are used, so + # that URLs containing percent-encoded surrogates can be converted. + encoding = sys.getfilesystemencoding() + errors = sys.getfilesystemencodeerrors() + path = os.fsdecode(os_helper.TESTFN_UNDECODABLE) + url = urllib.parse.quote(path, encoding=encoding, errors=errors) + self.assertEqual(nturl2path.url2pathname('///C:/' + url), + 'C:\\' + path) + + if __name__ == '__main__': unittest.main() diff --git a/Misc/NEWS.d/next/Library/2026-08-31-17-10-00.gh-issue-156713.Nt2URL.rst b/Misc/NEWS.d/next/Library/2026-08-31-17-10-00.gh-issue-156713.Nt2URL.rst new file mode 100644 index 00000000000000..1d21fc70fd9256 --- /dev/null +++ b/Misc/NEWS.d/next/Library/2026-08-31-17-10-00.gh-issue-156713.Nt2URL.rst @@ -0,0 +1,4 @@ +Fix :func:`!nturl2path.pathname2url` and :func:`!nturl2path.url2pathname`: +the filesystem encoding and error handler are now used for percent-encoding +and decoding, as in :mod:`urllib.request`. Previously paths containing +surrogate characters raised :exc:`UnicodeEncodeError`. _______________________________________________ Python-checkins mailing list -- [email protected] To unsubscribe send an email to [email protected] https://mail.python.org/mailman3//lists/python-checkins.python.org Member address: [email protected]
