gh-82367: Use FindFirstFile Win32 API in ntpath.realpath() (#110298)

moonsikpark · moonsikpark · commit 2b5a29ee1f04 · 2023-10-05T17:38:24.000+09:00
* Use `FindFirstFile` Win32 API to fix a bug where `ntpath.realpath()`
breaks out of traversing a series of paths where a (handled)
`ERROR_ACCESS_DENIED` or `ERROR_SHARING_VIOLATION` occurs.
* Update docs to reflect that `ntpath.realpath()` eliminates MS-DOS
style names.
diff --git a/Doc/library/os.path.rst b/Doc/library/os.path.rst
@@ -377,7 +377,8 @@ the :mod:`glob` module.)
 
    Return the canonical path of the specified filename, eliminating any symbolic
    links encountered in the path (if they are supported by the operating
-   system).
+   system). On Windows, this function will also resolve MS-DOS (also called 8.3)
+   style names such as ``C:\\PROGRA~1`` to ``C:\\Program Files``.
 
    If a path doesn't exist or a symlink loop is encountered, and *strict* is
    ``True``, :exc:`OSError` is raised. If *strict* is ``False``, the path is
diff --git a/Doc/whatsnew/3.13.rst b/Doc/whatsnew/3.13.rst
@@ -97,6 +97,9 @@ Other Language Changes
   if supported.
   (Contributed by Victor Stinner in :gh:`109649`.)
 
+* :func:`os.path.realpath` now resolves MS-DOS style file names even if
+  the file is not accessible.
+  (Contributed by Moonsik Park in :gh:`82367`.)
 
 New Modules
 ===========
diff --git a/Lib/ntpath.py b/Lib/ntpath.py
@@ -23,7 +23,6 @@
 import genericpath
 from genericpath import *
 
-
 __all__ = ["normcase","isabs","join","splitdrive","splitroot","split","splitext",
            "basename","dirname","commonprefix","getsize","getmtime",
            "getatime","getctime", "islink","exists","lexists","isdir","isfile",
@@ -601,7 +600,7 @@ def abspath(path):
             return _abspath_fallback(path)
 
 try:
-    from nt import _getfinalpathname, readlink as _nt_readlink
+    from nt import _findfirstfile, _getfinalpathname, readlink as _nt_readlink
 except ImportError:
     # realpath is a no-op on systems without _getfinalpathname support.
     realpath = abspath
@@ -688,10 +687,15 @@ def _getfinalpathname_nonstrict(path):
                 except OSError:
                     # If we fail to readlink(), let's keep traversing
                     pass
-                path, name = split(path)
-                # TODO (bpo-38186): Request the real file name from the directory
-                # entry using FindFirstFileW. For now, we will return the path
-                # as best we have it
+                # If we get these errors, try to get the real name of the file without accessing it.
+                if ex.winerror in (1, 5, 32, 50, 87, 1920, 1921):
+                    try:
+                        name = _findfirstfile(path)
+                        path, _ = split(path)
+                    except OSError:
+                        path, name = split(path)
+                else:
+                    path, name = split(path)
                 if path and not name:
                     return path + tail
                 tail = join(name, tail) if tail else name
diff --git a/Lib/test/test_ntpath.py b/Lib/test/test_ntpath.py
@@ -2,6 +2,7 @@
 import ntpath
 import os
 import string
+import subprocess
 import sys
 import unittest
 import warnings
@@ -637,6 +638,48 @@ def test_realpath_cwd(self):
         with os_helper.change_cwd(test_dir_short):
             self.assertPathEqual(test_file_long, ntpath.realpath("file.txt"))
 
+    @unittest.skipUnless(HAVE_GETFINALPATHNAME, 'need _getfinalpathname')
+    def test_realpath_permission(self):
+        # Test whether python can resolve the real filename of a
+        # shortened file name even if it does not have permission to access it.
+        ABSTFN = ntpath.realpath(os_helper.TESTFN)
+
+        os_helper.unlink(ABSTFN)
+        os_helper.rmtree(ABSTFN)
+        os.mkdir(ABSTFN)
+        self.addCleanup(os_helper.rmtree, ABSTFN)
+
+        test_file = ntpath.join(ABSTFN, "LongFileName123.txt")
+        test_file_short = ntpath.join(ABSTFN, "LONGFI~1.TXT")
+
+        with open(test_file, "wb") as f:
+            f.write(b"content")
+        # Automatic generation of short names may be disabled on
+        # NTFS volumes for the sake of performance.
+        # They're not supported at all on ReFS and exFAT.
+        subprocess.run(
+            # Try to set the short name manually.
+            ['fsutil.exe', 'file', 'setShortName', test_file, 'LONGFI~1.TXT'],
+            creationflags=subprocess.DETACHED_PROCESS
+        )
+
+        try:
+            self.assertPathEqual(test_file, ntpath.realpath(test_file_short))
+        except AssertionError:
+            raise unittest.SkipTest('the filesystem seems to lack support for short filenames')
+
+        # Deny the right to [S]YNCHRONIZE on the file to
+        # force nt._getfinalpathname to fail with ERROR_ACCESS_DENIED.
+        p = subprocess.run(
+            ['icacls.exe', test_file, '/deny', '*S-1-5-32-545:(S)'],
+            creationflags=subprocess.DETACHED_PROCESS
+        )
+
+        if p.returncode:
+            raise unittest.SkipTest('failed to deny access to the test file')
+
+        self.assertPathEqual(test_file, ntpath.realpath(test_file_short))
+
     def test_expandvars(self):
         with os_helper.EnvironmentVarGuard() as env:
             env.clear()
diff --git a/Misc/ACKS b/Misc/ACKS
@@ -1373,6 +1373,7 @@ Peter Parente
 Alexandre Parenteau
 Dan Parisien
 HyeSoo Park
+Moonsik Park
 William Park
 Claude Paroz
 Heikki Partanen
diff --git a/Misc/NEWS.d/next/Windows/2023-10-03-12-30-59.gh-issue-82367.nxwfMx.rst b/Misc/NEWS.d/next/Windows/2023-10-03-12-30-59.gh-issue-82367.nxwfMx.rst
@@ -0,0 +1,2 @@
+:func:`os.path.realpath` now resolves MS-DOS style file names even if
+the file is not accessible. Patch by Moonsik Park.
diff --git a/Modules/clinic/posixmodule.c.h b/Modules/clinic/posixmodule.c.h
diff --git a/Modules/posixmodule.c b/Modules/posixmodule.c
@@ -4809,6 +4809,37 @@ os__getfinalpathname_impl(PyObject *module, path_t *path)
     return result;
 }
 
+/*[clinic input]
+os._findfirstfile
+    path: path_t
+    /
+A function to get the real file name without accessing the file in Windows.
+[clinic start generated code]*/
+
+static PyObject *
+os__findfirstfile_impl(PyObject *module, path_t *path)
+/*[clinic end generated code: output=106dd3f0779c83dd input=0734dff70f60e1a8]*/
+{
+    PyObject *result;
+    HANDLE hFindFile;
+    WIN32_FIND_DATAW wFileData;
+    WCHAR *wRealFileName;
+
+    Py_BEGIN_ALLOW_THREADS
+    hFindFile = FindFirstFileW(path->wide, &wFileData);
+    Py_END_ALLOW_THREADS
+
+    if (hFindFile == INVALID_HANDLE_VALUE) {
+        path_error(path);
+        return NULL;
+    }
+
+    wRealFileName = wFileData.cFileName;
+    result = PyUnicode_FromWideChar(wRealFileName, wcslen(wRealFileName));
+    FindClose(hFindFile);
+    return result;
+}
+
 
 /*[clinic input]
 os._getvolumepathname
@@ -15961,6 +15992,7 @@ static PyMethodDef posix_methods[] = {
     OS__GETFULLPATHNAME_METHODDEF
     OS__GETDISKUSAGE_METHODDEF
     OS__GETFINALPATHNAME_METHODDEF
+    OS__FINDFIRSTFILE_METHODDEF
     OS__GETVOLUMEPATHNAME_METHODDEF
     OS__PATH_SPLITROOT_METHODDEF
     OS__PATH_NORMPATH_METHODDEF

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	+:func:`os.path.realpath` now resolves MS-DOS style file names even if
	`2`	`+the file is not accessible. Patch by Moonsik Park.`