Work around case-insensitivity issues with cwd on Windows

On Windows, file names are recorded case-sensitively, but looked up case-insensitively. Therefore, it is possible to switch to a directory by using incorrect case, e.g. cd documentation will still get you into the Documentation subdirectory.

In Powershell, doing so will however report the current directory with the specified spelling rather than the one recorded on disk, and Git will get confused.

To remedy that, we fixed this in Git for Windows more than three years ago, and needed only a small fix a couple of months later to accommodate for the diverse scenarios encountered by the many Git for Windows users.

Not only to keep the story closer to what happened historically, but also to make it easier to follow, I refrained from squashing these two patches.

Side note: the second patch is technically not battle-tested for that long:
it uses an API function that requires Windows Vista or later, and we only recently started to clean up Git for Windows' code to drop fallbacks for Windows XP. Read: this code used to load the GetFinalPathNameByHandle() function dynamically, and that is the only difference to the code that has been "battle-tested" for close to three years.

Anton Serbulov (1):
mingw: fix getcwd when the parent directory cannot be queried

Johannes Schindelin (1):
mingw: ensure getcwd() reports the correct case

 compat/mingw.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 48 insertions(+), 2 deletions(-)

base-commit: c4df23f7927d8d00e666a3c8d1b3375f1dc8a3c1
Published-As: https://github.com/gitgitgadget/git/releases/tags/pr-54%2Fdscho%2Fmingw-getcwd-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-54/dscho/mingw-getcwd-v1
Pull-Request: https://github.com/gitgitgadget/git/pull/54

gitgitgadget

From: Johannes Schindelin <[email protected]>

When switching the current working directory, say, in PowerShell, it is quite possible to use a different capitalization than the one that is recorded on disk. While doing the same in cmd.exe adjusts the capitalization magically, that does not happen in PowerShell so that getcwd() returns the current directory in a different way than is recorded on disk.

Typically this creates no problems except when you call

git log .

in a subdirectory called, say, "GIT/" but you switched to "Git/" and your getcwd() reports the latter, then Git won't understand that you wanted to see the history as per the GIT/ subdirectory but it thinks you wanted to see the history of some directory that may have existed in the past (but actually never did).

So let's be extra careful to adjust the capitalization of the current directory before working with it.

Reported by a few PowerShell power users ;-)

---
 compat/mingw.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

index 18caf2196..2c3e27ce9 100644
@@ -917,8 +917,15 @@ struct tm *localtime_r(const time_t *timep, struct tm *result)
 
 char *mingw_getcwd(char *pointer, int len)
 {
-	wchar_t wpointer[MAX_PATH];
-	if (!_wgetcwd(wpointer, ARRAY_SIZE(wpointer)))
+	wchar_t cwd[MAX_PATH], wpointer[MAX_PATH];
+	DWORD ret = GetCurrentDirectoryW(ARRAY_SIZE(cwd), cwd);
+
+	if (!ret || ret >= ARRAY_SIZE(cwd)) {
+		errno = ret ? ENAMETOOLONG : err_win_to_posix(GetLastError());
+		return NULL;
+	}
+	ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer));
+	if (!ret || ret >= ARRAY_SIZE(wpointer))
 		return NULL;
 	if (xwcstoutf(pointer, wpointer, len) < 0)
 		return NULL;

gitgitgadget

From: Anton Serbulov <[email protected]>

GetLongPathName() function may fail when it is unable to query the parent directory of a path component to determine the long name for that component. It happens, because of it uses FindFirstFile() function for each next short part of path. The FindFirstFile() requires List Directory and Synchronize desired access for a calling process.

In case of lacking such permission for some part of path, the GetLongPathName() returns 0 as result and GetLastError() returns ERROR_ACCESS_DENIED.

GetFinalPathNameByHandle() function can help in such cases, because it requires Read Attributes and Synchronize desired access to the target path only.

The GetFinalPathNameByHandle() function was introduced on Windows Server 2008/Windows Vista. So we need to load it dynamically.

CreateFile() parameters:
lpFileName = path to the current directory dwDesiredAccess = 0 (it means Read Attributes and Synchronize) dwShareMode = FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE
(it prevents Sharing Violation)
lpSecurityAttributes = NULL (default security attributes) dwCreationDisposition = OPEN_EXISTING
(required to obtain a directory handle)
dwFlagsAndAttributes = FILE_FLAG_BACKUP_SEMANTICS
(required to obtain a directory handle)
hTemplateFile = NULL (when opening an existing file or directory,
CreateFile ignores this parameter)

The string that is returned by GetFinalPathNameByHandle() function uses the \\?\ syntax. To skip the prefix and convert backslashes to slashes, the normalize_ntpath() mingw function will be used.

Note: GetFinalPathNameByHandle() function returns a final path. It is the path that is returned when a path is fully resolved. For example, for a symbolic link named "C:\tmp\mydir" that points to "D:\yourdir", the final path would be "D:\yourdir".

---
 compat/mingw.c | 39 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

index 2c3e27ce9..19addfa5d 100644
@@ -202,6 +202,31 @@ static int ask_yes_no_if_possible(const char *format, ...)
 	}
 }
 
+/* Normalizes NT paths as returned by some low-level APIs. */
+static wchar_t *normalize_ntpath(wchar_t *wbuf)
+{
+	int i;
+	/* fix absolute path prefixes */
+	if (wbuf[0] == '\\') {
+		/* strip NT namespace prefixes */
+		if (!wcsncmp(wbuf, L"\\??\\", 4) ||
+		    !wcsncmp(wbuf, L"\\\\?\\", 4))
+			wbuf += 4;
+		else if (!wcsnicmp(wbuf, L"\\DosDevices\\", 12))
+			wbuf += 12;
+		/* replace remaining '...UNC\' with '\\' */
+		if (!wcsnicmp(wbuf, L"UNC\\", 4)) {
+			wbuf += 2;
+			*wbuf = '\\';
+		}
+	}
+	/* convert backslashes to slashes */
+	for (i = 0; wbuf[i]; i++)
+		if (wbuf[i] == '\\')
+			wbuf[i] = '/';
+	return wbuf;
+}
+
 int mingw_unlink(const char *pathname)
 {
 	int ret, tries = 0;

@@ -925,6 +950,20 @@ char *mingw_getcwd(char *pointer, int len)
 		return NULL;
 	}
 	ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer));
+	if (!ret && GetLastError() == ERROR_ACCESS_DENIED) {
+		HANDLE hnd = CreateFileW(cwd, 0,
+			FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL,
+			OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
+		if (hnd == INVALID_HANDLE_VALUE)
+			return NULL;
+		ret = GetFinalPathNameByHandleW(hnd, wpointer, ARRAY_SIZE(wpointer), 0);
+		CloseHandle(hnd);
+		if (!ret || ret >= ARRAY_SIZE(wpointer))
+			return NULL;
+		if (xwcstoutf(pointer, normalize_ntpath(wpointer), len) < 0)
+			return NULL;
+		return pointer;
+	}
 	if (!ret || ret >= ARRAY_SIZE(wpointer))
 		return NULL;
 	if (xwcstoutf(pointer, wpointer, len) < 0)

gitgitgadget

s/of it/it/

BTW - It was "Stephen" that sent the email. :-)

Hi Stephen,

Thanks, fixed in v2,
Dscho

On Windows, file names are recorded case-sensitively, but looked up case-insensitively. Therefore, it is possible to switch to a directory by using incorrect case, e.g. cd documentation will still get you into the Documentation subdirectory.

In Powershell, doing so will however report the current directory with the specified spelling rather than the one recorded on disk, and Git will get confused.

To remedy that, we fixed this in Git for Windows more than three years ago, and needed only a small fix a couple of months later to accommodate for the diverse scenarios encountered by the many Git for Windows users.

Not only to keep the story closer to what happened historically, but also to make it easier to follow, I refrained from squashing these two patches.

Side note: the second patch is technically not battle-tested for that long:
it uses an API function that requires Windows Vista or later, and we only recently started to clean up Git for Windows' code to drop fallbacks for Windows XP. Read: this code used to load the GetFinalPathNameByHandle() function dynamically, and that is the only difference to the code that has been "battle-tested" for close to three years.

Changes since v1:

* Fixed a grammar mistake in the second commit message.

Anton Serbulov (1):
mingw: fix getcwd when the parent directory cannot be queried

Johannes Schindelin (1):
mingw: ensure getcwd() reports the correct case

 compat/mingw.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 48 insertions(+), 2 deletions(-)

base-commit: c4df23f7927d8d00e666a3c8d1b3375f1dc8a3c1
Published-As: https://github.com/gitgitgadget/git/releases/tags/pr-54%2Fdscho%2Fmingw-getcwd-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-54/dscho/mingw-getcwd-v2
Pull-Request: https://github.com/gitgitgadget/git/pull/54

Range-diff vs v1:

 1:  e13ae2338 = 1:  e13ae2338 mingw: ensure `getcwd()` reports the correct case
 2:  3e3b1c3b1 ! 2:  87ef9ac63 mingw: fix getcwd when the parent directory cannot be queried

@@ -4,7 +4,7 @@

GetLongPathName() function may fail when it is unable to query the parent directory of a path component to determine the long name
- for that component. It happens, because of it uses FindFirstFile()
+ for that component. It happens, because it uses FindFirstFile()
function for each next short part of path. The FindFirstFile() requires List Directory and Synchronize desired access for a calling process.

gitgitgadget

From: Johannes Schindelin <[email protected]>

When switching the current working directory, say, in PowerShell, it is quite possible to use a different capitalization than the one that is recorded on disk. While doing the same in cmd.exe adjusts the capitalization magically, that does not happen in PowerShell so that getcwd() returns the current directory in a different way than is recorded on disk.

Typically this creates no problems except when you call

git log .

in a subdirectory called, say, "GIT/" but you switched to "Git/" and your getcwd() reports the latter, then Git won't understand that you wanted to see the history as per the GIT/ subdirectory but it thinks you wanted to see the history of some directory that may have existed in the past (but actually never did).

So let's be extra careful to adjust the capitalization of the current directory before working with it.

Reported by a few PowerShell power users ;-)

---
 compat/mingw.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

index 18caf2196..2c3e27ce9 100644
@@ -917,8 +917,15 @@ struct tm *localtime_r(const time_t *timep, struct tm *result)
 
 char *mingw_getcwd(char *pointer, int len)
 {
-	wchar_t wpointer[MAX_PATH];
-	if (!_wgetcwd(wpointer, ARRAY_SIZE(wpointer)))
+	wchar_t cwd[MAX_PATH], wpointer[MAX_PATH];
+	DWORD ret = GetCurrentDirectoryW(ARRAY_SIZE(cwd), cwd);
+
+	if (!ret || ret >= ARRAY_SIZE(cwd)) {
+		errno = ret ? ENAMETOOLONG : err_win_to_posix(GetLastError());
+		return NULL;
+	}
+	ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer));
+	if (!ret || ret >= ARRAY_SIZE(wpointer))
 		return NULL;
 	if (xwcstoutf(pointer, wpointer, len) < 0)
 		return NULL;

gitgitgadget

From: Anton Serbulov <[email protected]>

GetLongPathName() function may fail when it is unable to query the parent directory of a path component to determine the long name for that component. It happens, because it uses FindFirstFile() function for each next short part of path. The FindFirstFile() requires List Directory and Synchronize desired access for a calling process.

In case of lacking such permission for some part of path, the GetLongPathName() returns 0 as result and GetLastError() returns ERROR_ACCESS_DENIED.

GetFinalPathNameByHandle() function can help in such cases, because it requires Read Attributes and Synchronize desired access to the target path only.

The GetFinalPathNameByHandle() function was introduced on Windows Server 2008/Windows Vista. So we need to load it dynamically.

CreateFile() parameters:
lpFileName = path to the current directory dwDesiredAccess = 0 (it means Read Attributes and Synchronize) dwShareMode = FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE
(it prevents Sharing Violation)
lpSecurityAttributes = NULL (default security attributes) dwCreationDisposition = OPEN_EXISTING
(required to obtain a directory handle)
dwFlagsAndAttributes = FILE_FLAG_BACKUP_SEMANTICS
(required to obtain a directory handle)
hTemplateFile = NULL (when opening an existing file or directory,
CreateFile ignores this parameter)

The string that is returned by GetFinalPathNameByHandle() function uses the \\?\ syntax. To skip the prefix and convert backslashes to slashes, the normalize_ntpath() mingw function will be used.

Note: GetFinalPathNameByHandle() function returns a final path. It is the path that is returned when a path is fully resolved. For example, for a symbolic link named "C:\tmp\mydir" that points to "D:\yourdir", the final path would be "D:\yourdir".

---
 compat/mingw.c | 39 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

index 2c3e27ce9..19addfa5d 100644
@@ -202,6 +202,31 @@ static int ask_yes_no_if_possible(const char *format, ...)
 	}
 }
 
+/* Normalizes NT paths as returned by some low-level APIs. */
+static wchar_t *normalize_ntpath(wchar_t *wbuf)
+{
+	int i;
+	/* fix absolute path prefixes */
+	if (wbuf[0] == '\\') {
+		/* strip NT namespace prefixes */
+		if (!wcsncmp(wbuf, L"\\??\\", 4) ||
+		    !wcsncmp(wbuf, L"\\\\?\\", 4))
+			wbuf += 4;
+		else if (!wcsnicmp(wbuf, L"\\DosDevices\\", 12))
+			wbuf += 12;
+		/* replace remaining '...UNC\' with '\\' */
+		if (!wcsnicmp(wbuf, L"UNC\\", 4)) {
+			wbuf += 2;
+			*wbuf = '\\';
+		}
+	}
+	/* convert backslashes to slashes */
+	for (i = 0; wbuf[i]; i++)
+		if (wbuf[i] == '\\')
+			wbuf[i] = '/';
+	return wbuf;
+}
+
 int mingw_unlink(const char *pathname)
 {
 	int ret, tries = 0;

@@ -925,6 +950,20 @@ char *mingw_getcwd(char *pointer, int len)
 		return NULL;
 	}
 	ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer));
+	if (!ret && GetLastError() == ERROR_ACCESS_DENIED) {
+		HANDLE hnd = CreateFileW(cwd, 0,
+			FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL,
+			OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
+		if (hnd == INVALID_HANDLE_VALUE)
+			return NULL;
+		ret = GetFinalPathNameByHandleW(hnd, wpointer, ARRAY_SIZE(wpointer), 0);
+		CloseHandle(hnd);
+		if (!ret || ret >= ARRAY_SIZE(wpointer))
+			return NULL;
+		if (xwcstoutf(pointer, normalize_ntpath(wpointer), len) < 0)
+			return NULL;
+		return pointer;
+	}
 	if (!ret || ret >= ARRAY_SIZE(wpointer))
 		return NULL;
 	if (xwcstoutf(pointer, wpointer, len) < 0)

gitgitgadget

"Johannes Schindelin via GitGitGadget" <[email protected]>

Thanks. I think this matches what I earlier queued on 'pu' with Stephen's typofix squashed in, so we are good already.

Hi Junio,

Perfect. Sorry that I forgot to check.

Ciao,
Dscho

Thanks for double-checking, and I do appreciate the final version, even if it matches the version I would have in 'pu' if I perfectly followed all what was said in the review thread, as I am not always perfect.