Charles Davis cdavis@mymail.mines.edu writes:
@@ -1828,13 +2054,24 @@ static NTSTATUS find_file_in_dir( char *unix_name, int pos, const WCHAR *name, i if (ret >= 0 && !used_default) { unix_name[pos + ret] = 0;
if (!stat( unix_name, &st ))
/* when checking case on a case insensitive but case preserving* fs, don't even bother doing the stat(), so we don't match the* file case-insensitively.*/if ((case_sensitive || !check_case) && !stat( unix_name, &st )) { if (is_win_dir) *is_win_dir = is_same_file( &windir, &st ); return STATUS_SUCCESS; }
This still doesn't make sense. There's no reason to ever skip the stat.
@@ -2455,6 +2713,10 @@ static NTSTATUS lookup_unix_name( const WCHAR *name, int name_len, char **buffer const WCHAR *end, *next; int is_win_dir = 0;
unix_name[pos] = 0;stat( unix_name, &st );unix_name[pos] = '/';
You can't add extra stat calls to the normal path. This is supposed to be an optimization.
On 10/18/10 8:28 AM, Alexandre Julliard wrote:
Charles Davis cdavis@mymail.mines.edu writes:
@@ -1828,13 +2054,24 @@ static NTSTATUS find_file_in_dir( char *unix_name, int pos, const WCHAR *name, i if (ret >= 0 && !used_default) { unix_name[pos + ret] = 0;
if (!stat( unix_name, &st ))
/* when checking case on a case insensitive but case preserving* fs, don't even bother doing the stat(), so we don't match the* file case-insensitively.*/if ((case_sensitive || !check_case) && !stat( unix_name, &st )) { if (is_win_dir) *is_win_dir = is_same_file( &windir, &st ); return STATUS_SUCCESS; }This still doesn't make sense. There's no reason to ever skip the stat.
When I wrote that, I was thinking, "If we do the stat() on a case-insensitive file system, it will succeed even if the filename has the wrong case. We don't want that when we're checking case, so we may as well read the dirents to see if the file exists with the right case." Is there something I'm missing? Would check_case even matter when case_sensitive is FALSE?
@@ -2455,6 +2713,10 @@ static NTSTATUS lookup_unix_name( const WCHAR *name, int name_len, char **buffer const WCHAR *end, *next; int is_win_dir = 0;
unix_name[pos] = 0;stat( unix_name, &st );unix_name[pos] = '/';You can't add extra stat calls to the normal path. This is supposed to be an optimization.
Yeah, you're right, as usual. I'll fix this in try 4.
Chip
Charles Davis cdavis@mymail.mines.edu writes:
When I wrote that, I was thinking, "If we do the stat() on a case-insensitive file system, it will succeed even if the filename has the wrong case. We don't want that when we're checking case, so we may as well read the dirents to see if the file exists with the right case." Is there something I'm missing? Would check_case even matter when case_sensitive is FALSE?
If the file system is not case sensitive then there's no case to check.
On 10/18/10 10:57 AM, Alexandre Julliard wrote:
Charles Davis cdavis@mymail.mines.edu writes:
When I wrote that, I was thinking, "If we do the stat() on a case-insensitive file system, it will succeed even if the filename has the wrong case. We don't want that when we're checking case, so we may as well read the dirents to see if the file exists with the right case." Is there something I'm missing? Would check_case even matter when case_sensitive is FALSE?
If the file system is not case sensitive then there's no case to check.
Only if the file system doesn't preserve case. Most modern case-insensitive file systems (FAT, NTFS, HFS) preserve case. CIOPFS preserves case if it's allowed to write xattrs to the underlying FS. This is why I even had the case-preserving handling in there in the first place.
So, if I understand correctly, you think that, if we determine that file system is case-insensitive, we should also assume it doesn't preserve case. Is that correct?
Chip
Charles Davis cdavis@mymail.mines.edu writes:
Only if the file system doesn't preserve case. Most modern case-insensitive file systems (FAT, NTFS, HFS) preserve case. CIOPFS preserves case if it's allowed to write xattrs to the underlying FS. This is why I even had the case-preserving handling in there in the first place.
So, if I understand correctly, you think that, if we determine that file system is case-insensitive, we should also assume it doesn't preserve case. Is that correct?
No. The check_case flag is used to specify Unix semantics, but Unix semantics on a case-insensitive filesystem are just the same as Windows. It doesn't make sense to try to be more case sensitive than the filesystem.
-
Alexandre Julliard -
Charles Davis