Summary: functions in std.file don't take symbolic links into
           Product: D
           Version: 2.040
          Platform: Other
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Phobos

--- Comment #0 from Jonathan M Davis <> 2010-02-24 
00:24:31 PST ---
Created an attachment (id=572)
Changes for std.file.d to make it work with symlinks.

At present, if you call isfile() or isdir() on a symbolic link, it reports on
what the link points to rather than the link itself. Presumably this is fine on
Windows since as far as I know, Windows doesn't have symbolic links. But it is
a big problem for Linux.

The other problem is rmdirRecursive(). At present, I believe that it currently
would follow symbolic links to directories rather than just deleting the
symlink itself. Ideally, it would not follow symlinks.

I have included a patch which I believe fixes the problems. Whether it's the
best solution to the problem, I don't know. I added getLinkAttributes() which
uses lstat instead of stat and make isdir() and isfile() use
getLinkAttributes() rather than getAttributes(). I also added issymlink() to go
with isdir() and isfile(). DirEntry was changed to use lstat as well. That
should fix, isfile(), isdir(), and rmdirRecursive(). Most of the other
functions like lastModified() continue to use fstat, so they continue to give
information on the file which is pointed to rather than the symlink, which I
think is what we want.

The one wart is that having DirEntry use lstat means that all of the attributes
that it gives are for the symlink itself which is not always what we want. So,
I added a followLink() function which returns a DirEntry with the values for
stat if it's a symbolic link or just returns a copy of the DirEntry if it's
not. followLink() can almost certainly been done in a cleaner manner. I'd have
used init, but I don't know how to get dirent for a single file rather than a
whole directory. So, it uses stat and then checks the mode to figure out which
value to set d_type for. DT_WHT is currently not used (it's commented out)
because I have no clue what it represents, so I don't know what mode value from
stat it corresponds to.

In any case, if my patch isn't quite what you'd want to use, hopefully it's at
least a step in the right direction. Hopefully the patch is in an appropriate
format. It has changes for std.file.d

Configure issuemail:
------- You are receiving this mail because: -------

Reply via email to