re: kernel32: Implement ReplaceFileA/ReplaceFileW (revised, resubmit)

[Dan Kegel]

Kudos for your persistance.  I hope this makes it in soon.

I see two remaining problems:

+        /*
+         * This case should never occur, we've already checked
permissions earlier
+         * and we are holding the file handle open.
+         */
+        ERR("Replacement file may not be deleted!\n");

You need to handle this case better; it happens in real
life sometimes that the filesystem lies when it tells
you you can write.  It should be a WARN, and you
need to return the proper error code.

Are you quite sure you need to do any file copying?
Wouldn't a series of renames be more efficient?
- Dan