Hello,

I have found a bug in "cp -rfs".

Steps to reproduce:

1. Given "path1" and "path2" are on different devices.

2. $ touch "path1/file"

3. $ cd path2/; ln -s path1/file

4. $ cp --symbolic-link --force --recursive path1/file .

Expected:

The link is overwritten with an exact copy.

Actual result:

cp shows an error:

cp: 'path1/file' and './file' are the same file

This bug was introduced in

Specifically this hunk:

diff --git a/src/copy.c b/src/copy.c
index e3832c2..9dbd536 100644
--- a/src/copy.c
+++ b/src/copy.c

@@ -46,6 +46,7 @@

 #include "file-set.h"

 #include "filemode.h"

 #include "filenamecat.h"

+#include "force-link.h"

 #include "full-write.h"

 #include "hash.h"

 #include "hash-triple.h"

@@ -1623,11 +1624,13 @@ same_file_ok (char const *src_name, struct stat const *src_sb,

}

}

-  /* It's ok to remove a destination symlink.  But that works only when we

-     unlink before opening the destination and when the source and destination

-     files are on the same partition.  */

-  if (x->unlink_dest_before_opening

-      && S_ISLNK (dst_sb_link->st_mode))

+  /* It's ok to remove a destination symlink.  But that works only

+     when creating symbolic links, or when the source and destination

+     are on the same file system and when creating hard links or when

+     unlinking before opening the destination.  */

+  if (x->symbolic_link

+      || ((x->hard_link || x->unlink_dest_before_opening)

+          && S_ISLNK (dst_sb_link->st_mode)))

     return dst_sb_link->st_dev == src_sb_link->st_dev;

   if (x->dereference == DEREF_NEVER)

Two patches that fix the issue are attached.

They are against the current master in https://github.com/coreutils/coreutils

The changes are also here:

Thank you, Illia Bobyr