[Bug-tar] --dereference behaviour differs for files and directories

[Vladimir A . Pavlov]

Hi!

In tar info page the following is written:

"
When reading from an archive, the `--dereference' (`-h') option
causes `tar' to follow an already-existing symbolic link when `tar'
writes or reads a file named in the archive.
"

I wrote a simple script and it seems that symlinks to files are not
followed (whereas symlinks to directories are followed).

In other words, if I extract a file using -h and a symlink with the same
name exists then the symlinks is just removed and a new file is created.
But if I extract a directory using -h and a symlink with the same name
exists, the symlinks is followed. So, the behaviour is different for files
and directories.

I'm writing some script and I wouldn't like the following versions of tar to
break it, so I have questions:
1. Is this behaviour a bug, or does it works as expected?
2. Is there a possibility this behaviour will be changed in future?

The following output demonstrates the issue.

In the output, the header "file (-h)" (and similar) means we're checking
how a _file_ overwrites a symlink to a _file_ if being extracted
with _-h_ option. Two groups of lines for each header are filesystem
status before and after extracting so we can see the change.

Tar 1.26 is used.

/tmp$ ./tar.sh

file ()

drwxr-xr-x 3 pkg users 4096 Nov  1 14:54 dst
drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst/dir
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/dir/regfile
lrwxrwxrwx 1 pkg users    7 Nov  1 14:54 dst/dir/symfile -> regfile

drwxr-xr-x 3 pkg users 4096 Nov  1 14:54 dst
drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst/dir
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/dir/regfile
-rw-r--r-- 1 pkg users    5 Nov  1 14:54 dst/dir/symfile

dir ()

drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst
lrwxrwxrwx 1 pkg users    1 Nov  1 14:54 dst/dir -> .
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/symfile

drwxr-xr-x 3 pkg users 4096 Nov  1 14:54 dst
drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst/dir
-rw-r--r-- 1 pkg users    5 Nov  1 14:54 dst/dir/symfile
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/symfile

file (-h)

drwxr-xr-x 3 pkg users 4096 Nov  1 14:54 dst
drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst/dir
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/dir/regfile
lrwxrwxrwx 1 pkg users    7 Nov  1 14:54 dst/dir/symfile -> regfile

drwxr-xr-x 3 pkg users 4096 Nov  1 14:54 dst
drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst/dir
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/dir/regfile
-rw-r--r-- 1 pkg users    5 Nov  1 14:54 dst/dir/symfile

dir (-h)

drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst
lrwxrwxrwx 1 pkg users    1 Nov  1 14:54 dst/dir -> .
-rw-r--r-- 1 pkg users    0 Nov  1 14:54 dst/symfile

drwxr-xr-x 2 pkg users 4096 Nov  1 14:54 dst
lrwxrwxrwx 1 pkg users    1 Nov  1 14:54 dst/dir -> .
-rw-r--r-- 1 pkg users    5 Nov  1 14:54 dst/symfile

The script in question follows:

/tmp$ cat tar.sh 
#!/bin/sh

for opt in '' '-h'; do

echo
echo "file ($opt)"
echo

rm -rf src dst

mkdir -p src/dir
echo file >src/dir/symfile

mkdir -p dst/dir
touch dst/dir/regfile
ln -s regfile dst/dir/symfile

find dst |xargs ls -ld
( cd src; tar -cf - . ) |tar -x $opt -f - -C dst
echo
find dst |xargs ls -ld

echo
echo "dir ($opt)"
echo

rm -rf src dst

mkdir -p src/dir
echo file >src/dir/symfile

mkdir -p dst
touch dst/symfile
ln -s . dst/dir

find dst |xargs ls -ld
( cd src; tar -cf - . ) |tar -x $opt -f - -C dst
echo
find dst |xargs ls -ld

done

--
pv4