diff -j for different handling of special files?

[Mark Aitchison]

I notice that people have for a while wanted diff to handle special 
files differently; I recently found I really wished diff would not just 
report two links as if they are files with the same contents, e.g.

  Files /usr/share/gnome/help/cheese/sv/figures/cheese-fullscreen.jpg 
and /usr/share/gnome/help/cheese/C/figures/cheese-fullscreen.jpg are 
identical

when one was a link to another, e.g.

  lrwxrwxrwx. 1 root root 34 2009-11-10 08:20 
/usr/share/gnome/help/cheese/sv/figures/cheese-effects.jpg -> 
../../C/figures/cheese-effects.jpg

So I am thinking of writing and submitting a patch for diff to have the 
option (it would maintain present behaviour without the link) of saying 
something like:

  File usr/share/gnome/help/cheese/sv/figures/cheese-fullscreen.jpg is 
a symbolic-link to file 
/usr/share/gnome/help/cheese/C/figures/cheese-fullscreen.jpg

or, in other circumstances:

  File usr/share/gnome/help/cheese/sv/figures/cheese-fullscreen.jpg is 
a regular-file [or device, directory, etc] symbolically linked from 
/usr/share/gnome/help/cheese/C/figures/cheese-fullscreen.jpg

or

  Files /usr/share/gnome/help/cheese/sv/figures/cheese-fullscreen.jpg 
and /usr/share/gnome/help/cheese/C/figures/cheese-fullscreen.jpg are 
links to the same file

This would mean a script that looks for duplicate files would not 
accidentally delete the only copy of a regular file (no, I didn't do 
that - but I did think of that happening with something like the was I 
was searching for diff lines reporting files being identical), and it 
may help in patch file generation in some situations, but I can imagine 
that even more enhanced ways or reporting special files in differences 
lists will be desired.  It also keeps the first file the second field in 
the output in each case, and the type of the first file in the first and 
second cases is given after "is a" in the same format that script would 
understand parsing something like:

  File file5/file1 is a directory while file file1 is a regular file

in the present diff output.

Before starting, I'd like to ask if this is a reasonable thing to do. 
I'm thinking of using the -j option (since it is one of the few letters 
unused, and I'd like to keep -z for processing zip/etc archive 
differences).  The changes in documentation would be something like 
(forgive the formatting):

    -j
    --just-files
            Just compare contents of real files - if one or more files 
to be compared is a link or any other "special" file just
            report the types of the files (and if one is a link to the 
other, or both a link to another file, report that).  With
            this option diff will never read either file if one is 
special, and never report the files "are identical".

    --just-files=filetype[filetype2...]
            Restricts the type(s) of files diff will consider, and if 
either is a special file diff will not read the contents but
            just report differences in file type. Parameters may be one 
or more of the following letters:

              b      block (buffered) special

              c      character (unbuffered) special

              d      directory

              h      hard link

              p      named pipe (FIFO)

              f      regular file

              l      symbolic  link

              s      socket

              D      door (Solaris)


I also think it would be nice (but I'm not volunteering to do it at the 
moment) to add options to diff to report any differences in the 
ownership/permissions (and SELinux security content) etc. Perhaps using 
-o.  And I am quite keep to see a -z option to report differences in 
(compressed) archives, reporting differences in files (even if they 
appear in a different order) between two .zip files for example, or any 
combination of .zip, .zoo, .tar, compressed tar, and real directory.

Thoughts?
Mark.