Hi,

thank you for your bug report.

On Sat, Feb 21, 2009 at 08:35:26PM +0000, Mark Sheppard wrote:
> If you run "deborphan --guess-section" it lists "libpam-unix2".  If
> you then remove that package Bad Things happen and no-one can log in,
> as I just discovered on a remote server!  So please can deborphan be
> made to not list that particular package?

I'm sorry to hear that but I don't think this will be changed.

deborphan(1):
| This method is in no way perfect  or even reliable, so beware when
| using this!

This part of the man page is very clear.

People who use --guess-* should be aware that it might find packages
which are important for them, this is why it is called "guess" and not
activated by default.  Especially --guess-sections is correct when it
shows you a possible library packages from a different section since
that is what you asked for when you used --guess-sections.

If I would exclude libpam* from --guess-section and add an option
--guess-pam then --guess-all (which I expect to be used more often than
--guess-section) would still show libpam*. Even when I would not add
--guess-pam, which would be wrong, at least --all-packages would still
show libpam*.

I could also exclude libpam* from being displayed at all, but then
people would complain that deborphan --weired-option does not list old,
unused pam modules anymore.  Even when I would implement such
a hardcoded keep list, where should I draw the line?  On my router
"deborphan -a -n" detects amongst other packages vlan, ifupdown,
dhcp-client, openssh-server, wireless-tools and udev (it would also
report the kernel if it wouldn't be in the firmware instead of the file
system).  Should all these packages be part of this hardcoded keep list
since removing any of them can lead to "bad things"?

To summarize this:  You got what you asked for and I don't think this is
a bug nor do I think that there is a sane way to prevent displaying
possible important packages (at least unless you remove all --guess
options).

Feel free to suggest a sane solution to fix this issue. :)

I currently don't think this is a bug at all, thus I will close this
report when I do not receive an answer to this mail.


Regards
Carsten



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Reply via email to