Re: "rm -rf" [4.0p] doesn't

[Don Buchholz]

Bob,

Thanks for the answer.  I'd check the Suns if I could, but
that was 1 job (and about 12 months) ago.  And, when performing
wholesale deletions, the odds are fairly good that I was
running as root (now that is a scary thought :-) ).  It's subtle,
but the 555 directories fouling things up seems reasonable.

Sorry about the false alarm.

Thanks again,

- Don Buchholz



Bob Proulx wrote:

> Don
>
>
> > My understanding of "-f" might be incorrect ... I thought "-f"
> > would force rm(1) into removing a file, even if it was mode 444.
> >
> > However, the version shipped w/ RedHat 6.2 (fileutils v4.0p)
> > doesn't work as expected.  For example, if I extracted the
> > attached (gzip'd) tar file, I would expect:
> >
> >         # tar xvfz breaks_rm-rf.tar.gz
> >         # /bin/rm -rf sis900
>
>
> Thanks for the report.  But that is not a bug.  The directories which
> contain the files are not writable and therefore the rm command will
> not remove the files.  The -f option will remove files mode 444 but
> the directory which contains those files must be writable.  Later when
> rm tries to remove the non-writable parent directory it finds it
> non-empty.
>
> Try these next commands on your example.  I used a find command
> instead of chmod -R so that only the directories would be changed
> writable to show that rm -f does remove files mode 444.
>
>   find sis900 -type d -print | xargs chmod u+w
>   rm -rf sis900
>
>
> > to simply, and quietly, remove the entire tree.  It doesn't.
> > (Maybe I spent too many years with SunOS and Solaris?)
>
>
> Check those Sun boxes again.  I don't have access to a Sun machine
> today to verify this but I believe it works the same there.  Certainly
> HP-UX and IBM AIX work this way as well.
>
> Are you sure you were not previously running the command as root?  For
> root, all directories as well as files are writable.  This case can
> only occur for a non-root user.
>
> Bob