Re: dpkg 18.104.22.168.1 bug
> Anders Hammarquist (firstname.lastname@example.org) wrote on 21 May 1998 18:24:
> >> Interesting, but why doesn't this happen on Intel machines??? I've
> >> just installed 2.1.102 on a dual PII and installed manpages without a
> >> glitch...
> >On intel, lchown has the same syscall number as the "old" chown (so
> >there the new syscall is the new behavoir of chown). On alpha, this
> >path was not chosen (possibly having to do with DU compatibility) so
> >binaries (such as tar) which expect chown to act as lchown fail.
> Ugh :-( :-(
> >Yes, it's a mess. There probably needs to be a more permanent fix for
> Obviously. In fact, linux on the alpha is a mess... There's this
> problem, there's no distribution of 2.0.x with the alpha patches
> integrated (that's why I'm using 2.1.x)...
> Anyway, THIS IS A BUG IN TAR/DPKG/WHATEVER. If the kernel behaves
> differently than applications should take this into account!!! I'll
> file a bug report for this.
Ok. I have some thoghts on a proper fix which you may want to include
in the bug report. The problem is in two parts (or may be anyway, I
haven't actually looked at libc to verify this). First, applications
(such as tar) which expect lchown behaviour should call lchown and not
chown (tar call chown as it is now). Second, libc needs to figure out
if the particular kernel has a lchown syscall or not and divert calls
to lchown to the appropriate syscall.
> BTW, submitting a bug report for the alpha is the same as for x86?
AFAIK, yes. Just make sure to state that the bug is in the alpha
version of the package.
To UNSUBSCRIBE, email to email@example.com
with a subject of "unsubscribe". Trouble? Contact firstname.lastname@example.org