rpmbuild FAIL while packaging symlinks

Michael Schwendt mschwendt at gmail.com
Sun Aug 19 21:23:15 UTC 2007


On 19/08/07, Andy Green <andy at warmcat.com> wrote:
> Somebody in the thread at some point said:
> > On 19/08/07, Andy Green <andy at warmcat.com> wrote:
> >> There are two "problems" here though, getting his build working at all
> >> and tracking down the segfault.
> >>
> >> I don't mind agreeing that resolving it with the unowned link thing
> >> taken care of is "more correct".  But so far you didn't get the guy over
> >> his actual problem in front of him.
> >
> > Then don't rush. Give him time to reply with a full listing of the
> > buildroot and rpmbuild output, so the problem can be looked at in more
> > detail in order to analyze it. Taking a look at a verbose listing of
> > the buildroot before creating a %files section is something every
> > packager ought to do.
>
> > Based on just the excerpt from his %install section, one cannot jump
> > to bullet-proof conclusions. It is likely that something with the
>
> Okay, summary: In this case the whole proposition is that the symlink
> makes the problem at rpmbuild time.  Removing the symlink from the
> rpmbuild action and doing it on the target at install time should stop
> the segfault happening if the report is to be believed at all.... but
> you don't seem to think that has any value for the guy experiencing the
> problem.  You think it is "false advice" and the product of some "rush"
> on my part.
>
> Have a nice evening!

No idea what your goal in this thread is.

Okay, summary: Somebody tries to add a symlink to a package, which
looks like a fully valid thing to do (and actually is done in lots of
Fedora packages), but it makes rpmbuild segfault. Although it is
unexpected and strange that rpmbuild would crash like that in this
case, you ask the person to "throw out the ln command altogether" as
if it were impossible to run "ln" at that place of the spec file and
instead suggest doing something that doesn't include the symlinks in
the package. You don't even ask whether that means that the person
possibly would return to a previous version of the spec file (one that
only had ldconfig post/postun scripts). Even if dropping the shell
code would make the package build, it doesn't answer the person's
question at the end of the original post: "I'm not sure, what's wrong
here, can't the file and the symlink to it be in the same directory
for rpmbuild packaing?"  I believe that finding out whether the
packager has made a mistake or whether there is a problem in rpmbuild
is the way to go and offers more value.




More information about the users mailing list