On Tue, 23 Dec 2003, Michal Pasternak wrote:

> Frederick Bruckman [Mon, Dec 22, 2003 at 05:55:31PM -0600]:
> > For xscreensaver, buildlink apparently fails in its purpose again: It
> > fails to exclude pam on FreeBSD, because pam is in the base system (PR
> > pkg/23843). Would it be OK to set --without-pam unconditionally? I.e.,
> > are there platforms/systems for which xscreensaver must link against
> > pam to function properly?
>
> I would just disable it for FreeBSD (just like the FreeBSD Ports Collection
> does it!).
>
> If there are problems on other platforms, people can send-pr(1).
>
> This is the safest thing possible -- I bet that if you disable PAM
> unconditionally for all other operating systems, someone after some time
> will e-mail tech-pkg@, asking why the support was dropped.

The problem with that, is that it's then a hidden dependency. In other
words, if it happens to be detected on the build system, the binary
packages won't even run on systems that don't have and don't need PAM.

> And, BTW, the fact this doesn't work has nothing to do with PAM on FreeBSD
> platform. xscreensaver's PAM .c file just uses "sigtimedwait", which on
> FreeBSD is defined in /usr/include/signal.h -- but doesn't occur anywhere
> else in /usr/lib, so the linker gets upset about it. It is just that
> --disable-pam option makes this file excluded from the build process.
>
> Again, I'd take the safest possible bet: turn it off only for FreeBSD. At
> least 99.9% of FreeBSD users, who are happy with FreeBSD Ports use such
> settings and nothing bad happens since...

It's makes no difference on NetBSD, since NetBSD doesn't support PAM,
and on FreeBSD, the xscreensaver support is clearly broken. Unless
someone speaks up to say that the PAM support works, and they need it,
I think it would be safer to simply disable it unconditionally.

Frederick