Subject: Re: dependency on gcc3-c++ and gcc3-c too strict
To: Jeremy C. Reed <email@example.com>
From: Todd Vierling <firstname.lastname@example.org>
Date: 02/10/2005 22:23:24
On Thu, 10 Feb 2005, Jeremy C. Reed wrote:
> > But is this something we really want to encourage? The general assumption
> > should be that, at runtime, you have libs at least as new as the build-time
> > dependency.
> I don't know if it should be -- and pkgsrc doesn't force anything like
> that now. (If we did then @pkgdep would become like @blddep but instead of
> being a specific version it would be a >= range.)
Actually, I've proposed exactly that on tech-pkg recently: Have pkg_add
treat @blddep as a >= dependency for dependency installation purposes.
There are times when the soname does not change (such as API additions,
since that's not a major version bump criterion), but you definitely need to
be at least at the build-time version level.
> > That said, this dependency could be switched to bl3 files using
> > BUILDLINK_DEPENDS and BUILDLINK_RECOMMENDED with all the usual connotations
> > of that.
> I was thinking about that and it seems to be a fine way.
Yes. I'm actually somewhat surprised that it's not a standard buildlink3.mk
after all, because that's more flexible anyhow. Other "system-level" bits
in mk/ use the bl3 framework to pull in dependencies....
-- Todd Vierling <email@example.com> <firstname.lastname@example.org>