Date: Wed, 30 Sep 2015 19:16:59 -0400 From: Lowell Gilbert <freebsd-ports-local@be-well.ilk.org> To: freebsd-ports@freebsd.org Subject: Re: Ports requires pkg 1.6.0, but 1.5.6 is the latest available Message-ID: <448u7nfqfo.fsf@lowell-desk.lan> In-Reply-To: <44eghffsoq.fsf@totally-fudged-out-message-id> (Roger Marquis's message of "Wed, 30 Sep 2015 15:11:52 -0700") References: <5609D023.70402@bluerosetech.com> <CAN6yY1tzo6QobOKqB53yKahC42rS3OnLMGKVTVnDOL%2BU3QfLhw@mail.gmail.com> <560A47FE.6010507@bluerosetech.com> <560AAD43.5000207@unfs.us> <560AF5CF.2080909@bluerosetech.com> <20150929213632.GA23442@ivaldir.etoilebsd.net> <44bncjsn5d.fsf@be-well.ilk.org> <44eghffsoq.fsf@totally-fudged-out-message-id>
next in thread | previous in thread | raw e-mail | index | archive | help
"Roger Marquis" <marquis@roble.com> writes: > For the vast majority of ports and packages it won't make any difference > whether they are installed are from head, quarterly and/or archives (in my > experience, options dependencies aside). Isn't the issue here a dependency on > the version of 'pkg' being enforced by 'pkg'? If so shouldn't this be fixed > in 'pkg'? Sort of. The issue is that trying to build some ports on head requires the version of pkg that's in head; the version that's on the quarterly branch won't do. pkg(8) is more sensitive to this than other ports, because it is used for installing into the production filesystem, but it could happen with many ports. The scenario is that a port on head depends on a recent version of another port, and that other port is still on an older version on the quarterly branch. Most ports will work, most of the time, but if you install some things from head and others from quarterly, there *will* be cases occasionally where the dependencies of a port on head cannot be met by the ports on quarterly. As far as I can see, if someone wants to use the quarterly branch with pkg, they will have fewer problems if they use the quarterly branch for building from source as well. That means checking out the quarterly branch from Subversion, which the Handbook doesn't currently explain how to do. [The portsnap ecosystem, as far I've been able to determine, does not support branches.] In short, there is nothing broken in pkg(8) per se. I am pretty sure this problem does not come up if you do everything from quarterly or everything from head; it's strictly an issue of conflicts between the two. Furthermore, I suspect that if the original poster had updated his quarterly-branch version of pkg to the head version, he probably would have been able to build the other port from the head tree. There are a few places where I may have gone wildly off the track. The most likely one is that real conflicts between quarterly and head may be vastly less likely than my estimate; in that case, the measures described by bapt@ would be the optimum achievable strategy.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?448u7nfqfo.fsf>