On Wed, Oct 28, 2009 at 12:45:15PM +0100, Toni Mueller wrote:
> Hi,
>
> On Wed, 28.10.2009 at 10:59:34 +0000, Stuart Henderson <s...@spacehopper.org>
> wrote:
> > if you're submitting new/updated ports, please run -current and keep
> > it fairly up-to-date so that you can check things work with what's in
> > the rest of the tree.
>
> I generally understand this argument, but there are some (local)
> problems with it:
>
> * I can't run a full -current system, at least not until OpenBSD will
> run w/o problems on a virtual machine (XEN or KVM, in my case) / not
> create gotchas that I can only collect flames with.
>
> * I usually create ports to complement what I have in a current -stable
> system, 4.6-stable in this case.
>
> * For ports like p5-common-sense, the difference to -current is most
> likely negligible, except if someone changes eg. the implementation
> of SHA256 checksums again (these changed from 4.5 to 4.6, as I
> discovered when I wanted to check out William's patch for nginx
> earlier today). I'm fully aware that this difference is much greater,
> and likely really significant with respect to packages that actually
> have dependencies and/or binary components, and I can't really
> imagine sending in a port for such w/o appropriate warning. I started
> to have a full -current ports tree in addition to my -stable ports
> tree, however, although I'm also aware that this has it's own brand
> of problems.
>
> So... should I stop trying to create and/or update ports?
I don't speak for anyone, but I can tell you that -current is very
stable and a pleasure to work with. I don't update during hackathons,
but I've never had a problem[1] otherwise.
Joachim
[1] But I do know how to build ports from source, and have my system set
up to make this easy. I usually use packages, but occasionally an
important security update to a port and a bump of a library major
version number may interfere to make the most recent packages
uninstallable on the most recent snapshot. This leads to pkg_add
complaining about "bad major" etc.
