brooks at one-eyed-alien.net
Thu May 4 19:23:30 UTC 2006
On Thu, May 04, 2006 at 03:15:12PM -0400, Kris Kennaway wrote:
> On Thu, May 04, 2006 at 11:39:36AM -0700, Brooks Davis wrote:
> > On Thu, May 04, 2006 at 12:57:27PM -0400, Kris Kennaway wrote:
> > > On Thu, May 04, 2006 at 11:41:55AM +0200, Ulrich Spoerlein wrote:
> > > > Aren Olvalde Tyr wrote:
> > > > > Perhaps it would be good, if, say on the corresponding port entry on the
> > > > > FreeBSD ports webpage, it listed all the options used for building the binary
> > > > > package. For example, for the "Package" link, instead of simply linking to
> > > > > the package, it could link to a page entry listing all of the build options
> > > > > used, with the package download link at the bottom. Or something like that.
> > > > >
> > > > > Just an idea. What do people think?
> > > >
> > > > I'd even go further. This is something I have been thinking about on and
> > > > off. Namely, a FLAVOUR system for packages. A maintainer specifies up to
> > > > three FLAVOURs per port, which set various flags for building the port.
> > >
> > > In FreeBSD land these are called "slave ports", and you can have as
> > > many as you like. I don't have any interest in adding a separate
> > > "flavour" system in parallel to this.
> > With MPI based parallel code there are times where I think a flavor
> > system might scale better, but I haven't done the work to expose the
> > non-scaling yet. The problem is that we've got ~5 versions of MPI
> > in the tree, but each one of those really should be buildable with
> > different C and Fortran compilers so you could easily see 50+ MPIs.
> > Multiple each applicaiton by that and things get crazy. :)
> Do all combinations really need packages? With or without flavours
> you wouldn't even think about building packages for all possible
> combinations of build options for a port.
All combinations don't need packages, but I'd like an easy way to build
as many as half a dozen versions on the same machine so users can use
the compiler and MPI version of their choice. At this point the easiest
way to handle that would be to build non-conflicting slave ports for the
combinations I wanted but that starts to waste a lot of inodes pretty
Another option that could work for me would be to make it easier to
maintain a local ports category so I could have my own slave ports.
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20060504/8394e58d/attachment.pgp
More information about the freebsd-ports