subports.conf?

[Ryan Schmidt]

On Jan 1, 2018, at 16:25, Andrew L. Moore wrote:

> Hi,
> I came across the discussion of using subports as dependencies rather than variants in MacPorts Trac #126 discussion (https://trac.macports.org/ticket/126).
> 
> Has splitting some ports into broad categories of subports, e.g., boost_gcc and boost_clang, been considered or even implemented?  I’m thinking that a ${prefix}/etc/macports/subports.conf that takes precedence over variants.conf  could be helpful.
> 
> My particular situation is that math/julia builds against gcc7, so that in variants.conf appears:
> 
>    +gcc7 +openmpi +openblas
> 
> But +gcc7 gets picked up when building devel/boost, and eventually source-highlight, which depends on boost, fails to build because its Portfile doesn't have a +gcc variant.
> 
> A workaround is to manually build source-highlight with gcc7, add a line to work/.macports.source-highlight.state:
> 
>    target: org.macports.build
> 
> then finally, run port install.  What I’d like to be able to do is add to a  ${prefix}/etc/macports/subports.conf something like:
> 
>    boost_clang
> 
> which would override the variant +gcc7 in variants.conf.
> 
> This is, admittedly, a pretty narrow use case.  Just thought I’d throw it out...

I'm having a really difficult time following this email, what problems you're experiencing and why, and how you're proposing to solve them.

It is a bug that the boost port's gcc variants came back. They should be removed because they cause problems. See https://trac.macports.org/ticket/55604

The fact that a port might be a subport is an implementation detail of relevance to developers only. To users, a port is a port is a port. Ports are things the user installs explicitly, or indirectly as a dependency of another port. They're not things that are listed in a conf file.

If a variant is specified in variants.conf, and a port the user installs does not have that variant, that variant is skipped and the port is installed without it; no error occurs.