Changes for pari 2.5.0

[Ryan Schmidt]

On Jul 1, 2011, at 19:25, Vincent Lefevre wrote:
> On 2011-07-01 04:18:56 -0500, Ryan Schmidt wrote:
>> We had decided years ago that we wanted standard variants to have
>> standard descriptions, hence the global variant descriptions file.
>> Do we really want to override the description of the x11 variant
>> here (or anywhere)? I assumed we would not.
> 
> There's a good reason to override it: give information specific to
> the port (e.g. what additional features x11 brings for the port).

Well that would be nice if ports did that. On the other hand, before we had global variant descriptions, a dozen port authors would describe the same feature a dozen different ways, which can be confusing for users as they try to work out whether the difference in wording indicates a difference in meaning or not. I took a look at the x11 variant situation, since that is the most recent variant to have a global description added; here are a couple of the descriptions currently in use for this variant:

Enable support for X11 (libsdl)
Enable rendering in X11 (reinteract)
Enable X support (libungif)
Enable X11 support (libwmf)
Enable use of X11 (octave-devel)
Build support for X11 (VLC)
Include X11 interface (links)
Enable X11 use in libraries, and build X11-related applications (heimdal)

Especially egregious are the four vtk ports, which describe the x11 variant in four different ways:

build with X11 (vtk-devel)
Build VTK with the MacPorts X11 libraries (vtk)
build VTK with MacPorts X11 (vtk44)
Use X11 (vtk5)

Since all of the above appear to be describing the same thing, I do think it would be to our users' advantage if these ports used the same wording for these variants, which can now be accomplished by simply deleting the custom variant description and letting the default x11 variant description take effect. That's why the warning is there.

We might consider adding additional global variant descriptions, for example for python or gcc version selection. These variants also notoriously have widely varying descriptions for exactly the same feature.

There may be legitimate cases for overriding the default variant description, to provide more info (like pari and perhaps glew). Not sure what we should do. Either leave things the way they are (and let the port author get a lint warning about it), remove the warning for overriding default variant descriptions (and thereby remove the automated way by which the authors of the above ports would become aware that they should not be writing those descriptions themselves anymore, if they have nothing to add), or provide a way in the portfile for the author to indicate that yes they really do mean to override the default description (seems like overkill).