tk @8.6.3 has no explicit +x11 variant for 10.5.8

[udbraumann]

I am trying to build R @3.2.2 including its +tcltk variant on 10.5.8, but while configuring it tests if tk was installed with +x11 being set. Without detecting a tk @8.6.3_0+x11 unfortunately R cannot be built with +tcltk, which in turn would be necessary to use the Rcmdr (R Commander: a graphical interface for R). As far as I could see, there was a +x11 variant in previous tk versions even for MacOSX versions lower than 10.6, but apparently it has been dropped (I guess since +quartz was removed, so an explicit +x11 may have appeared superfluous).

Here is what is returned on 10.5.8 about tk's variants:

$ sudo port variants tk
tk has the variants:
   universal: Build for multiple architectures

For comparison, here is what is available on 10.6.8:

$ sudo port variants tk
tk has the variants:
   quartz: Support for native Mac OS X graphics
     * conflicts with x11
   universal: Build for multiple architectures
[+]x11: Enable X11 support
     * conflicts with quartz

For compatibility reasons I would like to ask if the +x11 variant could be always signaled even for platforms below 10.6? In the Portfile I see that for 10.6 and upwards there is a mechanism to always select +x11 if +quartz has not been set. Something similar should be hard-wired for 10.5 and lower.

I have no idea if other ports besides R are relying on an active +x11 switch for tk, but at the moment I think the compatibility issue should be solved from the tk side, not the R side. Thank you!

[ryandesign (Ryan Carsten Schmidt)]

Makes sense. Thanks for the suggestion!

[mf2k (Frank Schima)]

Note that a "request" ticket type is only for requesting a new port.

[dstrubbe (David Strubbe)]

Yes there are various other ports requring tk +x11 which would have the same problem:

math/R/Portfile:161:    require_active_variants path:lib/pkgconfig/tk.pc:tk x11
science/magic/Portfile:35:require_active_variants tk x11
science/magic-devel/Portfile:37:require_active_variants tk x11
science/pymol/Portfile:56:require_active_variants tk "" quartz
science/xcrysden/Portfile:36:require_active_variants tk x11
x11/tix/Portfile:52:require_active_variants tk x11
x11/xcircuit/Portfile:44:require_active_variants tk x11 quartz

Only the way pymol writes the requirement would work on 10.5.

I think this patch should do the job.

I am mystified how the variant descriptions are getting set in this port -- are there some automatic default descriptions that get applied?

[dstrubbe (David Strubbe)]

Cc Me!

[ryandesign (Ryan Carsten Schmidt)]

Yes, there are default variant descriptions for some common variant names. See browser:trunk/dports/_resources/port1.0/variant_descriptions.conf​ .

I was already working on this, and have committed my version of the fix, along with an update to tk 8.6.4, in r141688.