gtk2 auto-fails (preventing automated tasks)

[mww@…]

gtk2 auto-fails if no variant is selected; this prevents automated package building and also creation of distfile mirrors.

I don't know how stable the quartz-backend is, but if it is not, x11 is probably the best default choice. If both choices make sense, it would be better to split gtk2 up into two ports, so a user can have both installed simultaneously. (also I don't know if all ports that require gtk2 do build happily with gtk2-quartz?)

[rhwood@…]

Here are the problems I encountered:

1. Both choices make sense, but are mutually exclusive.
2. Creating two separate ports would force us to fork every single port that depends (either directly or indirectly) on port:gtk2 into portfoo and portfoo-x11 ports.
3. Our variants engine is insufficiently subtle and insufficiently robust to handle automatically testing for the possibility of the correct dependent. This has been hashed over before, but without resolution (I think the macports code is too opaque to encourage new contributors and too complex to easily fix this problem).
4. The (for the port maintainer) obviously preferable default case (use quartz for the backend, not X11) would cause gtk2 to crap all over every single existing installation at upgrade (that had not been built with +quartz), and I'm not willing to be that mean to everyone. Better to be upfront about the issue than require that everyone go back and rebuild everything.

[rhwood@…]

OBE (overcome by events)

[(none)]

Milestone Port Bugs deleted