[MacPorts] #51643: qt5 portgroup overwrites supported_archs, defines universal variant when one is not desired

[MacPorts]

#51643: qt5 portgroup overwrites supported_archs, defines universal variant when
one is not desired
---------------------------+------------------------
  Reporter:  ryandesign@…  |      Owner:  mcalhoun@…
      Type:  defect        |     Status:  assigned
  Priority:  Normal        |  Milestone:
 Component:  ports         |    Version:  2.3.4
Resolution:                |   Keywords:
      Port:  qt5           |
---------------------------+------------------------

Comment (by rjvbertin@…):

 Replying to [comment:5 ryandesign@…]:

 > The problem is when the qt5 portgroup includes the muniversal portgroup
 and thus defines a universal variant for a port like qupzilla 2 which only
 builds for a single architecture -- x86_64 -- and thus should not have a
 universal variant.
 > And therefore how can the qt5 portgroup know whether or not it should
 include the muniversal portgroup? I don't know.

 I guess there's only one way to come up with an answer to that: determine
 if Qt *applications* require the muniversal PG for them to be built. A
 quick test with qmake and cmake shows that once you manage to get `-arch
 i386 -arch x8_64` into the C(XX)FLAGS straightforward universal binary
 builds work. That is, I get the expected universal binary .o files, and
 errors from the linker when it finds out that my Qt frameworks aren't
 universal.

 If this is confirmed for more than the very simple applications I tested
 the `muniversal` inclusion can go into the Qt Portfile itself.

 Even if it isn't confirmed, is there a way for the Qt PortGroup to oblige
 Portfiles to include `muniversal` if they want to support binary builds?
 I.e. can the Qt5 PortGroup remove the universal variant? It seems that the
 universal variant is available normally when one sets `universal_variant
 no` in the PortGroup, and then includes `muniversal` in a Portfile that
 wants to provide universal variants.

 If you can confirm that is how things are supposed to work then it seems
 the best solution yet. It would avoid the question how well the cmake
 PortGroup handles +universal, and even less more you'd get qmake to do
 universal builds.

 >
 > check `supported_archs` for the presence of archs other than i386 or
 x86_64 and raise an error in that case?
 >
 > I think that's an orthogonal issue, unrelated to this ticket.

 Probably, knowing now that your use case is x86_64 only.

-- 
Ticket URL: <https://trac.macports.org/ticket/51643#comment:6>
MacPorts <https://www.macports.org/>
Ports system for OS X