Please review my (hopefully) final Qt 5.3.2 commits

[Raphael Kubo da Costa]

Hi everyone (particularly Max and Alberto),

I believe all the work for updating Qt5 to 5.3.2 is done. I've been
committing smaller parts of the required changes for a while now, but
left the big, disruptive changes for the end.

The pending diff was quite big to manually divide into separate commits
for SVN, so I ended up doing them in git for now, and will commit them
in the right order to area51 once someone else looks at and OK's them.

The commits are here:
https://github.com/rakuco/freebsd-ports/commits/qt-5.3.2

Specifically, I'm interested in feedback on 12b281c ("Call qmake from
the root of the ${WRKSRC}") and 2787663 ("bsd.qt.mk: Set QMAKESPEC
instead of QMAKEPATH"). Even after looking at several commits in area51
(such as r8728, r8841 and r10019) it wasn't clear why we were doing so
many weird things with qmake and qtbase's configuration script (like
calling qmake directly in pre-configure and post-configure while passing
-dont-process to the configure script). I believe the explanations to my
changes make sense, but I'd really like you guys to see if I got
everything right since you were doing the Qt5 work before.

In addition to that, I don't think we need to change QMAKE_ARGS with the
contents of QT_CONFIG and QT_DEFINES, but I didn't find a good
explanation for why we were doing that in the first place (r8782), but
at least the Poudriere builds of the qtbase ports without those worked
fine.

Lastly, I'm fairly convinced 7492a2c ("Stop explicitly passing
${LOCALBASE} to the compiler") is the only way to fix ports/194088, but
feedback is also welcome.

[Max Brazhnikov]

Post by Raphael Kubo da Costa
Hi everyone (particularly Max and Alberto),
I believe all the work for updating Qt5 to 5.3.2 is done. I've been
committing smaller parts of the required changes for a while now, but
left the big, disruptive changes for the end.
The pending diff was quite big to manually divide into separate commits
for SVN, so I ended up doing them in git for now, and will commit them
in the right order to area51 once someone else looks at and OK's them.
https://github.com/rakuco/freebsd-ports/commits/qt-5.3.2

Everything looks quite reasonable, so I'm fine with all changes.

Post by Raphael Kubo da Costa
Specifically, I'm interested in feedback on 12b281c ("Call qmake from
the root of the ${WRKSRC}") and 2787663 ("bsd.qt.mk: Set QMAKESPEC
instead of QMAKEPATH"). Even after looking at several commits in area51
(such as r8728, r8841 and r10019) it wasn't clear why we were doing so
many weird things with qmake and qtbase's configuration script (like
calling qmake directly in pre-configure and post-configure while passing
-dont-process to the configure script). I believe the explanations to my
changes make sense, but I'd really like you guys to see if I got
everything right since you were doing the Qt5 work before.

I only updated it from 5.2.0-beta to 5.2.1 (according to the logs). I didn't
dive deep into the building system. r10019 was a temporal solutions, which
actually runs qmake from the root of ${WRKSRC} (otherwise ports fail to
build due to missing files which are generated by calling qmake),
but I didn't find time for proper fixing it.

Post by Raphael Kubo da Costa
In addition to that, I don't think we need to change QMAKE_ARGS with the
contents of QT_CONFIG and QT_DEFINES, but I didn't find a good
explanation for why we were doing that in the first place (r8782), but
at least the Poudriere builds of the qtbase ports without those worked
fine.

I don't know what they are needed for. I'd like to see Alberto's comment
on this matter.

Post by Raphael Kubo da Costa
Lastly, I'm fairly convinced 7492a2c ("Stop explicitly passing
${LOCALBASE} to the compiler") is the only way to fix ports/194088, but
feedback is also welcome.

This may break 3rd-party software and ports which relay on this includes,
but I don't consider this as a drawback: passing additional flags when you
need them is much easier, than removing flags when they bring problems.

Max

[Raphael Kubo da Costa]

Post by Max Brazhnikov

Post by Raphael Kubo da Costa
Lastly, I'm fairly convinced 7492a2c ("Stop explicitly passing
${LOCALBASE} to the compiler") is the only way to fix ports/194088, but
feedback is also welcome.

This may break 3rd-party software and ports which relay on this includes,
but I don't consider this as a drawback: passing additional flags when you
need them is much easier, than removing flags when they bring problems.

The idea of setting LIBRARY_PATH and CPATH in CONFIGURE_ENV and MAKE_ENV
is to make the change transparent to 3rd-party software: the only change
is that while we explicitly passed -I/usr/local/include and
-L/usr/local/lib before via mkspecs/freebsd-*/qmake.conf, they are now
passed implicitly via those environment variables.

I may have to move those definitions from bsd.qt.mk to Uses/qmake.mk
though, as the previous setup only affected ports which used qmake and
the new approach should be limited to qmake as well.