In preparation for updating Boost to 1.60, add include guards from more Boost
headers to the list of macros that moc automatically defines when processing
files. As explained in r408911, Qt4's moc cannot parse some constructs used by
a few Boost headers, so we define their include guards to make moc skip them.
This is a cleaner approach that allows us to largely revert r408472, r408473,
r408474, r408475, r408502, r408773 and r408419, which added several patches to
many ports to work around this moc bug.
PR: 199601
PR: 208322
* PORTNAME devel/py-pyicu should be used rather than devel/py-icu, per policy [1]
* Add LICENSE=MIT as per LICENSE file contained in distfile
* Tidy up MAKE_ENV variable assignment (pet portlint)
* Update WWW in pkg-descr to point to PyPI site, since the old site is down
* Update other ports which depend on devel/py-icu to point to devel/py-pyicu
[1] https://wiki.freebsd.org/Python/PortsPolicy
PR: 207367
Approved by: maintainer timeout
Announcement: https://blog.kitware.com/cmake-3-5-0-available-for-download/
Most ports have already been fixed for this update, but science/orthanc was
included in this commit because I am not entirely sure the changes in it work
with CMake < 3.5.0. In short, FindDCMTK.cmake was rewritten and the way orthanc
uses the DCMTK variables was incompatible with the use upstream intended.
PR: 208033
Install x11/kdelibs4's headers into include/kde4 instead of include (which
consequently causes several other ports to have their installation paths
changed too).
The idea behind this is to reduce path conflicts between KDE4 ports and the
upcoming KDE Frameworks 5 ports that will be installed into include/KF5. If
we continue installing the KDE4 headers into include/, we can end up in a
situation like this:
c++ [...] -I/usr/local/include -I/usr/local/include/KF5 file.cpp
If the KDE4 and KF5 versions of a port have the same headers, the KDE4 port
will unintentionally be picked up first and the build will fail.
Most of this huge patch is just PORTREVISION bumps, pkg-plist changes and a
few patches to FooConfig.cmake files to make them look into the kde4/
subdirectory in include/.
Changes which don't fit into the above are:
- deskutils/kdepimlibs4: Import an upstream patch to remove some double
semicolons that cause base GCC to fail. They have always been present, but
since the faulty header was referenced via -isystem /usr/local/include
this never caused any problems.
- devel/subversion, devel/subversion18: Update patch-configure. The current
kwallet changes there date back to 2011 (r272490), at a time when the
build could fail when both KDE3 and KDE4 were installed. Replace those
bits with a change I've submitted upstream to use the kde4-config program
to determine where KDE4's headers and libraries are installed instead of
assuming the headers are always in include/.
Once again, huge thanks to Tobias Berner <tcberner@gmail.com> for being the
first one to notice this problem when working on the KDE Frameworks 5 ports,
coming up with the solution and bugging me until I had time to work on this
and ask for the exp-run :-)
PR: 207906 (exp-run)
This update to the port also adds a lot of dependencies on system libraries
instead of using the versions bundled with ITK.
PR: 206977
Submitted by: Ben Woods <woodsb02@gmail.com> (maintainer)
This port checks the release date, and subtracts it from the current
date. If the difference is more than 6 months, it interactively asks
the user if they want to upgrade.
The post-patch target removes this possibility, ensuring the port never
reverts to an interactive build.
approved by: infrastructure blanket (interactive builds discouraged)
There are numerous graphviz related packages in PyPI, including:
graphviz, graphviz-python and pygraphviz.
This port, py-graphviz, is not and does not provide the same package
as the 'graphviz' package on PyPI (CHEESESHOP).
Further, it blocks the creation of a correctly named py-graphviz port,
is a POLA violation in terms of users expecting to find and recieve
the package they expect based on a standard search, and is likely to
create confusion among other ports if they depend on a python graphviz
package, which could result in incorrect packages/dependencies.
This is why among other reasons, Python packages should be named
exactly what they are called in PyPI, or inside install_requires,
setup_requires or tests_requires as dependencies in other packages.
This change moves py-graphviz to py-pygraphviz to correct the above
incorrect naming.
While I'm here:
- Remove backup (http) MASTER_SITES, CHEESESHOP is highly-available
and provides a Geo-aware CDN by Fastly.
- Remove {BUILD,RUN}_DEPENDS in favour of the actual LIB_DEPENDS
- Use PYDISTUTILS_* framework variables to customise the build instead
of hack patching setup.py
- Enable autoplist, fixing builds on Python 3.x [1]
- Enable concurrent (Python version) installation
- Hack patch setup.py to preclude docs installation, as it adds a
directory to --record and breaks pkg-plist output [2]
- Remove DOCSDIR override (no longer necessary)
- Assign MAINTAINER'ship to python@
- Add TEST_DEPENDS and test target
- pkg-descr: Match WWW URL to setup.py:homepage field
- pkg-descr: Improve package description text (matching upstream)
While I'm sweeping dependencies:
- Fix a typo in science/gramps: s|>-|>=
[2] https://wiki.freebsd.org/Python (Open task for install_data bug)
Reported by: Ben Woods (via IRC) [1]
