Created attachment 190040 [details] Proposed patch This is a much smaller update than the previous one. Patch (also attached): https://people.freebsd.org/~rakuco/exp-runs/qt-5.9.4-v1.patch
New failures on 10.3: http://package22.nyi.freebsd.org/data/103amd64-default-PR225436/2018-01-27_20h23m47s/logs/nomacs-3.6.1.log http://package22.nyi.freebsd.org/data/103amd64-default-PR225436/2018-01-27_20h23m47s/logs/antimicro-2.23_1.log http://package23.nyi.freebsd.org/data/103i386-default-PR225436/2018-01-27_20h23m42s/logs/nomacs-3.6.1.log http://package23.nyi.freebsd.org/data/103i386-default-PR225436/2018-01-27_20h23m42s/logs/antimicro-2.23_1.log Not-new failure on 10.3 that you can give a look at: http://package22.nyi.freebsd.org/data/103amd64-default-PR225436/2018-01-27_20h23m47s/logs/flameshot-0.5.0.log http://package23.nyi.freebsd.org/data/103i386-default-PR225436/2018-01-27_20h23m42s/logs/flameshot-0.5.0.log
I've sent a patch for flameshot here: https://reviews.freebsd.org/D14100 I know what's wrong with nomacs and antiword, I'll fix them tomorrow.
A commit references this bug: Author: rakuco Date: Mon Jan 29 11:40:33 UTC 2018 New revision: 460285 URL: https://svnweb.freebsd.org/changeset/ports/460285 Log: Add a patch to drop CMake calls to QT5_WRAP_CPP(). The USE_QT5 code path already unconditionally sets CMAKE_AUTOMOC to on. Calling QT5_WRAP_CPP() used to be just redundant, as antimicro_HEADERS_MOC was never actually added as a source dependency of the antimicro target. In other words, CMake's own automoc infrastructure was actually being used and the moc invocations from QT5_WRAP_CPP() were not being made at all. Starting with Qt 5.9.4, calling QT5_WRAP_CPP() disables the AUTOMOC property on the macro's input files, which means neither CMake's automoc infrastructure not QT5_WRAP_CPP()'s code were being used and we ended up with several 'undefined reference to vtable' errors when linking. Sent upstream: https://github.com/AntiMicro/antimicro/pull/207 PR: 225436 Changes: head/x11/antimicro/files/patch-CMakeLists.txt
A commit references this bug: Author: rakuco Date: Mon Jan 29 11:45:43 UTC 2018 New revision: 460286 URL: https://svnweb.freebsd.org/changeset/ports/460286 Log: Add a patch to drop CMake calls to QT5_WRAP_CPP(). cmake/Utils.cmake already sets CMAKE_AUTOMOC to on. Calling QT5_WRAP_CPP() used to be just redundant, as QUAZIP_MOC_SRC and LIBQPSD_MOC_SRC were never actually added as source dependencies in MacBuildTarget.cmake and UnixBuildTarget.cmake. In other words, CMake's own automoc infrastructure was actually being used and the moc invocations from QT5_WRAP_CPP() were not being made at all. Starting with Qt 5.9.4, calling QT5_WRAP_CPP() disables the AUTOMOC property on the macro's input files, which means neither CMake's automoc infrastructure not QT5_WRAP_CPP()'s code were being used and we ended up with several 'undefined reference to vtable' errors when linking. Sent upstream: https://github.com/nomacs/nomacs/pull/187 PR: 225436 Changes: head/graphics/nomacs/files/patch-cmake_Unix.cmake
A commit references this bug: Author: rakuco Date: Mon Jan 29 12:37:06 UTC 2018 New revision: 460296 URL: https://svnweb.freebsd.org/changeset/ports/460296 Log: Update Qt5 to 5.9.4. Announcement: https://blog.qt.io/blog/2018/01/23/qt-5-9-4-released/ This is a minor update and a lot easier to land than the previous 5.7.1 -> 5.9.3 commit. Thanks to antoine for the exp-run. PR: 225436 Changes: head/Mk/bsd.qt.mk head/devel/qt5/distinfo head/devel/qt5-core/pkg-plist head/devel/qt5-qmake/pkg-plist head/graphics/qt5-3d/pkg-plist head/misc/qt5-doc/Makefile head/misc/qt5-doc/distinfo head/misc/qt5-doc/pkg-plist head/misc/qt5-examples/distinfo head/misc/qt5-l10n/pkg-plist head/x11-toolkits/qt5-quickcontrols2/pkg-plist