While most mkspecs use hardcoded compiler names / options, qmake in
meta-qt5 is setup to get them from the environment. While this works
fine with openembedded, it does break normal workflows outside of it.
For example to get the SDK working with Qt Creator either the environmental
script most be sourced first, making it impossible to switch between
different Qt versions or the OE_QMAKE must be set for every project,
otherwise you get cryptic errors like 'make: c: Command not found',
since the compiler name is not set and hence an attempt is made to
execute only the arguments.
Likewise on the target itself, qmake doesn't work without first setting
the OE_QMAKE variables.
This patch adds default toolchains options, so it is no longer required
to set them, but keeps the flexibility to alter them from the environment.
Signed-off-by: Jeroen Hofstee <jhofstee@victronenergy.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This was never defined for nativesdk, which then broke sdk which
had nativesdk-qtwayland included. Rather than add it there, remove
its use everywhere, since it was not adding anything meaningful to
the build.
Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The expansion is now again optional, but for brief time
it was required. Add the default value (True) so that
the layer can be used also with older oe-core versions.
Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Previous patch was removed, but PACKAGECONFIG was not changed
to match the new feature flag.
Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Different HostPrefix was used when building Qt and in external
SDK's qt.conf. This caused e.g., wrong path in qmodule.pri for
PKG_CONFIG_EXECUTABLE.
Task-number: QTBUG-63097
Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Using Qt based SDK for stripping ARM binaries results in errors.
Generated Makefile contains:
STRIP = \$(OE_QMAKE_STRIP)
but due to the missing export of OE_QMAKE_STRIP variable STRIP will be unset and
no strip will be perfomed:
install -m 755 -p lib/libSigUdev.so.1.0.0 /opt/poky/2.2.1/sysroots/cortexa9hf-neon-poky-linux-gnueabi/usr/lib/libSigUdev.so.1.0.0
strip-unneeded /opt/poky/2.2.1/sysroots/cortexa9hf-neon-poky-linux-gnueabi/usr/lib/libSigUdev.so.1.0.0
make: strip-unneeded: Command not found
Makefile:576: recipe for target 'install_target' failed
make: [install_target] Error 127 (ignored)
adding OE_QMAKE_STRIP=$STRIP to generated qt5.sh solve the issue.
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
ptest.bbclass provides a set of standard do_*_ptest_base tasks. Each of
them calls a corresponding regular do_*_ptest task, that components are
supposed to override for own implementation. When PTEST_ENABLED is not
set, an anonymous python function removes all do_*_ptest_base tasks from
the queue. qt5-ptest.inc adds a special case for do_populate_sysroot and
re-arranges the order or ptest tasks due to that. But, unfortunately,
do_install_ptest was added directly, not do_install_ptest_base, hence when
"ptest" is disabled and all other do_*_ptest_base tasks are removed, this
one do_install_ptest task is left w/o dependencies and gets scheduled for
execution very early on, even before pseudo-native gets built and stages
"fakeroot" functionality. The fix is to add do_install_ptest_base task,
which calls do_install_ptest only when "ptest" is enabled.
Signed-off-by: Denys Dmytriyenko <denys@ti.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
A useless file was added mistakenly by commit ec80b092:
[ qtbase: Add OE clang specific mkspecs ]
Drop it.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
If we specify neither -system-pcre nor -qt-pcre then qtbase will first see
if it can find the pcre library in the system and fall back to compiling
its own if that's not possible.
Unfortunately, this means that if qtbase-native happens to be built on a
machine with system pcre available then the resulting binaries such as
bin/qt5/uic in sstate files won't work correctly on machines that don't
have pcre available. Even ignoring that, it doesn't help for packages to
build differently depending on how the build system is configured.
We could depend on pcre-native and pass -system-pcre, but this doesn't work
by default because qtbase actually requires pcre16 which is an optional
(via PACKAGECONFIG) feature of pcre so the default there would need to be
changed too.
So, let's make qtbase-native match the default of qtbase by passing
-qt-pcre in order to ensure that qtbase-native is always built with its
own pcre implementation.
Signed-off-by: Mike Crowe <mac@mcrowe.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Since update to 5.9.2 native recipes in meta-qt5-extra fail due to strange path
to mkspecs. This patch fixes this the hard way.
Signed-off-by: Andreas Müller <schnitzeltony@googlemail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
We can not piggy back clang anymore on existing
OE mkspecs since starting 5.9 the configure is
asking compiler for include paths and it needs to
know if platform is clang-linux or g++-linux
Fixes:
ERROR: failed to parse default search paths from compiler output
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* remove
qtdeclarative/0001-Fix-QQmlExpression-leaking-QQmlError-objects.patch
qtwayland/0002-Fix-initial-window-property-values-being-propagated.patch
which is already applied and fuzzy patch just apply the same section
twice
* restore
qttools/0001-add-noqtwebkit-configuration.patch
which was removed from SRC_URI in 5.9 upgrade
* add comment with link to meta-qt5 repository where the meta-qt5 .patch
files are maintained, include branch and tag name so that it's more
clear from where these changes are exported (with:
git format-patch --no-numbered --no-signature)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* adapt QtWebEngine recipe to use GN instead of GYP
* add QtRemoteObjects and QtWebView as a new Qt modules
* update available QtBase configure arguments
* remove obsolete patches
* patch all .pc files to remove build paths
* include generated QML cache files in packages
* the patch "configure paths for target qmake properly" could not
be applied anymore and support must be done differently
* QtWebEngine now requires gcc-multilib to be installed on the host
system, because the host tools are built to the same bitness as
the target (arm -> x86, aarch64 -> x86-64)
* refresh the patches to match with b5.9* branches on:
https://github.com/meta-qt5/qtbasehttps://github.com/meta-qt5/qtwebengine
and 56-based branch on
https://github.com/meta-qt5/qtwebengine-chromium
Signed-off-by: Samuli Piippo <samuli.piippo@qt.io>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* meta/conf/distro/include/security_flags.inc adds -Wformat and
-Wformat-security, but openh264 adds -Wno-format after that causing
build to fail
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This file lists the licenses in use by OE-Core and it not a
proper license file, just use the COPYING.MIT.
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
With
commit efa8aaf82e
Author: Andreas Müller <schnitzeltony@googlemail.com>
Date: Thu Feb 9 00:26:09 2017
qmake5_base.bbclass: set qt.conf by environment variable again
we stopped pointing ${OE_QMAKE_QTCONF_PATH} at a valid file and
instead directed it to a path which was intended not to exist.
The motivation was to permit qtbase/qtbase-native/nativesdk-qtbase
to build again after Qt 5.8 started paying attention to the
contents of this file.
The change as done in efa8aaf works well enough for Morty and
subsequent releases' copies of Bitbake, but fails on earlier
releases because they lack the following change:
commit 2afcbfef2cd1ca568e5225884a8021df38ee3db0
Author: Ross Burton <ross.burton@intel.com>
Date: 2016-07-14 13:56:22
bitbake: build: don't use $B as the default cwd for functions
The result is that when we build with Krogoth or prior, the body
of do_generate_qt_config_file() runs with a cwd of ${B}, which was
_not_ the intent of efa8aaf. Because the working directory is ${B},
${OE_QMAKE_QTCONF_PATH} is written in there too. do_configure() --
whose cwd is by design also ${B} -- then finds the file 'foodummy',
and the build breaks for the reasons outlined in efa8aaf.
This change simply shifts the implementation tactics to suppress
the creation of ${OE_QMAKE_QTCONF_PATH} during qtbase rather than rely
on unspecified behavior about the cwd of do_generate_qt_config_file().
(cherry picked from commit a17ff281aa8d99f770b0a049cb798235005fb93e
at https://codereview.qt-project.org/yocto/meta-qt5.)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* upload the patches to b5.8* branches on:
https://github.com/meta-qt5/qtbase
* refresh the patches
* notice that some patches were in section which is meant to be
common for target, native and nativesdk qtbase but were added
only in target recipe, I've updated them to be included in all
3, hopefully they can be used in all 3 (I don't use nativesdk at all)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
This adds a new command line switch to the configure script to make sure
the right paths to moc, uic, rcc are used. It also makes sure OpenGL is
enabled if the qt5 packageconfig is in use.
Note that the moc/uic/rcc path configure switches were introduced
in GStreamer 1.12.2, so this version is a requirement.
Signed-off-by: Carlos Rafael Giani <dv@pseudoterminal.org>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
qtbase-native has been included by QT5TOOLSDEPENDS, so this line is
redundant and leading multiple qtbase-native existing in DEPENDS.
Signed-off-by: Ming Liu <peter.x.liu@external.atlascopco.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The license checksum has changed as a static linking exception has been
added on top of the LGPL 2.1 license.
Signed-off-by: Jonathan Liu <net147@gmail.com>
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
The expand parameter to DataSmart.getVar() defaults to True since
yocto 2.3 (pyro) [0]. Stable Yocto releases 2.2 (morty) and 2.1
(krogoth) had no default value. Older, now unsupported releases
defaulted to False [1] [2].
Not passing the expand parameter here means all recipes including
qt5-ptest.inc won't build with yocto 2.2 and 2.1 with this error
message:
$ bitbake qtxmlpatterns -vfc populate_sysroot
[...]
ERROR: Error executing a python function in exec_python_func() autogenerated:
The stack trace of python calls that resulted in this exception/failure was:
File: 'exec_python_func() autogenerated', lineno: 2, function: <module>
0001:
*** 0002:do_populate_sysroot(d)
0003:
File: '[...]/poky/meta/classes/staging.bbclass', lineno: 227, function: do_populate_sysroot
0223: p = p.replace("/", "_")
0224: with open(provdir + p, "w") as f:
0225: f.write(pn)
0226:}
*** 0227:
0228:do_populate_sysroot[vardeps] += "${SYSROOT_PREPROCESS_FUNCS}"
0229:do_populate_sysroot[vardepsexclude] += "MULTI_PROVIDER_WHITELIST"
0230:
0231:SSTATETASKS += "do_populate_sysroot"
Exception: TypeError: getVar() missing 1 required positional argument: 'expand'
DEBUG: Python function do_populate_sysroot finished
ERROR: Function failed: do_populate_sysroot
Fixing this in a user layer requires duplicating the entire
do_populate_sysroot_append() in the .bbappend of all packages
including qt5-ptest.inc, i.e. currently qtdeclarative, qtquick1 and
qtxmlpatterns.
Avoid the hassle for those users by explicitly passing expand=True.
[0] http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=2631c375b8a025ee2e4329f6d93900bb7b5dadf7
[1] http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=c7610aa353e31977c099ce574acb188a3f4219e0
[2] http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=4e5e501155a21fe64a470c626b3afe6d9748308e
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
* it was partially fixed in previous change, but there was still:
qwt-qt5-6.1.3: qwt-qt5-dbg: found library in wrong location:
/usr/plugins/designer/.debug/libqwt_designer_plugin.so [libdir]
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
