The console-bridge is a cmake installation, but not a catkin package.
In its current state, the recipe compiles and installs, but the
class_loader package fails and complains with
| catkin_package() CATKIN_DEPENDS on 'console_bridge', which has been found
| in
| '(...)/build/tmp-eglibc/sysroots/qemux86-64/usr/share/console_bridge/cmake/console_bridge-config.cmake',
| but it is not a catkin package
This commit provides the current state for further discussion and
joint investigation.
| CMake Warning at /build/v2012.12/build/tmp-angstrom_v2012_12-eglibc/sysroots/x86_64-linux/usr/share/catkin/cmake/test/nosetests.cmake:90 (message):
| nosetests not found, Python tests can not be run (try installing package
| 'python-nose')
Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
The last patched version assumes that the variable msg include dirs
is only one directory and not a list of directories. This leads to
errors if generate_message(DIRECTORY ...) occurs twice in the
CMakeLists files.
Make Error at build/devel/share/cmake/...-msg-paths.cmake:2 (if):
| if given arguments:
|
| "NOT" "IS_ABSOLUTE" "" ""
Fixed issue #72.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
The python binary path for genmsg_py.py and gensrv_py.py are hard
coded in the CMake file and don't work during cross compile. Setup
them via find_program during runtime.
This fix#69 and the following error:
*** No rule to make target `/usr/lib/genpy/genmsg_py.py', needed by
`devel/lib/python2.7/site-packages/std_msgs/msg/_Bool.py'. Stop.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
When importing the rosbag module in python, it fails in a genpy python module
complaining about missing the shutil python module. The shutils module is
provided by python-shell.
The recipes are supposed to cross-compile and install the chatter
example. In this commit, the recipe do not contain any dependencies
and configure does not run through yet.
The recipe python-pprint rdepends on python-io. As this was not reflected in
the oe-core repository, we added the dependency in genmsg.
Now after the commit abe7bf9992e298f1b53e790eee7b064a9e4e8589 in oe-core,
we can remove the runtime dependency in genmsg.
This commit resolves issue #54.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@oss.bmw-carit.de>
The topic-tools fails during the do_compile step at:
.../sysroots/x86_64-linux/usr/lib/libboost_thread-mt.so: could not read symbols: File in wrong format
The failure comes from the rostest-native package which delivers a
header file with boost dependencies and therefore add the relevant
library to the link path. Add rostest to the DEPENDS to use the
header and libraries from the cross sysroot and keep the native
rostest for the binary.
Fixed#51
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
For starting roscore, it requires a number of runtime dependencies.
At the moment, all runtime dependencies, even recursive runtime dependencies,
are added only in the roslaunch recipe. Hence, the recipe contains all
runtime dependencies.
At a later stage, one should move the recursive runtime dependencies into
the appropriate recipes.
Rostest depends on its own package and therefore needs to search
in the devel space for the rostestConfig.cmake during native
build.
| CMake Error at /.../usr/share/catkin/cmake/find_program_required.cmake:5 (message):
| rostest not found
|
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
All ROS packages with install tools in the bin directory or install
python code with is used by other packages should be used as native
version. Adapte the DEPENDS accordingly.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
Adapt genmsg to handle msg files relatively to the current
cmake path. Without this modification genmsg failed to generate
messages because it pass a path without sysroot to the python
code.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
Catkin require a .catkin file in all CMAKE_PREFIX_PATH to add them
to the CATKIN_WORKSPACES. Remove this requirement as it couldn't
be satisfy when catkin packages are installed into the common usr
prefix. Without this adaption catkin couldn't find all files.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
The recipes were created by manually extracting the information
from the corresponding package.xml.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@oss.bmw-carit.de>
After the release of ROS groovy, there is no strong reason to
continue to work on the cross-compilation of the previous ROS
version fuerte. Hence, this commit removes the initial
non-functioning recipe and the reference in the README file.
For cross-compiling ROS fuerte, the repository at
https://github.com/yida/yida-collections
might be a better starting point anyway.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@oss.bmw-carit.de>
Add the package to the dependency because the do_configure fails
as it tries to find the package.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
Cmake should search in the native sysroot for rosunit. Remove
NO_CMAKE_FIND_ROOT_PATH from the find_program call to allow
cmake to search in the native sysroot.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
Add the package to the dependency because the do_configure fails
as it tries to find the package.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
The license checksum was never tested before as the do_configure
failed until now and do_qa_configure is run after do_configure.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
The catkin framework need some modification to support cross compiling
packages via yocto. When we want to upstream patches we need to provide
patches against the current groovy-devel branch. Use this branch as base
for our work to simplify the tests and to unify the developing process
of catkin.
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>
* Use configuration options instead of patches
* Use relative paths in CMake installation files
* Use catkin class
Signed-off-by: Stefan Herbrechtsmeier <stefan@herbrechtsmeier.net>