You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The comment mentions avoid finding the library downloaded in WORKSPACE B when building workspace A which I'm not sure I fully understand, however forcing everyone to either install yaml-cpp in the system or to download and build it in place does not seem a universally valid approach.
There are plenty of situations where yaml-cpp would be already available and it should be used (for example if a package manager such as conan is being used), rather than introducing conflicting dependencies in the system.
What do you think?
Do we still need that flag? If yes, can we add a CMake variable to optionally disable it?
The text was updated successfully, but these errors were encountered:
So that change was originally part of a sequence of events. We first put in #8 to make things use the system installed yaml-cpp, which was reverted in #15, and then was re-applied with NO_CMAKE_PACKAGE_REGISTRY in #16. I think #16 has some of the breadcrumbs of why we applied the fix to begin with. @ivanpauno maybe you can shed some light on why we needed this fix, if you remember?
IIRC, the issue is that was finding the package in the cmake registry, but it was causing issues at runtime because the directory was not added to LD_LIBRARY_PATH. See discussion in #16 for more details.
Currently the trick is that we add an enviroment hook when the package is being built
The presence of
NO_CMAKE_PACKAGE_REGISTRY
flag in https://github.com/ros2/yaml_cpp_vendor/blob/master/CMakeLists.txt#L71-L75 results inyaml-cpp
to be searched only in system locations.The comment mentions
avoid finding the library downloaded in WORKSPACE B when building workspace A
which I'm not sure I fully understand, however forcing everyone to either installyaml-cpp
in the system or to download and build it in place does not seem a universally valid approach.There are plenty of situations where
yaml-cpp
would be already available and it should be used (for example if a package manager such as conan is being used), rather than introducing conflicting dependencies in the system.What do you think?
Do we still need that flag? If yes, can we add a CMake variable to optionally disable it?
The text was updated successfully, but these errors were encountered: