On Mon, 2011-11-28 at 16:17 -0500, Marc Ferland wrote: > Hi, > > I'm currently writing a recipe for the visualization toolkit (vtk). > > To compile, this library invokes executables that are generated > on-the-fly by the compilation process (a little bit like Qt and > qmake). > > The path to these executables can be specified to cmake when building > vtk. So far I was able to make it work by hard-coding this "tools" > directory in my recipe to a path on my local machine. It works, but it > is not very portable. > > What's the official way to handle such libraries? > > Should I first do a native build, then use this native build directory > when cross-compiling? If so, is there any examples I should look into?
Yes, a -native recipe to build the tools and then use those native tools in the cross built is the way to go. Take a look at any of the recipes which have a DEPENDS on a native version of themselves. A simple example is say, the bison recipe and the logic is something like: DEPENDS = "bison-native" DEPENDS_virtclass-native = "" BBCLASSEXTEND = "native" (simplified slightly) Cheers, Richard _______________________________________________ yocto mailing list yocto@yoctoproject.org https://lists.yoctoproject.org/listinfo/yocto
