Ville Voutilainen via SG15 <sg15_at_[hidden]> writes:

> But that also doesn't necessarily mean dynamic build graph nodes. I'm
> not a build system expert, but I do know how to construct my (CMake)
> builds so that the generative build is done before the builds that
> need its results, so that the generation results are there when the
> next step is configured as part of an overall build (the build, not
> the configure-step of the overall build), and it then feeds Ninja a
> perfectly static build graph.

How would you handle the Apache Thrift example I mentioned in a
sibling reply? Do you expect the user to re-run CMake every time
they change their IDL file?