Jonathan Schwender

> My goal is for Corrosion to handle both setting the environment variable and adding the include path so the user doesn't even need to know about these Sorry, but...

> I never proposed using include_directories, nor do I ever recommend using those legacy non-target-specific CMake functions. Then I just really don't understand how you expect corrosion to know which...

Ah thanks, I think I understand a bit better now what you want. Basically you want to set `INTERFACE_INCLUDE_DIRECTORIES` to the location of the generated headers, so that CMake can...

https://github.com/corrosion-rs/corrosion/pull/244 will be a first step for cxx integration.

> If corrosion would like to stick non-Cargo build steps downstream of a Cargo-based invocation of cxx-build, the only correct way to implement that is by sufficiently accurately pretending to...

Is compiling in release mode required for rftrace?

If it's not required for performance reasons, I'd suggest to compile in debug mode instead. This way commit c476784 "remove lto support [...]" is not needed and can be reverted.

@stlankes In that case I'd recommend adding a new [custom profile](https://doc.rust-lang.org/nightly/cargo/reference/unstable.html#custom-named-profiles) for rftrace, so that we don't have to disable lto in release mode for an optional feature.

@jbreitbart That's a good idea! `--release` still needs the correct RUSTFLAG to be set, and until cargo implements [per profile rustflags](https://github.com/rust-lang/cargo/issues/7878), adding a new profile for lto (e,g, `release-lto`) is...

@stlankes Could you add the `enhancement` label?