Gary Miguel

Removing `includes` does indeed fix it. I'm a little confused as to why users of the library are able to use system-include syntax without it though. I.e. `#include `. I...

If I change it to `includes = [""]`, which I think is correct, the toolchain's local_cuda includes clobber my custom thrust's includes. You can see the behavior here: https://github.com/garymm/rules_cuda/tree/3b973fa/examples So...

I just tried that and it doesn't work for some reason... will investigate.

I updated the issue description with the latest status. I'm really not sure what's going wrong. I tried using the `-M` and `-H` options for the compiler, and I also...

Hmm, indeed there's something about the sandbox that changes the behavior. Building with `--sandbox_strategy=local` works.

And for some reason removing `includes = ["."]` from the thrust and cub rules works, even though the resulting change in the nvcc command seems like the opposite of what...

I can contribute an example that uses custom thrust and cub with no `includes =` line and then close this, but I'm a bit at a loss as to why...

@cloudhan are you OK with me removing the flag and just relying on `--dynamic_mode`? Or if not, then having the default be to respect `--dynamic_mode`?

If you assign this to me I can work on making it respect `--dynamic_mode` by default, and we can leave the flag to override.

@ezyang I've been told you have been invited to our regular meeting on 09-02 and we can discuss then. However this scope feature is needed in the short term and...