checking whether package can be installed... NOTE Installation took CPU time x times elapsed time

[aitap]

Problem: Your package got a NOTE saying that during installation, it took more CPU time than elapsed time.

Cause: CRAN checks run on big, multi-core computers, running many R CMD check processes in parallel. Every process is allowed to use two CPU cores to exercise parallel code in examples and unit tests (policy: "If running a package uses multiple threads/cores it must never use more than two simultaneously"). Something that ran as part of the installation process (most likely called from src/Makevars*) started more than two child processes (or threads) at the same time.

• Example cause seen with Rust: running cargo without explicit -j 2 uses all available CPUs by default [1, 2]
• Possible cause when using CMake and/or Ninja: when using cmake -GNinja and then cmake --build, ninja will by default use all available CPUs.
• Possible cause when using Make: some third-party software may call make -j`nproc` in its build scripts to speed up the build process.

Solution: Specific to the build system your package is using, but many modern build systems use the -j flag that had originated in Make:

• For cargo, make sure to specify -j 2 any time your package is built on CRAN.
• For cmake --build, also make sure to specify -j 2 at least on CRAN builds to avoid the underlying build system implicitly using all CPUs.
• For Make, locate the explicit uses of the -j flag and make sure it doesn't exceed two at least on CRAN builds.

[Baltmann01]

Thanks for the contribution. This issue (to some extend) is already on our ToDo list. I suspect it will be added next week.
I will look at your suggestions and include them in the recipe when writing it.

[aitap]

I think I see it now. Will check the ToDo list next time I decide to file something here, thanks!

[Baltmann01]

Just as a heads-up: I just created a PR (#63) about this issue. Thanks for all your suggestions! I added more details than I had planned due to this post :)