Can I override the `check` path?

[yoni]

I'm running continuous integration builds on a Jenkins server and need the R CMD check output to remain in the project working directory. Otherwise, I can't diagnose the issues on build failures.

I've tried a couple of approaches, including passing check(..., args=c('--output', 'my_check_dir')). The problem is, I'm pretty sure devtools overrides the path separately, resulting in a conflict.

[hadley]

It doesn't look like it. What about patching check() to return r_cmd_check_path in the case of failure, and then you could add custom handling? Alternatively, we could add the check_dir param to check() and pass that on down to check_r_cmd()

[yoni]

Thanks for the response, @hadley. :)

Returning r_cmd_check_path implies that the caller needs to add error handling to their code. I like the top-down check_dir option better.

[yoni]

Thanks for adding this, @hadley! It was on my TODO list, but I never quite got the time.