#rstats package dev question: can I make R CMD check exit with a status code 1 when there are warnings but no errors? Alternatively, how does CRAN distinguish warning and error status on package check pages?
1
3
1
If I recall correctly, the `rcmdcheck` utility package by @GaborCsardi gives you a summary object you can act upon. I have been using for ages via the `rcc.r` shell / command-line wrapper in my `littler` package (but do not act on the summary there).
2
2
Replying to @eddelbuettel @thosjleeper @GaborCsardi
As to 'how does CRAN ...': It's all in the open in base R. If you look ati `R CMD check`, it dispatches to `tools::check_packages()`. These are all great resources for study (and while I don't remember if I saw it write out intermediate results I recommend these for study).

Nov 24, 2021 · 10:52 PM UTC

1
2
Replying to @eddelbuettel @GaborCsardi
Right, I've looked at that code. I think, though, the package check results page post-processes that in order to produce the OK, NOTE, WARNING, ERROR summary table. And I'm not sure that code is open sourced.
1
I lean towards taking the other side. Look at check.R, checktools.R, CRANtools.R, ... which cover a lot. I.e. I just "discovered" `summarize_CRAN_check_status()` (which is called too at the of the main checking routine). But twitter is the wrong place for this so I stop now.
1
1
more replies