[R-pkg-devel] Suggestion: improving checks during actual package submission

[Uwe Ligges]

On 16.09.2025 22:10, Leo Mada via R-package-devel wrote:
> Dear Duncan,
> 
> Unfortunately, I am not fully versed in the build project. I did use devtools::check.
> 
> Looking at the args, it seems that cran = TRUE is set by default:
> args(devtools::check)
> # function (pkg = ".", document = NULL, build_args = NULL, ...,
> #    manual = FALSE, cran = TRUE, remote = FALSE, incoming = remote,
> #    force_suggests = FALSE, run_dont_test = FALSE, args = "--timings",
> #    env_vars = c(NOT_CRAN = "true"), quiet = FALSE, check_dir = NULL,
> #    cleanup = deprecated(), vignettes = TRUE, error_on = c("never",
> #        "error", "warning", "note"))
> 
> However, users may have older versions of R. So I wouldn't bet on the results which users get on their machines.

This is explained in the policies: A recent version of R should be used, 
at least R-release, better R-devel.

People may be used to accept policies unconditionally. For CRAN, please 
really read the policies.

Best,
Uwe Ligges



> 
> GitHub Check:
> I have set up some scripts for automatic checking on the GitHub page of Rpdb (at the sugestion from this list). But they didn't catch the date error either (and I do not remember if I uploaded the 1st resubmitted version in order to detect the version issue). That was a quick-fix (though I did run the local devtools:check).
> 
> Sincerely,
> 
> Leonard
> 
> ________________________________
> From: Duncan Murdoch <murdoch.duncan using gmail.com>
> Sent: Tuesday, September 16, 2025 8:51 PM
> To: Leo Mada <leo.mada using syonic.eu>; List r-package-devel <r-package-devel using r-project.org>
> Subject: Re: [R-pkg-devel] Suggestion: improving checks during actual package submission
> 
> On 2025-09-16 11:58 a.m., Leo Mada via R-package-devel wrote:
>> Dear CRAN Team,
>>
>> Please find below some ideas how to improve the checks during a package submission.
>>
>> I recently submitted a new version of package Rpbd, but made some trivial mistakes during the first 2 attempts. Both could have been easily caught during the actual submission process - before even running any tests.
>>
>> 1. Forgot to update date
>> The user should be warned if date is older than 1 week (and raise error if date is older than ... whatever triggers the error).
>>
>> 2. Version with leading zeroes
>> I resubmitted as version 2.4.01 - as it was just a minor update and was not aware of the generated error.
>>
>> Both fields can be easily checked when uploading the file with the package: the DESCRIPTION file is processed anyway. The user can easily correct these errors, avoiding unnecessary checks, including additional manual checks.
> 
> You shouldn't have been uploading the package unless you had already run
> the checks locally (or on one of the several different available check
> services).
> 
> After all, you did check the box saying "I have checked the submission
> using R CMD check --as-cran and a current version of r-devel, as
> mandated by the CRAN Repository Policy. (You could do so using the
> win-builder service at https://win-builder.r-project.org)".
> 
> Duncan Murdoch
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel