Skip to content

[R-pkg-devel] R CMD check yielding different results for me than CRAN reviewer

7 messages · Adah Zhang, Ben Bolker, Uwe Ligges +2 more

Original
Adah Zhang
# 
Hello,

I am running R CMD check (and also with the --as-cran flag), which results
in all OK's in my 00check.log. However, after submitting the package, the
CRAN reviewer tells me of a warning/error regarding one of my examples
(which is easily fix-able if I got the warning myself).

How do I ensure that all warnings, errors, and notes are addressed if I do
not receive these when running the standard checks or when testing the
examples myself? Is there a standard order of commands that a CRAN reviewer
goes through aside from 'R CMD check'?

Thanks,
Adah
Ben Bolker
# 
The most common scenario is that you're using R-release and the
warnings are coming from the R-devel version, which often has a bunch
of new, additional tests.

The other frequent answer to this question (is there a FAQ list for
this mailing list yet?) is that using CRAN's win-builder service is a
very easy way to get access to the R-release and R-devel tests without
changing anything in your local setup.
On Tue, Jan 26, 2016 at 2:03 PM, Adah Zhang <asz25 at case.edu> wrote:
Dirk Eddelbuettel
#
On 26 January 2016 at 14:03, Adah Zhang wrote:
| I am running R CMD check (and also with the --as-cran flag), which results
| in all OK's in my 00check.log. However, after submitting the package, the
| CRAN reviewer tells me of a warning/error regarding one of my examples
| (which is easily fix-able if I got the warning myself).
| 
| How do I ensure that all warnings, errors, and notes are addressed if I do
| not receive these when running the standard checks or when testing the
| examples myself? Is there a standard order of commands that a CRAN reviewer
| goes through aside from 'R CMD check'?

"Use a different computer."

That is not as flippant as it sounds because

 -- thanks to Uwe you have r-release and r-devel on Windows just one ftp
    submission away

 -- thanks to Carl and yours truly you can also try a variety of Docker
    containers for R including r-devel prebuilt and even SAN/ASAN/UBSAN
    flavours

 -- the remainer of campus should offer some choices too though that relies
    on your local contacts.

Sometimes these things are of course also due to something else but as you
supplied nothing concrete or reproducible I cannot be more specific either.

Hope this helps.  It's the right list for the question.

Dirk
Uwe Ligges
#
On 26.01.2016 20:03, Adah Zhang wrote:
The reviewer uses a recent version of R-devel (i.e. a daily built), a 
recent set of packages (i.e. run update.packages() first) and then

R CMD check --as-cran (well, at least similar, but you will get the 
warnings you talk about if you do that, as I already told you when you 
asked CRAN).

Best,
Uwe Ligges
Sven E. Templer
# 
Dear Adah,

win-builder is fairly easy, but once you've setup a github repository using

https://github.com/metacran/r-builder

you might save even that few amount of work.

Best,
Sven
Adah Zhang
# 
Thanks for all the suggestions! I'm working off of a Mac, so I don't think win-builder would work for my case.
Dirk Eddelbuettel
#
On 26 January 2016 at 16:13, Adah Zhang wrote:
| Thanks for all the suggestions! I'm working off of a Mac, so I don't think win-builder would work for my case.

Wrong. Read http://win-builder.r-project.org.  It is a _service_ that you
submit _source_ tarballs to.  Which is why I called it just one ftp away:
| >> On 26 Jan 2016, at 20:54, Dirk Eddelbuettel <edd at debian.org> wrote:
| >> -- thanks to Uwe you have r-release and r-devel on Windows just one ftp
| >>   submission away

Dirk