[R-pkg-devel] advice on a rejected package (I think, because time was 13min, in excess of 10min limit)

[Duncan Murdoch]

On 11/05/2022 4:26 p.m., Daniel Kelley wrote:
> Thanks, Dirk.  On my machine, the test suite takes 85 seconds.  Assuming the CRAN machines are about half as fast, eliminating all the tests might get oce under the 10 minute limit.  I've gone ahead and made all the tests be local, by relying on the presence of a file that is in github, but not included in the R tarball.  Tomorrow I will also take a close look at code examples, to see what I can eliminate.  (I had already done some of that for the previous release, but I will cut closer to the bone this time.).
> 
> Hopefully these reductions will get oce under the limit. On that assumption, I will try submitting a new version tomorrow, with fingers crossed.  If I really have to, I can eliminate all the examples from the documentation.  I'd rather have the package on CRAN even without tests and examples, because it requires compilation, and not all my users are set up for that.

CRAN will object if many of your help pages don't have examples.

Duncan Murdoch

> 
> Thanks again.  I appreciate your swift and clear response.
> 
> Dan.
> 
>> On May 11, 2022, at 4:27 PM, Dirk Eddelbuettel <edd using debian.org> wrote:
>>
>> CAUTION: The Sender of this email is not from within Dalhousie.
>>
>> On 11 May 2022 at 19:07, Daniel Kelley wrote:
>> | Today I tried to update the CRAN oce package to version 1.7-3, to address a problem with 1.7-2 on one of the CRAN build platforms.  I used devtools::release() to submit it.  I just received the text I've pasted at the bottom of this email.  I've checked all the links, and my impression is that the sole problem is that the build took 13 minutes, which is over a 10-minute limit.
>> |
>> | Since oce is a big package, I am not sure that removing the tests and documentation examples would reduce the time by 30%. There is a lot of C, C++ and fortran code that takes a while to build. And a package lacking code examples and a test suite is not likely to please my users, many of whom are scientists without deep coding experience.
>> |
>> | During the previous submission, this time limit flag was also raised, and I asked R-package-devel for advice on what to do.  Someone suggested that I wait a week, in case a CRAN member would notice and put the package through. That worked, but my concern is whether this is the approach that makes for less work on the CRAN side.
>> |
>> | So, I wonder whether the advice on this list is again to leave it alone and wait, or to contact CRAN?
>>
>> The 'Overall checktime 13 min > 10 min' suggests to me that you should
>> condition some tests on an environment variable possibly present on your home
>> system, your CI setup, ... but not CRAN so that the default gets closer to
>> five minutes.
>>
>> Of course it sucks that we have to reduce coverage but CRAN is also testing
>> 18.5k packages, often recursively or in bulk when changes are considered to
>> time matters. So it is a defensible tradeoff.
>>
>> Dirk
>>
>> --
>> dirk.eddelbuettel.com | @eddelbuettel | edd using debian.org
> 
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel