[R-SIG-Mac] [Rd] My first package

[Simon Urbanek]

On Jan 20, 2014, at 10:34 AM, Federico Calboli <f.calboli at imperial.ac.uk> wrote:

> 
> On 20 Jan 2014, at 15:15, Gábor Csárdi <csardi.gabor at gmail.com> wrote:
> 
>> On Mon, Jan 20, 2014 at 9:58 AM, Federico Calboli <f.calboli at imperial.ac.uk> wrote:
>> [...]
>> I do that *with the current release of R* and I never had an issue whatsoever.  Incidentally the words 'This should be done with the current version of R-devel (or if that is not possible, current R-patched or the current release of R' translate in plainspeak 'run R CMD check --as-cran with whatever version of R provided it's not outdated'.
>> 
>> That's an interesting "translation".
> 
> That's as interesting as the English language makes it -- please note of the expression 'should' as opposed to 'must', and the list of three options of R versions.
> 

Yes, in that order. Note also *if that is not possible*. The whole point is that the check on CRAN *will* be run against R-devel which is what you have to satisfy. By not using R-devel you're taking the chance of not passing some tests and creating extra work for the CRAN maintainers to give you the output that you could have gotten yourself and request fixes - that's why it says "should". Given the number of package submissions, the effect of people that choose to ignore the rules compounds and thus creates a significant preventable workload which the above rule attempts to minimize.


>> Are you a CRAN maintainer?
> 
> I maintain two packages on CRAN, if that's what you're asking.
> 
>> 
>> I guess you were just lucky so far. Most packages are small and not affected by changes between R-release and R-devel. But some of them are. 
>> 
>> Also, R-devel typically has more rigorous package checks than R-release.
> 
> Does it?  
> 

Yes, often.


>> My package was delayed several times because it did not pass checks in R-devel (it did in R-release). I guess this happened to other packages as well, hence the sentence above was included in the policies.
> 
> I presume that the issue is this: some packages are written solely in R, hence, unless one uses a particular and outdated syntax, there are no differences between R-devel and and R-current,

Wrong


> whereas packages that call C/C++/Fortran code are subject to the vagaries of compilers and operating systems, and using R-devel should help future proofing the package, while maintaing compatibility with R-current.
> 

You're asked to use R-devel and on OS X it's readily available including binary package dependencies.

Cheers,
Simon