[Rcpp-devel] Ancient Rcpp for OS X on CRAN [Was: Building/linking trouble with cxxfunction()]

[Simon Urbanek]

On Feb 22, 2011, at 6:03 PM, Simon Urbanek wrote:

> 
> On Feb 22, 2011, at 5:51 PM, <Ken.Williams at thomsonreuters.com> <Ken.Williams at thomsonreuters.com> wrote:
> 
>> 
>> 
>> On 2/22/11 4:33 PM, "Simon Urbanek" <simon.urbanek at r-project.org> wrote:
>> 
>>> 
>>> On Feb 22, 2011, at 5:29 PM, <Ken.Williams at thomsonreuters.com>
>>> <Ken.Williams at thomsonreuters.com> wrote:
>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> On 2/22/11 4:25 PM, "Simon Urbanek" <simon.urbanek at r-project.org> wrote:
>>>> 
>>>>> On Feb 22, 2011, at 5:12 PM, Dirk Eddelbuettel wrote:
>>>>> 
>>>>>> 
>>>>>> Anyway---I will point our OS X user to source installs.
>>>>>> 
>>>>> 
>>>>> Well, I don't think that helps in any way - the test will still fail
>>>>> for
>>>>> all 10.5 users. Why don't you just fix the test? It's up to you, but
>>>>> removing Rcpp from CRAN won't help anyone (well, almost ;)).
>>>> 
>>>> 
>>>> Actually it helps a great deal, it means that 10.6 users can install
>>>> something.
>>>> 
>>> 
>>> Can you explain? You can always install from sources (provided you have
>>> all the tools etc.) - regardless whether there is a binary or not...
>> 
>> That's true - but for most OS X users, myself included until just a couple
>> hours ago, it looks like (when using the R.app GUI install tools) the
>> oldest successfully-built binary is the "latest available version", unless
>> someone like Dirk points us to the source install and says it should work
>> fine.  So it certainly helped me because I didn't know what I was missing.
>> 
> 
> Ok, I'll take it down, but note that this will have a snowball effect since it implies that all depending packages will fail as well.

Actually, thinking of it, I could initiate the build manually, skipping checks. Given that we know what fails it should be safe. However, there are currently two issues: a) I can do it only for the current version, i.e. if you update Rcpp make sure the fix is in, otherwise it won't build and b) the build machine is currently catching up R-devel packages so I can only do it when it's done -- could mean tomorrow.

Cheers,
Simon



> I can change the policy such that check failures result in the removal of the old binary - from what you said it may be useful since the users will be more likely to bug the maintainers for a fix. The drawback is that in the meantime there is no binary. Personally, either is fine with me.
> 
> 
>>>> 
>>>> What's this about removing Rcpp from CRAN though?  Just a joke I assume?
>>>> 
>>> 
>>> No, why?
>> 
>> Because it's a useful package, of course.  And CRAN is where someone
>> should go to find useful R packages.
>> 
> 
> Well, if that was true, CRAN would have a much fewer packages ;).
> 
> Cheers,
> Simon
>