[Rd] Problem with dyn.load() under Windows 64bit at CRAN

[Uwe Ligges]

On 02.07.2010 16:10, Dominick Samperi wrote:
> 2010/6/30 Uwe Ligges<ligges at statistik.tu-dortmund.de>
>
>> On 30.06.2010 15:44, Dominick Samperi wrote:
>>
>> Another odd thing about this is that everything worked under Windows 64bit
>>> before the changes were made to serialize the build of packages that
>>> depend on each other.
>>>
>>   That's untrue. I try to serialize some things on winbuilder.
>>
>
> There was a problem building packages that depended on Rcpp due to the
> parallel build process used at CRAN (and a bug in GNU make).


Well, sometimes, *not* always.
And *only* for Windows, it was fine on other platforms.


> The CRAN
> build process was changed (by serializing dependent packages) to fix this
> problem.

*No*! It was *not* changed, as I already told you in my last reply on 
the list. I aimed at and still have not found the time. And again, only 
Windows.

I fear you do not understand what I meant: I am just stopping to build 
some packages in parallel. Other stuff is not affected, hence nothing 
you or this list need to worry about.
I should not have told you, as I understand now. I just tried to give 
you an explanation why your packages failed twice in the past due to bad 
luck.
I just should keep quite in future and answer with a simple "no idea" 
given all the information and even internal guesses are reported in a 
different context and used for list-wide speculation in the end.



> The package cxxPack built fine under Windows 64bit before this
> change was implemented. Now it fails.

No, that's not the reason: the change has not yet been implemented.


> Today the build of cxxPack fails under most of the OS's, and it appears that
> the problem is due to a new release of Rcpp.
>
> Perhaps the idea of wrapping libs inside packages simply does not work,
> at least not without extra support and attention from the CRAN team.


Why should the CRAN team care about it?

Really, given my lack of time and your annoying posts to the lists 
containing misleading information, I probably should stop to care about 
your package at all. There are > 2000 other packages and some more 
important tasks to tackle. Consider the CRAN team would take so much 
time for each single package ...

Uwe


> One possible resolution would be to incorporate whatever I need from Rcpp
> into cxxPack to eliminate the package dependency.
 >
> Thanks,
> Dominick
>