[Rd] [External] Re: Update on rtools4 and ucrt support

[Simon Urbanek]

Avi,

thanks. Yes, the whole point of the developer blog posts by R-core are for uses to provide feedback, so that's great - it's odd that it required a somewhat orthogonal post to start the discussion several months later, but I'm glad we got here.

Note that the point of the switch is to iron out all issues that may be encountered and provide real CRAN testing - that's why both R-core and CRAN is involved in all this. This only affects R-devel since we want to be ready for the R 4.2.0 release, it won't affect the current release builds.

Anyway, now I'll leave it to the Windows experts to address the details and work together.

Cheers,
Simon


> On Aug 24, 2021, at 10:03 AM, Avraham Adler <avraham.adler using gmail.com> wrote:
> 
> 
> Thank you, Simon. That was valuable. Skimming that quickly, I get a bit concerned. I’ve been building from source and then using OpenBLAS in my R source for many, many years now, and it looks like its support is tenuous in the experimental chain. Similarly with packages like nloptr, where I build NLOPT 2.6+ and have adjusted Jelmer’s code for it to work in R for Windows. I maintain packages with Fortran/OpenMP and Rcpp(Parallel). I hope that should the decision be made to switch, it will be done when the build process is more streamlined, especially for some fundamental packages. 
> 
> That being said, I must take the opportunity again to thank R-core, Tomas, Jeroen, Duncan among many others who have built an infrastructure that allows amateur programmers to contribute to the statistical infrastructure. 
> 
> Thanks again,
> 
> Avi