[Rd] r-devel windows gcc10-UCRT machine questions

Tomas Kalibera tom@@@k@||ber@ @end|ng |rom gm@||@com
Thu Mar 11 16:46:23 CET 2021

On 3/11/21 4:21 PM, Dirk Eddelbuettel wrote:
> I saw two (unchanged, long-existing) tests of main fail narrowly on this new
> platform (relative to the tolerance argument set).
Thanks for already looking at the results. I'd be happy to have a look 
if you point me to the example.
> Attempting to change the tolerance if .Platform$OS.type == "windows" failed
> (any idea why that test would evaluate to FALSE?).  Could it be that
> capabilities("long.double") returns FALSE?
.Platform$OS.type == "windows" and capabilities("long.double") are both 
true on the test system.
> As always, it is a little challenging to keep our stacks in order without
> access to the platforms added.  So with thanks in advance for any pointers.
> And of course for working towards utf8 on windows.

There is no checking service yet for this, but it should be simple to 
install things to Windows 10 following documentation at


If you just want to install a CRAN version of your package and run some 
R code using it, you should only need to install R from the binary 
installer, and from it install your package (there are binary builds of 

I'd recommend to do this in a fresh VM. One can get a free VM (time 
limited) from Microsoft for testing and run it in say VirtualBox.

If doing this on a system with the normal R-devel installed, it is 
better to remove it first including the package library directory, and 
vice-versa, remove the UCRT-based R-devel before installing the normal 
one. Or use some other usual measures to make sure that each 
installation has its own package library directory.

> Dirk

More information about the R-devel mailing list