[R-SIG-Mac] Problems with Rmpi

[Prof Brian Ripley]

On 22/01/2026 06:55, Ken Beath wrote:
> I’ve received an e-mail from Kurt Hornik, that they are going to remove the doRNG package because of build problems.

That is not what his message says -- the biggest issue is that it is 
unmaintained (except for multiple patches by CRAN) and the message asked 
for a volunteer maintainer.

Rarely are packages archived (not removed) from CRAN because of failures 
on macOS.  In the only instances I recall it was because such failures 
were easily corrected but the package 'maintainer' did not do so.

> Looking at it, it is only because the Rmpi package fails to build on MacOS and hence any packages that use it directly or indirectly fail to build. Is that something that is fixable?

It does build, and I use it for the M1mac checks.  But as the README 
there says

"External libraries are where possible installed via minor
modifications to Simon Urbanek's 'recipes' at
https://github.com/R-macos/recipes .  The main exceptions are those
which need to use dynamic libraries (such as openmpi ..."

I have never seen a Mac setup where MPI would fit well -- it is most 
advantageous in clusters of many CPUs with fast interconnects, the realm 
of modern supercomputers.


-- 
Brian D. Ripley,                  ripleybd using icloud.com
Emeritus Professor of Applied Statistics, University of Oxford