[R-SIG-Mac] M1+ R package test failures (local and Mac Builder)

[Prof Brian Ripley]

Apple Clang 14.0.3 and its associated SDK change to MacOSX13.3.sdk broke 
about 25 CRAN packages (3 with segfaults) and 17 remain broken (and also 
with Xcode/CLT 15 beta 4 which uses a newer build of 14.0.3).  But in my 
checks mmrm 0.2.2 is not one of them.

Packages should aim to be platform-independent. So rather than expect 
all check systems to run the same toolchain, make sure your package 
works on all that are available.  I believe mac-builder uses different 
ones for 'release' and for 'development' but you write as if it uses 
just one -- and I know that there have been changes since R 4.3.0 was 
released.

It is possible to download different CLTs from 
https://developer.apple.com/download/all/ and switch between them (for 
the compiler you will need to install them in turn but you can switch 
between installed SDKs).  I have 14.2, 14.3.1 and 15 beta 4 downloaded 
and switch between them.

Given the several recent security alerts, if you are running macOS 
13.3.1 I would suggest you update to 13.5.

On 25/07/2023 14:50, Sabanes Bove, Daniel via R-SIG-Mac wrote:
> Hi R using Mac developers,
> 
> I switched yesterday to a new M2 MacBook, coming from a 3+ year old Intel
> MacBook.
> The concrete problem is that my R package mmrm (including C++ and in
> particular Eigen/TMB code) successfully compiles, but the tests fail, i.e.
> the compiled and installed R package does not behave correctly.

You give us no idea of the problems.  Please do re-read the posting guide.

> First I thought that this is just my local toolchain which is not yet
> appropriately configured, but I tried to read more on that and got OpenMP
> and gfortran and it should be ok.
> So I thought ok, maybe I will try the Mac Builder website and find out in
> which commit the R package broke.
> 
> Unfortunately, the problem is that the Mac Builder does not help us
> sufficiently here, because even the current CRAN release (0.2.2, where
> tests based on CRAN binary pass fine locally and on CRAN) leads to test
> failures on Mac Builder.
> 
> I do see a few differences between the CRAN reported setup and the mac
> Builder:
> - CRAN is running under macOS Big Sur 11.6.7, while Mac Builder is running
> under macOS Ventura 13.3.1 (same as my local machine)
> - CRAN uses C++ compiler ‘Apple clang version 13.0.0 (clang-1300.0.29.30)’
> (and not 14.0.0 as used for the R compilation!) for the package build,
> while Mac Builder uses ‘Apple clang version 14.0.3 (clang-1403.0.22.14.1)’
> (this is the same as on my local machine)
> 
> At least Mac Builder and my local machine thus use a similar setup it
> seems (even though M1 vs M2 difference exists)
> 
> My question is: what are your tips on how to move forward here? How can we
> set up the Apple clang version and maybe other build flags locally - as
> well as on Mac Builder for the whole community - to really match the CRAN
> configuration sufficiently well and thus allow for successful package
> builds?
> 
> Thanks all,
> best regards
> Daniel
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac

-- 
Brian D. Ripley,                  ripley using stats.ox.ac.uk
Emeritus Professor of Applied Statistics, University of Oxford