[R-SIG-Mac] Installing CRAN binary packages with R 4.0 installed from source crashes R

[Prof Brian Ripley]

On 02/04/2020 09:34, Simon Urbanek wrote:
> Hervé,
> 
> "both" is a fairly recent addition and my guess would be that it has been guarded specifically since it is the default and installing binaries only works for the CRAN version. I didn't look at the new "both" code to see how it knows that it's the CRAN version - there is really no special "CRAN" flag. At some point we were guarding binary installs in general by checking the OS and R, but it was fragile - you could be building using the same system as we do and yet use a different compiler, so I think it's in general impossible unless we introduce some extra identification of the binaries. So, yes, if you compile R from sources yourself it is not guaranteed to be compatible with CRAN package binaries. Those are only built and tested with the CRAN R binary.

It is simple: type = 'both' has to know what the two types are.  Only 
the CRAN binaries have the default type set to "mac.binary": building 
from the sources gives you a default type of "source".  See ?.Platform.

> 
> Cheers,
> Simon
> 
> 
>> On 2/04/2020, at 7:47 PM, Hervé Pagès <hpages using fredhutch.org> wrote:
>>
>> Hi Simon,
>>
>> After installing R 4.0 alpha from source on a macOS Mojave system, R won't let me use type="both" to install CRAN packages. I get:
>>
>>   Error in install.packages("rJava", type = "both", repos = "https://cran.r-project.org") :
>>     type == "both" can only be used on Windows or a CRAN build for macOS
>>
>> OK so this suggests that the CRAN binary packages for R 4.0 are not compatible with my R. Surprisingly though using type="mac.binary" doesn't complain and lets me install these binaries. But then trying to load them causes segfaults. I've tried this with rJava, Rcpp, ggplot2, and doing library() on any of them crashes my session. Note that installing all these packages from source works without any problem.
>>
>> So my questions are: is it the case that CRAN binary packages are not meant to be used with an R 4.0 installed from source? If yes then why isn't type="mac.binary" blocking this like type="both" does?
>>
>> Thanks,
>> H.
>>
>>> sessionInfo()
>> R version 4.0.0 alpha (2020-04-01 r78132)
>> Platform: x86_64-apple-darwin18.7.0 (64-bit)
>> Running under: macOS Mojave 10.14.6
>>
>> Matrix products: default
>> BLAS:   /Users/biocbuild/bbs-3.11-bioc/R/lib/libRblas.dylib
>> LAPACK: /Users/biocbuild/bbs-3.11-bioc/R/lib/libRlapack.dylib
>>
>> locale:
>> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
>>
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>
>> loaded via a namespace (and not attached):
>> [1] compiler_4.0.0
>>
>>
>> -- 
>> Hervé Pagès
>>
>> Program in Computational Biology
>> Division of Public Health Sciences
>> Fred Hutchinson Cancer Research Center
>> 1100 Fairview Ave. N, M1-B514
>> P.O. Box 19024
>> Seattle, WA 98109-1024
>>
>> E-mail: hpages using fredhutch.org
>> Phone:  (206) 667-5791
>> Fax:    (206) 667-1319
>>
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac using r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>>
> 
> _______________________________________________
> 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