Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Ken Beath
Simon, I’m going to investigate further, but the problem is that 4.0 works fine but 4.1 doesn’t. I will check what is going wrong. Ken > On 20 May 2020, at 1:47 pm, Simon Urbanek wrote: > > Ken, > > R can be built with just Xcode and GFortran (see also > http://mac.r-project.org/tools/ ).

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Simon Urbanek
Adelchi, as pointed out you have conflicting libraries in /usr/local - likely from some conflicting package manager. I would recommend re-naming /usr/local to remove the issue and installing a clean GNU Fortran (see http://mac.r-project.org/tools/ or from the official GNU Fortran pages which

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Simon Urbanek
Ken, R can be built with just Xcode and GFortran (see also http://mac.r-project.org/tools/ ). If you have issues, those are very often due to conflicting package managers (Homebrew, macoports, fink ...) so a good exercise is to simply re-name /usr/local/ and setup a clean environment as above

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Ken Beath
Plus one of the duties of package managers is to make sure that they build and pass CRAN checks. Doing that with a non standard environment seems a bit risky. At the moment I can’t get anything to build on R-Devel which is making life difficult and I may raise a question if it still isn’t

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread David Winsemius
On 5/19/20 1:54 PM, Kevin Ushey wrote: The main thing that stands out to me in that error: Referenced from: /usr/local/lib/libquadmath.0.dylib Expected in: /usr/local/lib/libgcc_s_x86_64.1.dylib R bundles its own copies of libquadmath and libgcc:

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread David Winsemius
On 5/19/20 1:46 PM, Yoshihiko Baba via R-SIG-Mac wrote: > Adelchi, > > Why don’t you try Fink version of R? > > Fink build mnormt on R 4.0 as well (although it is experimental). > https://github.com/fink/fink-distributions/issues/623 This should be considered contrary to recommended practice:

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Kevin Ushey
The main thing that stands out to me in that error: Referenced from: /usr/local/lib/libquadmath.0.dylib Expected in: /usr/local/lib/libgcc_s_x86_64.1.dylib R bundles its own copies of libquadmath and libgcc: kevinushey@Kevins-MBP:/Library/Frameworks/R.framework/Resources/lib $ ls -la total

Re: [R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Yoshihiko Baba via R-SIG-Mac
Adelchi, Why don’t you try Fink version of R? Fink build mnormt on R 4.0 as well (although it is experimental). https://github.com/fink/fink-distributions/issues/623 Fink allows you to install different versions of R (3.6, 3.5, 3.4, 3.3, 3.2 and 3.1). For different versions, CRAN packages are

[R-SIG-Mac] Cannot compile Fortran on macOS

2020-05-19 Thread Adelchi Azzalini
Hi. In essence, my problem is that I cannot install a package which requires Fortran-77 compilation on my Mac. More in detail, this is my own package 'mnormt' which exists on CRAN since 2006; it is currently at version 1.5-7. It has been developed in successive versions on a Linux machine,