On Oct 6, 2015, at 3:12 AM, Rainer M Krug <rai...@krugs.de> wrote: > Simon Urbanek <simon.urba...@r-project.org> writes: > >> On Oct 5, 2015, at 3:34 AM, Rainer M Krug <rai...@krugs.de> wrote: >> >>> peter dalgaard <pda...@gmail.com> writes: >>> >>>> <Round of applause, please!> >>> >>> Agreed. >>> >>> I just would like to dd that in the case of using homebrew, which >>> installs everything under /usr/local/... which did not cause any >>> problems at all. I moved the contents of /usr/local to an other location >>> before upgrade because the upgrade to Yosemite took because of a large >>> /usr/local several hours instead of about half an hour, moved it back >>> afterwards and everything worked fins without problems. >>> >>> One question concerning mixed usage of homebrew and the Official R >>> installers: >>> >>> do the Official R installers fail if an R binary already exist, or do >>> they raise a warning? >>> >>> To make this easier, would it be possible, to install the binaries under >>> /usr/local/bin/R.X.Y.Z and then just create links in /usr/local/bin ? >>> Tghis would make the whole process more transparent and easier to switch >>> between different versions and means of installation. >>> >> >> R installer never installs any R binaries outside of the >> framework/app. The only thing we provide on 10.10 is a softlink for R >> and Rscript in /usr/local/bin into the framework (removing anything >> else in that name if it exists). In earlier OS X versions this applies >> to /usr/bin instead. > > OK - thanks for the clarification. > > Maybe I am overly careful, but I would very much prefer that the user is > asked how to proceed if something would be overwritten. > > Due to e.g. homebrew /usr/local/bin becomes used quite a bit, so an > approach which does not overwrite existing files / links in > /usr/local/bin would be, in my opinion, a plus. > > Homebrew does for example installs the files, but does issue an error if > the links in e.g. /usr/local/bin would overwrite existing files / links. >
Homebrew doesn't really abide by any Apple rules and it's command line tool, so whatever it does is quite irrelevant here. Users use Homebrew entirely at their own risk. R is using the standard Apple Installer and for obvious reasons AI doesn't pester the user about making low-level decisions (since 99% of users just want it to work - just imagine AI asking about every file when you upgrade OS X ;)). Also if you mix native R and Homebrew R you're already in trouble, so you better know what you're doing. That said, one thing we could do would be to re-name whatever we find to replace so you can move it back if you so desire. (Of course if you install it twice your backup is gone etc.). Cheers, Simon > Cheers, > > Rainer > >> >> Cheers, >> Simon >> >> >>> Thanks for the clarifications, >>> >>> Rainer >>> >>>> >>>> -pd >>>> >>>>> On 04 Oct 2015, at 18:53 , Prof Brian Ripley <rip...@stats.ox.ac.uk> >>>>> wrote: >>>>> >>>>> The machine which provides the 'r-devel-osx-x86_64-clang' checks on >>>>> the CRAN check farm has been upgraded from Yosemite to El Capitan >>>>> and a complete round of checks has been run. >>>>> >>>>> >>>>> 1) There is a lot of misinformation around about 'System Integrity >>>>> Protection' aka 'rootless'. >>>>> >>>>> Upgrading to El Capitan moves files which are not allowed under /usr >>>>> to /Library/SystemMigration/usr, so you will be able to see what was >>>>> lost. This includes /usr/bin/R, /usr/bin/Rscript (but the installer >>>>> installs these under /usr/local/bin on El Capitan as from R 3.2.2), >>>>> /usr/X11R6, /usr/texbin . Contrary to reports from betas, the link >>>>> /usr/X11 is preserved. >>>>> >>>>> If an installer tries to create a disallowed file such as >>>>> /usr/bin/R, this is silently ignored (at least in the cases we >>>>> tested). So you can install e.g. R 3.1.3 but the executables will >>>>> not appear in the default Terminal path (more details in the current >>>>> manual). >>>>> >>>>> >>>>> 2) After updating you need to re-install the Command Line Tools and >>>>> R (to get the links in /usr/local). I did not need to re-install >>>>> Java nor XQuartz. >>>>> >>>>> >>>>> 3) All the 'Mavericks' binary packages tested worked. The source >>>>> packages of rJava and rgl (only) cannot be installed and the >>>>> maintainers have patched versions available. >>>>> >>>>> >>>>> There is updated information in the latest 'R Installation and >>>>> Administration' manual in R-patched and R-devel (in the sources, or >>>>> the online versions at https://cran.r-project.org/manuals.html will >>>>> update in a day or two). >>>>> >>>>> -- >>>>> Brian D. Ripley, rip...@stats.ox.ac.uk >>>>> Emeritus Professor of Applied Statistics, University of Oxford >>>>> 1 South Parks Road, Oxford OX1 3TG, UK >>>>> >>>>> _______________________________________________ >>>>> R-SIG-Mac mailing list >>>>> R-SIG-Mac@r-project.org >>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac >>> >>> -- >>> Rainer M. Krug >>> email: Rainer<at>krugs<dot>de >>> PGP: 0x0F52F982 >>> _______________________________________________ >>> R-SIG-Mac mailing list >>> R-SIG-Mac@r-project.org >>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac > > -- > Rainer M. Krug > email: Rainer<at>krugs<dot>de > PGP: 0x0F52F982 _______________________________________________ R-SIG-Mac mailing list R-SIG-Mac@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-mac
