On Nov 13, 2012, at 1:59 PM, Prof Brian Ripley <rip...@stats.ox.ac.uk> wrote:
> Marc, > > Start with 'which R', or run /usr/bin/R explicitly. Hi Prof Ripley, Thanks. ~ which R /usr/bin/R /usr/bin/R called directly launches 32 bit R. R64 launches 64 bit R as one might expect: ~ ls -l /usr/bin/R64 lrwxr-xr-x 1 root wheel 49 Nov 13 12:15 /usr/bin/R64 -> /Library/Frameworks/R.framework/Resources/bin/R64 > That should be a symlink to > > /Library/Frameworks/R.framework/Resources/bin/R > ~ ls -l /usr/bin/R lrwxr-xr-x 1 root wheel 47 Nov 13 12:14 /usr/bin/R -> /Library/Frameworks/R.framework/Resources/bin/R > and that should be a symlink, If it is not like > > lrwxr-xr-x 1 root admin 3 28 Oct 08:03 > /Library/Frameworks/R.framework/Resources/bin/R@ -> R64 > > or is linked to R32, change it to point to R64. ~ ls -l /Library/Frameworks/R.framework/Resources/bin/R -rwxrwxr-x 1 root admin 8775 Oct 26 11:22 /Library/Frameworks/R.framework/Resources/bin/R > > It is the installer which determines what it links to. Specifically > https://svn.r-project.org/R-dev-web/trunk/CRAN/QA/Simon/R-build/packaging/leopard/scripts/postflight > . > > If running /Library/Frameworks/R.framework/Resources/bin/R64 is not x86_64, > come back to us. That does run 64 bit R, as expected. Looks like a key flag in the shell script in SVN is: ~ /usr/sbin/sysctl hw.cpu64bit_capable hw.cpu64bit_capable: 1 Going to try to work my way through the shell script a bit more, but wanted to get back with this information. Thanks, Marc > > Brian > > On 13/11/2012 19:06, Marc Schwartz wrote: >> Hi All, >> >> I am on a fully updated MacBook Pro, running 10.8.2. I get: >> >> uname -m x86_64 >> >> from the terminal, as would be expected. >> >> I have been using the CRAN OSX binaries for some time now, rather >> than building from source, which I had been doing previously. >> >> Somewhere along the way in the past few weeks, apparently since I >> installed R 2.15.2, the default architecture under which R is running >> is now 32 bit, not 64 bit, which was the case previously. I just >> noted this today due to some funny encoding issues that I had not >> seen before and have spent the past few hours trying to figure out >> what changed. >> >> I initially thought something was amiss with the latest 2.15.2 >> release .pkg file. I completely removed R (Framework and symlinks to >> the startup scripts, etc.) and re-installed. Same thing. 32 bit R was >> the default link from 'R'. >> >> So I removed R again and re-installed 2.15.1 (getting the older >> binary from CRAN), since that was the last version of R that I had >> installed which defaulted to 64 bit. >> >> Funny, same thing, it defaulted to 32 bit R. >> >> Then I wondered if there was something related to some anti-virus >> software (Avast) that I had recently installed due to some events >> that had occurred recently. I completely removed the AV software, >> rebooted, removed R and then re-installed R. Same thing, 32 bit R as >> the default. >> >> What am I missing here? What is the installation program and/or the R >> startup script itself looking for that determines whether 64 or 32 >> bit R should the default when one simply uses 'R' to start it up? A >> read of the R startup script suggests that the output (as above) of >> 'uname -m' being 'x86_64' may be all that is needed, but perhaps I am >> missing something else. >> >> I am also attaching the full installation log file here (for 2.15.2). >> I did not see anything there obvious to my eyes. >> >> I can't recall the timeline well enough right now to consider whether >> some OSX update changed something, or if there is something strictly >> unique to my MBP that is causing this problem. >> >> Thanks for any insights. >> >> Regards, >> >> Marc _______________________________________________ R-SIG-Mac mailing list R-SIG-Mac@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-mac
