[R-SIG-Mac] problem with options("pkgType") in a home-built R 2.2.0

[stefano iacus]

On 15/ott/05, at 00:18, Don MacQueen wrote:


> David,
>
> Stefan has already answered.  None the less, here's my perspective.
>
> Like you, I build from source, and have also discovered the change in
> default 'type' for
> install.packages() and friends. Like you, I left the default
> --with-aqua unchanged.
>
> Here's a relevant excerpt from the NEWS file.
>
>    o    There is now support of installing from within R both  
> source and
>     binary packages on MacOS X and Windows.     Most of the R
>     functions now have a 'type' argument defaulting to
>     getOption("pkgType") and with possible values "source",
>     "win.binary" and "mac.binary".    The default is "source" except
>     under Windows and the CRAN GUI build for MacOS X.
>
> I would think that the configure option --with-aqua causes a build to
> be a "CRAN GUI" build, because
> --with-aqua is what gets you the option of using the double-clickable
> app that goes in your /Applications folder.
>

let's say it provides  cocoa/aqua callbacks, that's all about the -- 
with-aqua flag.
The fact that it also sets mac.binary it is unwanted behaviour and we  
need to change this. The default should be "source" as from the NEWS  
file.

mac.binary is just for CRAN binaries and, as I said, we decided to  
set it for the cran build (but unfortunately it is also set for non  
cran build) because we want to be sure about compatibility between  
cran R binary for OS X and binary packages for both BioC and CRAN.
We need this because otherwise feedback from user is difficult to  
manage as there are several binary R around (CRAN, Jan's, fink,  
darwinport and I've also seen a binary version of R shipped with  
GRASS for OS X).
We got lots of feedback and R-sig-mac is just the top of the iceberg :)

stefano


>
> I decided to live with having to set type to 'source' each time I use
> it; it's not that much of an extra burden. I'd suggest building
> without aqua and finding out if that sets the default back to
> 'source'.
>
> I build from source on my primary machine because I also am also the
> local installer for R on Solaris, and I find it easier when things
> are similar on both platforms. I normally send other users to the
> Aqua version, and for them the default of 'mac.binary' makes sense.
>
> -Don
>
> At 10:32 PM +0100 10/14/05, David Firth wrote:
>
>
>> I have built R 2.2.0 from source, on Mac OS X 10.4, with the
>> configure step as follows:
>>
>> ./configure --with-x --with-tcltk F77=gfortran --with-blas="-
>> framework vecLib" --with-lapack
>>
>> (yes, I know some of that is redundant now!)
>>
>> Al goes fine with the build.  But when I run R I find
>>
>>
>>
>>> version
>>>
>>>
>>           _
>> platform powerpc-apple-darwin8.2.0
>> arch     powerpc
>> os       darwin8.2.0
>> system   powerpc, darwin8.2.0
>> status
>> major    2
>> minor    2.0
>> year     2005
>> month    10
>> day      06
>> svn rev  35749
>> language R
>>
>>
>>> options("pkgType")
>>>
>>>
>> $pkgType
>> [1] "mac.binary"
>>
>>
>>> contrib.url(options("repos"))
>>>
>>>
>>                                                           CRAN
>> "http://cran.uk.r-project.org/bin/macosx/powerpc/contrib/2.2"
>>
>> This creates problems for me when I try to use for example
>> install.packages:
>>
>> 1.  If I try to run install.packages, the URL that gets used does not
>> exist (as I also verified outside of R):
>>
>>
>>> install.packages("eha")
>>>
>>>
>> Warning in install.packages("eha") : argument 'lib' is missing:
>> using /Users/david/Library/R
>> Warning: unable to access index for repository http://cran.uk.r-
>> project.org/bin/macosx/powerpc/contrib/2.2
>> Warning in download.packages(pkgs, destdir = tmpd, available =
>> available,  :
>>           no package 'eha' at the repositories
>>
>> 2.  My reading of ?options is that options("pkgType") for me should
>> be "source" not "mac.binary", since it says
>>       'pkgType': The default type of packages to be downloaded and
>>            installed - see 'install.packages'. Possible values are
>>            '"source"' (the default except under the CRAN Mac OS X  
>> build)
>>            and '"mac.binary"'.
>> I do not think I am running "the CRAN Mac OS X build" -- I have never
>> downloaded the R binary from CRAN, and I assume (correctly?) that
>> that is what is meant there.  If I do options(pkgType = "source") all
>> is fine.  But should I have to do that myself?
>>
>> This behaviour seems to be new to R 2.2.0.  Have I blundered
>> somewhere in building R?  I noticed that --with-aqua is now a default
>> configure option, but I didn't negate that because (a) I wasn't sure
>> what it does [what does it do?], and (b) I supposed it was the
>> default for good reason.
>>
>> I got the same behaviour when building under Mac OS 10.3.9, with
>> ./configure --with-x --with-tcltk --with-tcl-config=/sw/lib/
>> tclConfig.sh --with-tk-config=/sw/lib/tkConfig.sh  --with-blas="-
>> framework vecLib" --with-lapack
>>
>> I would welcome any advice.  I build R myself mainly because I like
>> to use ESS as my interface.
>>
>> Yours hopefully,
>> David
>>
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>>
>>
>
>
> -- 
> --------------------------------------
> Don MacQueen
> Environmental Protection Department
> Lawrence Livermore National Laboratory
> Livermore, CA, USA
>
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>
>