[R-SIG-Mac] Problems loading Rcmdr on a Mac 10.7.2 and a Mac 10.6.8

[Simon Urbanek]

John,

On Sep 19, 2013, at 10:29 AM, John Fox wrote:

> Dear Simon,
> 
> Thank you very much for these clarifications (and also for responding to
> David and Sarah's posts in this thread). I think that I understand what's
> going on now, and will add a note to the Rcmdr installation instructions to
> run Software Update prior to installing R, the Rcmdr, and (if necessary)
> XQuartz.
> 
> (Sarah: You could try asking your student(s) to run Software Update, and
> then reinstall R and the Rcmdr package. If you do this, can you report to
> the list whether it works?)
> 
> I do have one further question: I understand now that R will use an older
> X-11 in preference to XQuartz if XQuartz doesn't install a symbolic link to
> /usr/X11. Is there a reason for this? That is, why not have R use XQuartz in
> preference to another X-11 if XQuartz is installed -- or to pick the newest
> X-server, if it's possible to ascertain that?
> 

It's not. This is not about preferences - R has no choice, it has to links against something and we picked something that's more likely to exist. You cannot pick and choose at run time since the libraries (with paths) are linked in. Note that this is *not* about the X11 that the user will be running - this is just for internal libraries use by R - these are two entirely separate things (server vs client). Also XQuartz was really a development project, so it didn't seem like a good idea to require it - in particular since we would ask people to install parallel X11 implementation to an existing one - and have to answer the "why?". In the meantime Apple has removed system X11 completely from recent OS X so it may be worth re-visiting that option and just require XQuartz. However, there is no guarantee that they will support the targets we support (although so far they do). Finally, XQuartz will act as system X11 but not vice-versa, so it's much safer to use the system X11 location as it supports more setups.

FWIW: I think this whole discussion is getting off the original topic: if you don't keep your system up to date, nothing is guaranteed to work - it has nothing to do with X11, it affects all system libraries and we have seen that before.

Cheers,
Simon


> Thanks again for your help,
> John
> 
>> -----Original Message-----
>> From: Simon Urbanek [mailto:urbanek at research.att.com]
>> Sent: Thursday, September 19, 2013 9:25 AM
>> To: John Fox
>> Cc: David Winsemius; Sarah Hardy; r-sig-mac
>> Subject: Re: [R-SIG-Mac] Problems loading Rcmdr on a Mac 10.7.2 and a
>> Mac 10.6.8
>> 
>> On Sep 19, 2013, at 9:05 AM, John Fox wrote:
>> 
>>> Dear David,
>>> 
>>> On Thu, 19 Sep 2013 01:04:18 -0700
>>> David Winsemius <dwinsemius at comcast.net> wrote:
>>>> 
>>>> On Sep 18, 2013, at 8:47 PM, John Fox wrote:
>>>> 
>>> 
>>> . . .
>>> 
>>>>> It also seems apparent from the error message that there's a
>> mismatch between the version of Tcl/Tk, presumably the one supplied
>> with the R distribution, and the version of X-Windows.
>>>> 
>>>> That is not what I am reading. libfreetype.6 version 13 is the
>> reported problem with  a need to update to version 14.
>>>> 
>>> 
>>> Do you know where libfreetype.6 comes from? Is it installed
>> independently of Tcl/Tk and XQuartz? If so, do you know how to update
>> it?
>>> 
>>> . . .
>>> 
>>>> I'm not really a student but maybe this will be useful anyway. There
>> is a version of Tk at:
>>>> 
>>>> http://r.research.att.com/src/tk8.6.0-src.tar.gz
>>>> 
>>>> ... which appears to be the most up-to-date version. I'm actually
>> running version 8.5 with OSX 10.7.5
>>>> 
>>> 
>>> Version 8.6-0 of Tcl/Tk for X-Windows is included in the R 3.0.1
>> installer for Mac OS X. Is it possible that it doesn't get installed if
>> there's already a Tcl/Tk for X-Windows installed?
>>> 
>>> . . .
>>> 
>>>> 
>>>> The error message below says that libfreetype.6 is an old version
>> and that you need version 14.0 for libtk8.6
>>>> 
>>>> From Unix cmd line:
>>>> 
>>>> otool -L /opt/local/lib/libfreetype.6.dylib
>>>> 
>>>> # ---- reports ___________
>>>> /opt/local/lib/libfreetype.6.dylib:
>>>> 	/opt/local/lib/libfreetype.6.dylib (compatibility version 16.0.0,
>> current version 16.0.0)
>>>> 	/opt/local/lib/libz.1.dylib (compatibility version 1.0.0, current
>> version 1.2.7)
>>>> 	/opt/local/lib/libbz2.1.0.dylib (compatibility version 1.0.0,
>> current version 1.0.6)
>>>> 	/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current
>> version 125.2.11)
>>>> 
>>> 
>>> So tcltk and the Rcmdr 2.0-0 work on your system because you have a
>> sufficiently up-to-date libfreetype.6. It would be additionally helpful
>> to be able to answer the following questions:
>>> 
>>> (1) Why is it that some users have old versions of libfreetype.6.
>>> 
>> 
>> guess: they don't use Software Update and thus have outdated system
>> libraries
>> 
>> 
>>> (2) Why doesn't installing the newest versions of R (which includes
>> Tcl/Tk) and XQuartz, not solve this problem?
>>> 
>> 
>> guess: they are running an older version or didn't install R properly
>> from the CRAN package
>> 
>> 
>>> (3) What can users experiencing this problem do to fix it?
>>> 
>> 
>> 
>> Run Software Update for (1) , re-install R for (2)?
>> 
>> We have only 3rd hand report here so we're reduced to guessing so far.
>> As I said, one important thing is that we uses system X11 and not
>> XQuartz so installing XQuartz has no effect if there is a system X11
>> installed. However, on an up-todate system there should be no problems.
>> 
>> Cheers,
>> S
>> 
>> 
>>> If I knew the answers to these questions, I could update the Rcmdr
>> installation notes to help users avoid the problem.
>>> 
>>> Thank you for your help,
>>> John
>