[R-SIG-Mac] Possibly fixed R GUI, please test

[Simon Urbanek]

Saptarshi,

On May 2, 2006, at 2:07 PM, Saptarshi Guha wrote:

> Thank you for the fix - it appears to be working so far. May I know  
> what the problem was?

Sure: Info.plist in the GUI used to set DYLD_LIBRARY_PATH to $R_HOME/ 
lib for historical reasons - this was necessary before we started  
linking libR.dylib with the full path (AFAIR that was R 2.0.0 or  
earlier). R 2.3.0 now comes with custom gcc libraries (from gcc  
4.0.3) thus setting DYLD_LIBRARY_PATH overrides the default search  
path so the corresponding system libraries are no longer used even if  
other applications require them.
My tests didn't show it, because it doesn't hurt a vanilla system -  
R.app works and there are no other modules, but if the user installed  
some extra modules that are loaded at run-time (e.g. input helpers,  
haxies etc.), those require the system version of the libraries, so  
they fail, sometimes even killing R.app as well.

Cheers,
Simon


> On May 2, 2006, at 1:58 PM, Simon Urbanek wrote:
>
>> To all users that had problems with the GUI and R 2.3.0 (error  
>> messages in the console or 'hangs' at startup), please try the  
>> current nightly build of the GUI from
>> http://R.research.att.com/
>> and send me your feedback. The full URL is
>> http://R.research.att.com/R-GUI-3114-2.3-Deployment.dmg
>> This should also fix problems for people that used 3rd party  
>> haxies and plugins.
>>
>> Thanks,
>> Simon
>>
>> On Apr 27, 2006, at 11:40 AM, Matthieu Dubois wrote:
>>
>>> I worked with R-2.2.1. I tried to install R-2.3.0 and R.GUI-1.15  
>>> from a binary, available at CRAN, without removing the previous R- 
>>> version. When double-clicking on the R.app icon, the R.GUI  
>>> launched but tried unsuccessfully to start R. After few minutes,  
>>> I need to force the R.GUI to stop. However, R-2.3.0 was  
>>> successfully installed and easily started in the Terminal. What  
>>> can I do ? Any idea would be helpful.
>>
>
>