patch for consideration

## Sorry - just realized I replied to the last message off-list ##

OK, I see what's going on. R CMD is setting DYLD_LIBRARY_PATH but  
after Python is already running so nothing happens.

Looks like I'l be setting those paths by hand.

Thanks for your help,
Kjell

On Sep 30, 2009, at 6:10 PM, Simon Urbanek wrote:

On Sep 30, 2009, at 11:25 , Konis Kjell wrote:

What I am trying to do is get the full path to libR.dylib encoded in
package shared objects so they can be loaded with Rpy.  Would it be
best to just do this by hand after installing R?

Yes - that's what the framework version does. You can only do that
after installation. However, any embedding application has to run R
CMD anyway so it should not matter since that sets DYLD_LIBRARY_PATH
so the path in the binary is irrelevant ...

Cheers,
Simon

Kjell

On 30 sept. 2009, at 17:05, Simon Urbanek wrote:

On Sep 30, 2009, at 10:34 , Konis Kjell wrote:

Hello,

The attached file patches the configure script so that the full
install path is used in libR.dylib when doing a non-Framework  
build.
I  have tested it under 10.5 and 10.6 using the FSF version 4.3
compiler  family. Please feel free to include it in R if you think
it would be  useful.

What is the point of this exercise? R is relocatable until  
installed,
so you cannot embed any paths and configure time and the above  
patch
won't work (also try make install rhome=/foo and see it break).  
To be
precise R won't even care because DYLD_LIBRARY_PATH overrides the
path
but then the question is why to embed it at all since it is
incorrect?

Cheers,
Simon

On Sep 30, 2009, at 12:57 , Kjell Konis wrote:

## Sorry - just realized I replied to the last message off-list ##

OK, I see what's going on. R CMD is setting DYLD_LIBRARY_PATH but
after Python is already running so nothing happens.

That shouldn't be a problem - DYLD_LIBRARY_PATH is dynamic on Darwin
(unlike LD_LIBRARY_PATH on some systems) so changes *do* become active
immediately.

Are you sure you're not setting it in a different
environment? This seems more like an issue in Rpy, not in R ...

Cheers,
Simon

(I should add that this all is only an issue when not using the
framework which is much easier since you can query everything from the
bundle which is why R.app can work even without R CMD)

Looks like I'l be setting those paths by hand.

Thanks for your help,
Kjell

On Sep 30, 2009, at 6:10 PM, Simon Urbanek wrote:

On Sep 30, 2009, at 11:25 , Konis Kjell wrote:

What I am trying to do is get the full path to libR.dylib  
encoded in
package shared objects so they can be loaded with Rpy.  Would it be
best to just do this by hand after installing R?

Yes - that's what the framework version does. You can only do that
after installation. However, any embedding application has to run R
CMD anyway so it should not matter since that sets DYLD_LIBRARY_PATH
so the path in the binary is irrelevant ...

Cheers,
Simon

Kjell

On 30 sept. 2009, at 17:05, Simon Urbanek wrote:

On Sep 30, 2009, at 10:34 , Konis Kjell wrote:

Hello,

The attached file patches the configure script so that the full
install path is used in libR.dylib when doing a non-Framework
build.
I  have tested it under 10.5 and 10.6 using the FSF version 4.3
compiler  family. Please feel free to include it in R if you  
think
it would be  useful.

What is the point of this exercise? R is relocatable until
installed,
so you cannot embed any paths and configure time and the above
patch
won't work (also try make install rhome=/foo and see it break).
To be
precise R won't even care because DYLD_LIBRARY_PATH overrides the
path
but then the question is why to embed it at all since it is
incorrect?

Cheers,
Simon