[R-SIG-Mac] Fwd: os x / g95 package namespace load problem

[Simon Urbanek]

Joel,

On Jun 23, 2005, at 3:39 PM, Joel Bremson wrote:

> I have a working package for linux, including fortran 95 code  
> compiled with g95, that I need to port to OS X. The package works  
> on Linux and seems to load on the Mac,but when I try to run a  
> function that calls C or Fortran I'm told that the symbol is not  
> loaded.

A high-level question first: what are you trying to do here? (I'm  
assuming you're the author of the package - if not, send all this to  
the author ;)). Using a fixed Makefile is a really bad idea, because  
you lose all the flags that R configuration conveniently detected for  
you, so that package is likely to run only on one specific system and  
nowhere else. In most cases you don't need any Makefile at all,  
because R supplies all that is needed. You can still specify any  
additional flags or dependencies using Makevars. Finally, even if you  
have a very complex setting, it's easier to use configure and  
Makevars than building the entire Makefile (which involves using  
autoconf anyway). Actually from the Makefile you provided it seems to  
me that it's entirely superfluous, because you use no complex  
settings at all.

> I was told to use the .dylib extension instead of .so

Who told you that? If you used the regular process for compiling  
package code that R provides, you'd see that libraries in packages  
still use .so, because they are loaded using dlopen so the name  
doesn't have to conform to OS' conventions.

Please consult 'Writing R Extensions' ( http://cran.r-project.org/doc/ 
manuals/R-exts.html ) - it provides guidance and examples. Is also  
says "Note that this mechanism [using Makevars] should be general  
enough to eliminate the need for a package-specific Makefile. If such  
a file is to be distributed, considerable care is needed to make it  
general enough to work on all R platforms".

Cheers,
Simon