R 2.1.1 available in DarwinPorts

R-SIG-MAC is an "R Special Interest Group on Macintosh Development  
and Porting, both for MacOS 8.6 - 9.x and MacOS X"  Why does what  
we're doing not count?
I don't see either of 'Fink' or 'Darwin Ports' in the list {MacOS  
8.6-9.x, MacOS X}.

Once you install Fink or Darwin Ports, you are no longer doing UNIX  
style builds on MacOS X, because you effectively replace many of the  
system libraries and, so it appears with Darwin Ports, the  
compilers.   You are now building with different UNIX system.

Back in the days of MacOS X 10.0 and 10.1, Fink and Darwin Ports were  
arguably useful.  With Tiger, they just get in the way and create  
trouble like the ones you are experiencing.  Almost all the basic GNU/ 
BSD libraries are available on Tiger.  About the only ones now  
missing will never be there. Apple will not ship GPL licensed  
libraries, because they might accidentally pollute a commercial  
developers code.  That is why there is no GNU readline for instance.

The open source libraries that are not included are now trivial to  
build without needing the wholesale changes of Fink or Darwin Ports.

Some one wrote that one needed to install the whole of FSF gcc to get  
g77.  This is not correct.  Everything needed to build R from source  
on Panther or Tiger is included in the Tcl/Tk and g77 packages of the  
R 2.1.1 binary distribution.

Building gfortran from Apple sources on Tiger is not hard.  However,  
it is necessary to include recent patches from the FSF gfortran  
gcc-4.0.x.  You can find an Apple gcc-4 compiler with gfortran which  
will build R on www.swarm.org.  I hope to build a more recent one as  
soon as the fixes for the equivalence problems in gfortran get  
committed to the cvs.  That should be within a few days.

There are those from the Linux camp who think gcc-4 is not such a  
good compiler as the recent gcc-3.x.  As recent tests published by  
Anandtech show, while this may be the case on x86 Linux, it is simply  
not true on MacOS X.

Bill Northcott
I assume the fsf gfortran patches do not wind up in the apple
cvs branch, unless apple explicitly puts them there.

I am also interested if the patched gfortran works with vecLib
and supports its complex blas.

Building gfortran from Apple sources on Tiger is not hard.  However,
it is necessary to include recent patches from the FSF gfortran
gcc-4.0.x.  You can find an Apple gcc-4 compiler with gfortran which
will build R on www.swarm.org.  I hope to build a more recent one as
soon as the fixes for the equivalence problems in gfortran get
committed to the cvs.  That should be within a few days.
==========================================================
Jan de Leeuw, 11667 Steinhoff Rd, Frazier Park, CA 93225, 661-245-1725
              http://www.cuddyvalley.org/dunklen.mov
==========================================================
              Und die Einen sind im Dunkeln
              Und die Andern sind im Licht
              Doch man sieht nur die im Lichte
              Die im Dunkeln sieht man nicht       (Brecht)
I assume the fsf gfortran patches do not wind up in the apple
cvs branch, unless apple explicitly puts them there.
It just gets updated when Apple do their merges with the FSF tree.   
The latest corresponds to the gcc-4.0.1 release.  All I do to get the  
patches is diff four directories (gcc/fortran, gcc/testsuite/ 
gfortran.dg, gcc/testsuite/gfortran.fortran-torture and libgfortran)  
against the current gcc-4_0-branch from the cvs.  The only thing to  
watch is that you don't overwrite the couple of 'APPLE LOCAL'  
modifications.

Then all you have to do to build gfortran on a G5 is to add 'f95' (or  
it may have changed to gfortran) to the list of langauges in  
build_gcc.  Finally when it is all built you need to manually copy  
the fortran runtime library to the distribution tree.   Otherwise  
your code won't run!
I am also interested if the patched gfortran works with vecLib
and supports its complex blas.
That sounds like a good question, but rather beyond my competence to  
answer.

Bill

On 08/09/2005, at 9:45 AM, Kjell Konis wrote:

R-SIG-MAC is an "R Special Interest Group on Macintosh Development
and Porting, both for MacOS 8.6 - 9.x and MacOS X"  Why does what
we're doing not count?

I don't see either of 'Fink' or 'Darwin Ports' in the list {MacOS
8.6-9.x, MacOS X}.

I think that there is no reason why we should not help Kjell to build  
R in his set up provided that we can help and that Kjell follow our  
suggestions (Simon has written several times what's going wrong and  
why... thus why not considering these suggestions?).

As for daily testing of darwinports: r-sig-mac is probably NOT the  
right place, maybe darwinports has it's own list for that purpose. Or  
maybe Kjell can correspond with some testers privately and report  
here success or ask for solutions.

What Simon says about compiler dependencies REALLY matters and it is  
something to take care in particular for binary distributions.

Apple is doing great work in including more and more system  
libraries, and there's no reason why not using these if available.
Apple is also helping other open source project in having their  
products build out-of-the box on OS X for the reason explained by  
Bill in this mail.

stefano
Once you install Fink or Darwin Ports, you are no longer doing UNIX
style builds on MacOS X, because you effectively replace many of the
system libraries and, so it appears with Darwin Ports, the
compilers.   You are now building with different UNIX system.

Back in the days of MacOS X 10.0 and 10.1, Fink and Darwin Ports were
arguably useful.  With Tiger, they just get in the way and create
trouble like the ones you are experiencing.  Almost all the basic GNU/
BSD libraries are available on Tiger.  About the only ones now
missing will never be there. Apple will not ship GPL licensed
libraries, because they might accidentally pollute a commercial
developers code.  That is why there is no GNU readline for instance.

The open source libraries that are not included are now trivial to
build without needing the wholesale changes of Fink or Darwin Ports.

Some one wrote that one needed to install the whole of FSF gcc to get
g77.  This is not correct.  Everything needed to build R from source
on Panther or Tiger is included in the Tcl/Tk and g77 packages of the
R 2.1.1 binary distribution.

Building gfortran from Apple sources on Tiger is not hard.  However,
it is necessary to include recent patches from the FSF gfortran
gcc-4.0.x.  You can find an Apple gcc-4 compiler with gfortran which
will build R on www.swarm.org.  I hope to build a more recent one as
soon as the fixes for the equivalence problems in gfortran get
committed to the cvs.  That should be within a few days.

There are those from the Linux camp who think gcc-4 is not such a
good compiler as the recent gcc-3.x.  As recent tests published by
Anandtech show, while this may be the case on x86 Linux, it is simply
not true on MacOS X.

Bill Northcott

_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac

I assume the fsf gfortran patches do not wind up in the apple
cvs branch, unless apple explicitly puts them there.

I am also interested if the patched gfortran works with vecLib
and supports its complex blas.
As of complex blas, there is a work-around now in the most current R- 
devel. The ABI change in fortran was deliberate so I don't think it  
will go away.

Cheers,
Simon