Message-ID: <C82D9261-EC3C-44F8-8164-A7D6EF072900@unsw.edu.au>
Date: 2005-09-27T01:13:40Z
From: Bill Northcott
Subject: R 2.1.1 tiger compile problem
In-Reply-To: <1253d67a0509261156f6a2db2@mail.gmail.com>
On 27/09/2005, at 4:56 AM, Joel Bremson wrote:
> I was able to build R-2.1.1 on the Tiger/G5 with configure command you
> gave, adding in F77=gfortran but my package will not build
> correctly. It dies
> with a namespace load error during the R CMD CHECK. If I ignore the
> namespace error and just install it the code segfaults.
>
> However, the package does build and run correctly on a Tiger/G4
> machine. The difference is
> that on the G4 we run gcc-4.0 apple build 4061 and gfortran 4.1.0
> v20050517. On the
> G5 we're running gcc-4.0 apple build 5026 and gfortran apple build
> 5202.
>
> Since the code compiles and runs properly on the G4 I'm assuming the
> code is fine. Now we're thinking of running the same compilers on
> the G5.
The gfortran code from a few months back had a number of crucial
bugs, one or more of which is probably your problem. The gfortran
code in apple-gcc-5202 is the gcc-4.0.1 release version. Your
gfortran 4.1.0v20050517 code would probably have less bugs.
I think you could get even better code by waiting a few days for the
gcc-4.0.2 release, which will have a lot of important bug fixes, and
would be much more consistent with the officially released Apple
gcc-5026.
Bill Northcott