Re: Building R-2.6.0 fails on macppc

[jim smith <n0oct%sbcglobal.net@localhost>]

On Sat, 9 Feb 2008 00:44:08 +0100
Hernani Marques Madeira <hernani%vecirex.net@localhost> wrote:

> On Fri, Feb 08, 2008 at 12:45:57PM -0600, jim smith wrote:
> 
> > Apparently that last bit, "relocation truncated to fit:" is the key.
> > The compiler [as I understand it], can't handle it on this platform.
> > One also cannot add "makeoptions CFLAGS="-mlong-calls" to the kernel
> > conf and rebuild, as the -mlong-calls is not an option for the compiler
> > on the macppc platform.
> > 
> > Anybody had any luck getting around this for any packages?
> 
> Can you try using CFLAGS+=-fPIC in your Makefile, after cleaning, for
> testing?
> 
> gcc(1) at me says:
>       [...]
>        -shared
>            Produce a shared object which can then be linked with other objects
>            to form an executable.  Not all systems support this option.  For
>            predictable results, you must also specify the same set of options
>            that were used to generate code (-fpic, -fPIC, or model suboptions)
>            when you specify this option.[1]
>       [...]
> 
> To exclude that the problem is not of general nature I compiled R-2.6.0 on 
> two 
> platforms: i386 (NetBSD v3.1.1_PATCH) and amd64 (NetBSD v4.99.52) -- there 
> anything went well.
> 
> --
> hernani
> 

That did the trick!  Thanks!  And, if anyone wants it, I'm building the
binary package of this for macppc.

-- 
jim smith <n0oct%sbcglobal.net@localhost>