Weekly MacOS X binaries for gfortran!

[Andrew Pinski]

On Dec 23, 2004, at 8:35 AM, FX Coudert wrote:

>> no success here (iBook G3 600 MHz) compiling a simple definition file 
>> from the netCDF distro.
>>  >gfortran -c -O2  typeSizes.f90
>> /var/tmp//ccsgchto.s:1:Unknown pseudo-op: .machine
>> /var/tmp//ccsgchto.s:1:Rest of line ignored. 1st junk character 
>> valued 112 (p).
>> /var/tmp//ccsgchto.s:8:Unknown pseudo-op: .subsections_via_symbols
>> gmake: *** [typeSizes.o] Error 1
>
> OK. I had the same problem when compiling gfortran, but I thought it 
> would not be present at runtime.
> It is fixed by installing an up-to-date version of cctools : 
> ftp://gcc.gnu.org/pub/gcc/infrastructure/cctools-528.5.dmg
>
> To people with knowldedge of MacOSX: is there a way to avoid requiring 
> this? or should we assume that anyone willing to use gfortran will 
> have to install it?

Yes by reverting two (or is it three) patches by that means you 
don'tget some of the goodies associated with the updated cctools and 
GCC.  Like
"dead" function stripping (which is what . subsections_via_symbols is
used for) and you don't have to use -force_cpusubtype_ALL when you are
compiling code which uses some optional instructions.

Thanks,
Andrew Pinski