[SciPy-user] Build regression: return of restFP/saveFP problems on OS X?

[David M. Cooke]

Robert Kern <robert.kern at gmail.com> writes:

> David M. Cooke wrote:
>> Robert Kern <robert.kern at gmail.com> writes:
>> 
>>>Zachary Pincus wrote:
>>>
>>>>Hi folks,
>>>>
>>>>I've been building svn versions of scipy just fine for some months  
>>>>now, following Chris Fonnesbeck's instructions on the wiki. (Short  
>>>>version: use gcc3.3, get g77 from the hpc repository.)
>>>>
>>>>Today I updated to the latest svn scipy, and I now cannot get scipy  
>>>>to compile because of link errors saying that restFP and saveFP are  
>>>>not defined. These problems used to be endemic with scipy and OS X,  
>>>>but they were fixed some time ago. It looks like they have been re- 
>>>>introduced recently, though.
>>>
>>>Uncomment lines 114-115 in numpy/distutils/fcompiler/gnu.py .
>>>
>>>David (Cooke), can you explain why you commented them out?
>> 
>> Whoops. I was testing whether I could build using gfortran 4.1.
>> I compiled gcc 4.1 using darwinports, which doesn't support
>> cc_dynamic. As far as I can tell, there's not actually a library by
>> that name; I think the linker does something special when it finds it?
>
> [~]$ ls -l /usr/lib/libcc_dynamic.a
> lrwxr-xr-x   1 root  wheel  27 Nov  7 18:08 /usr/lib/libcc_dynamic.a ->
> gcc/darwin/default/libgcc.a

Ah hah, that symlink is only made when you've got 3.3 selected by
gcc_select (it's not made for gcc 4.0).

-- 
|>|\/|<
/--------------------------------------------------------------------------\
|David M. Cooke                      http://arbutus.physics.mcmaster.ca/dmc/
|cookedm at physics.mcmaster.ca