On Dec 19, 2013, at 5:44 PM, Satish Balay <[email protected]> wrote:

> On Thu, 19 Dec 2013, Barry Smith wrote:
> 
>> 
>> On Dec 19, 2013, at 1:52 PM, Matthew Knepley <[email protected]> wrote:
>> 
>>> is there a good reason not to add rules for all the fortran stubs with the 
>>> right dependencies, and call bfort individually?
>> 
>>   Can all the generated files also go in ${PETSC_ARCH} instead of within the 
>> source code so that I can have different branches with different arches and 
>> switch between them without having to regenerate (and remove) the fortran 
>> stubs appropriate for that particular arch — branch.
> 
> One issue is distributing and compiling sowing.
> 
> For ex: on windows it currently builds with g++ [not MS C]. And on big
> machines [and batch systems] we don't want to be building it with
> mpicc.

   Valid point. Perhaps they compiled stubs could be included somewhere in the 
tar ball and copied over to the PETSC_ARCH directory when using the tar ball 
and only generated when using git.

   Barry

> 
> Satish

Reply via email to