[petsc-dev] BuildSystem: PCC=mpicc, CXX and FC

[Barry Smith]

On Sep 14, 2010, at 11:03 PM, Satish Balay wrote:

> I don't remember previous conversation regarding it. However wrt
> python I strongly felt the python used by configure should be used by
> 'full path' in generated scripts. [for eg check: reconfigure.py]
> 
> I'm fine with using full paths for default searches done by configure
> [cc,mpicc etc from PATH]. However if user provides 'cc=gcc' - this
> should not be expanded to full PATH automatically by configure..

   How come?

   Barry

> 
> Satish
> 
> On Tue, 14 Sep 2010, Barry Smith wrote:
> 
>> 
>> On Sep 14, 2010, at 8:54 PM, Matthew Knepley wrote:
>> 
>>> Do we really want the full path to all compilers? We discussed this before and did not
>>> do it. Satish, do you remember why?
>> 
>>  With the full path I like the fact that the compiler choices are set at at ./configure time and won't mysteriously change (and result in failure) later if someone changes their path or installs a new MPI into the path.
>> 
>>  I do not know why you opted not to always have the full path.
>>> 
>>>   Matt
>>> 
>>> On Tue, Sep 14, 2010 at 8:51 PM, Lisandro Dalcin <dalcinl at gmail.com> wrote:
>>> On 14 September 2010 22:26, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>> 
>>>> On Sep 14, 2010, at 8:24 PM, Lisandro Dalcin wrote:
>>>> 
>>>>> On 14 September 2010 21:55, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>>>> 
>>>>>> On Sep 14, 2010, at 6:24 PM, Lisandro Dalcin wrote:
>>>>>> 
>>>>>>> 1) If mpicc is in $PATH, then petscvariables will get PCC=mpicc.
>>>>>>> Perhaps it should get the full path instead?
>>>>>>> 
>>>>>>> 2) I think CXX and FC variables should be renames to PCXX and PFC. The
>>>>>>> former names can easily conflict with CXX and FC defined in user
>>>>>>> makefiles for use with non-MPI sources.
>>>>>> 
>>>>>>  Then we should also get rid of CC
>>>>>> 
>>>>>>   Essentially you are saying we should remove all "traditional" use of these variables from our makefiles? I agree doing that might be a good idea. But doesn't it contradict your attempt to try hard to conform to standard usage for things?
>>>>>> 
>>>>> 
>>>>> Sorry, I need a clarification: Could you tell me why PETSc makefiles
>>>>> uses PCC instead of the "traditional"  CC? Is it just because PCC
>>>>> could be CC or CXX depending on --with-c-language?
>>>> 
>>>>  Yes that is one reason and may be the only reason.
>>>> 
>>> 
>>> In such case, forget my comments, except for the part of setting
>>> PCC=/path/to/mpicc (and perhaps the same for CC, CXX, FC if they point
>>> to MPI compiler wrappers). What do you think about this? Please note
>>> I'm not objecting the current status, just asking about the full path
>>> alternative.
>>> 
>>> 
>>> --
>>> Lisandro Dalcin
>>> ---------------
>>> CIMEC (INTEC/CONICET-UNL)
>>> Predio CONICET-Santa Fe
>>> Colectora RN 168 Km 472, Paraje El Pozo
>>> Tel: +54-342-4511594 (ext 1011)
>>> Tel/Fax: +54-342-4511169
>>> 
>>> 
>>> 
>>> -- 
>>> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
>>> -- Norbert Wiener
>> 
>> 
>