[petsc-dev] Okay, petscmath.h is a complete piece of crap and now its breaking my build

[Matthew Knepley]

On Tue, Mar 15, 2011 at 1:32 PM, Jed Brown <jed at 59a2.org> wrote:

> On Tue, Mar 15, 2011 at 19:22, Matthew Knepley <knepley at gmail.com> wrote:
>
>> Does yours? My Emacs does not follow #if, and it would not know what
>> branch anyway.
>>
>
> Semantic can parse petscconf.h, then it knows. You are familiar with M-x
> cpp-highlight-mode, right?
>
>
>>
>>
>>> How would it be easier to audit whether all possible combinations were
>>> valid if the code was generated?
>>>
>>
>> 1) I would see exactly what I got (I know I can do this by hand with -E)
>
>
> But only for one configuration at a time, the preprocessor statements would
> be simple if there were a small number of possible configurations. It is
> nested deeply because there is a combinatorial number of possibilities. How
> is auditing that faster with code generation. In particular, when you change
> the generator, how do you confirm that all possible combinations still
> result in correct code?
>

Verification of preprocess code is exactly as intensive, since you have to
compile each version. What is your point here?


>
>> 2) I could look at the configure log to see what decisions were made. This
>> is difficult with the #if model.
>>
>
> LOL.
>
> > 3) The problem turns out to be silent failure due to the need to
> reconfigure. This would not happen in the
>     generated model.
>
> If everything was generated, you would always need to reconfigure. Why
> don't you just always reconfigure now?
>

The code I had would have continued to work, rather than break.

  Matt

-- 
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110315/16446732/attachment.html>