One last time: annoying warning

[Richard B. Kreckel]

Hi again,

On Thu, 26 Sep 2002, Roberto Bagnara wrote:
[...]
> You are violating [17.4.3.1.2], for instance.

LOL!  I was actually anticipating this to come up.  Anything else, Sir?

[...]
> a) writing header files so as to suit the needs of as many users
>     as possible, and

Exactly.  If those leading underscores pose a problem in any realistic
compilation environment I will happily patch it.  Until then there is
nothing to worry about.

> b) ask for and provide clearly specified interfaces

Good.  So let me next specify `lsolve()' and it's relation to the matrix
class.

Just some thoughts: that whole notion of "clearly specified interfaces"
is not always entirely clear.  As soon as one allows general expressions
lie `x+sin(x)+cos(2*x)' one opens a pandora's box of controversial
evaluations, both anonymous and named ones.  It is not always easy to
clearly specify the intent since that appears clearly as people work with
the system and not earlier.  Remember that `sqrfree()' discussion and the
change in behaviour between GiNaC-1.0.0 and GiNaC-1.0.1?  As a developer,
I find a set of regression tests more useful than a set of formal
specifications since a problem is much easier trapped there.  I hope you
do the same.

However: what is unclear about the notion of polynomials?

Regards
     -richy.
-- 
Richard B. Kreckel
<Richard.Kreckel at GiNaC.DE>
<http://www.ginac.de/~kreckel/>