On Fri, Jun 11, 2004 at 10:37:21AM +0200, [log in to unmask] wrote:
> I think that's the wrong way to go, because it hides the bug in the code
> from the user.
The object is not to use the workaround all the time. The object is to
use the workaround to find that there is a bug.
> Better all variables should be set to some invalid data (like NaN), unless
> explicitely initialized.
There is no equivalent to NaN for integers. Yes, it's handy to have
NaN initialization for floating point. Personally, I like to be able
to have initialization to a constant, and vary the constant to see if
the program behavior changes.
-- greg
|