Re: [Gm2] no overflow checking ?

[Gaius Mulley]

"Fischlin  Andreas" <address@hidden> writes:

> Ok, but AFAIK that does not mean one could not have a compiler option
> which overrules that exception raising. I would implement such a
> standard then by having by default the checking on, but a good
> compiler would allow for full control in the manner I described
> earlier. Besides, the rules for raising an exception for integer
> overflow or floating point overflow or underflow, respectively, should
> be differentiated and not dealt with in one sweep. ;-)
>
> Regards,
> Andreas

sure, this is the way to go.  ISO also provides source code directives
which will be used <* *>

to turn on/off aspects of range checking.  I propose that if -fiso is
used then all ISO checking will be on by default, whereas if -fpim is
used this may not necessarily be the case (although for simplicity sake
it might be better to enable by default).  Certainly there will be
command line options to disable/enable all aspects of checking (as just
the integer checking alone will be expensive - too expensive for some
embedded targets)

regards,
Gaius