That’s always been my understanding too. I use asserts liberally to test whether my initial assumptions about “things which should never go wrong” are correct. That includes checking virtually all method/function call return values using ATLVERIFY() or equivalent (which keeps PC-Lint warning 534 out of my hair ;) ).

Any assumptions which turn out to be invalid (or could be influenced by user or system behaviour) are then looked at in more detail and release build error checking added if necessary.

If a failed assumption would cause a crash (e.g. an unguarded pointer or a divide by zero) I will also usually back up the assert with a following conditional – call me paranoid, if you like, but I’d rather be sure that an unguarded pointer etc. I’ve missed isn’t going to cause a crash at runtime.

FWIW I’m hoping to do a session on PC-Lint at next year’s ACCU Conference. :)

The argument for putting in extra code is – we don’t always do everything right, and a reasonable (but incorrect in some minor way perhaps) result or recovery may be better than a crash or a wild result.

Obviously circumstances vary.

If the circle function is in a screensaver or a game background graphic, drawing a million circles, the consequences of an incorrect radius circle are minor, and a badly drawn circle is better than a crash.

If the circle function is calculating something your life depends on, maybe a recovery to an incorrect result is worse than a crash.

And so on…

I watched a very moving documentary a while back about the battle of Prokhorovka (part of Kursk). I think it was a BBC ‘Timewatch’ documentary. Well worth watching. Especially when the bits where they interviewed the veterans.

e.g.

void Circle::setRadius( double radius )
{
assert( radius > 0.0 );

// if radius <= 0.0 then program must be badly fubar’ed
// should I try to recover, or just continue in release?
if ( radius <= 0.0 )
{
m_radius = 1.0 ; // reasonable value that won’t crash the program
Log( “Bad Radius…”, etc… ) ;
return ;
}

m_radius = radius;
}

BTW, I have a site about the Battle of Kursk. Only has a couple of real pages on it, because I never got round to adding more, but if interested it’s

http://www.operationcitadel.com/