Function call, postfix inc and exception behavior

[Goran Sliskovic]

Hi,
I have following piece of code:

#include <iostream>
using namespace std;

void f(int i) {
throw (int)0;
}

int main()
{
int i = 0;
try {
f(i++);
}
catch(int x) {
}
cout << i << endl;
return 0;
}

Using VC7.1 I get 1 as a result in debug mode, 0 in release. I thought this
was defined behaviour as i++ should be evaluated prior to function call and
have all side-effects completed. Am I missing something (exception thrown
makes this undefined?) or is it a bug in the compiler?

Regards,
Goran

[Victor Bazarov]

Re: Function call, postfix inc and exception behavior

"Goran Sliskovic" <gsliskov@yahoo .com> wrote...[color=blue]
> I have following piece of code:
>
> #include <iostream>
> using namespace std;
>
> void f(int i) {
> throw (int)0;
> }
>
> int main()
> {
> int i = 0;
> try {
> f(i++);
> }
> catch(int x) {
> }
> cout << i << endl;
> return 0;
> }
>
> Using VC7.1 I get 1 as a result in debug mode, 0 in release. I thought[/color]
this[color=blue]
> was defined behaviour as i++ should be evaluated prior to function call[/color]
and[color=blue]
> have all side-effects completed. Am I missing something (exception thrown
> makes this undefined?) or is it a bug in the compiler?[/color]

Looks like a bug. Probably an optimization issue, since you get
different behaviour in Debug and Release mode.

I've added m.p.vc.lang to the list of newsgroups, perhaps somebody
at MSoft will see it.

Victor

[Goran Sliskovic]

Re: Function call, postfix inc and exception behavior


"Victor Bazarov" <v.Abazarov@com Acast.net> wrote in message
news:nRrpc.5447 1$iF6.4920610@a ttbi_s02...[color=blue]
> "Goran Sliskovic" <gsliskov@yahoo .com> wrote...[color=green]
> > I have following piece of code:
> >
> > #include <iostream>
> > using namespace std;
> >
> > void f(int i) {
> > throw (int)0;
> > }[/color][/color]
....[color=blue][color=green]
> > Using VC7.1 I get 1 as a result in debug mode, 0 in release. I thought[/color]
> this[color=green]
> > was defined behaviour as i++ should be evaluated prior to function call[/color]
> and[color=green]
> > have all side-effects completed. Am I missing something (exception[/color][/color]
thrown[color=blue][color=green]
> > makes this undefined?) or is it a bug in the compiler?[/color]
>
> Looks like a bug. Probably an optimization issue, since you get
> different behaviour in Debug and Release mode.
>
> I've added m.p.vc.lang to the list of newsgroups, perhaps somebody
> at MSoft will see it.
>
> Victor
>[/color]

Thnx, I have found out it is a bug, and also already known one. So I guess
my understanding of standard was ok.

Regards,
Goran