Please consider making compile-time shift calculation more consistent with...

Superseded by http://www.ti.com/tool/msp430-gcc-opensource

Status: Inactive

Brought to you by: cliechti, diwil, pabigot

Milestone: 4.7.x_development

Status: open

Owner: nobody

Labels: None

Priority: 2

Updated: 2013-12-01

Created: 2013-04-22

Private: No

Following code produces unexpected result (0) for BITS=16:

#define BITS 16
#define MOD(x) ((x) & ((1 << BITS) - 1))

int v = 5;
int main()
{
    printf("%d\n", MOD(v));
}

Despite shifts to values bigger than bit size of type are undefined per C standard, there're:

1) Arithmetical definition of shift operation;
2) Behavior of other well-known gcc ports (x86);
3) User expectations;
4) Principle of least surprise.

So, it would be nice to change compile-time (and only compile-time!) behavior to more consistent with mathematical definition.

This is obviously not a priority request.