Re: BUG #5592: list of integer undefined behaviors

On 8/3/2010 4:08 PM, Tom Lane wrote:
> However, most of the cases that seem of interest so far are in fairly
> small, stable functions. Would it be reasonable to attach a "checked
> for overflow problems" label to these functions as a whole?

This should work great. I'll get my clang hacker to start working on it.

I see that pgsql has support already in place to def out attributes when
non-GCC ocmpilers are used.

Just to be clear we're talking about putting something like this in your
header files:

int
bms_first_member(Bitmapset *a)
__attribute__((no_integer_overflow_checks));

Pick whatever name you like for this attribute, it doesn't matter to us.

John