>> I do not think that it is really worth fixing, but I will not prevent anyone
>> to fix it.
>
> I still think it does. Well, if there is consensus to address this one
> and optionally the other integer overflows even on back branches, I'll
> write a patch and let's call that a deal. This is not a problem from
> my side.
My point is just about the cost-benefit of fixing a low probability issue
that you can only encounter if you are looking for it, and not with any
reasonable bench script.
Now adding somewhere a test might just help closing the subject and
do more useful things, so that would also be a win.
/* these would raise an arithmetic error */ if (lval == INT64_MIN && rval == -1) { fprintf(stderr, "cannot
divideor modulo INT64_MIN by -1\n"); return false; }
This may be backpatched to old supported versions.
--
Fabien.
