Hi Andres,
Sorry for such a late reply.
On Tue, Feb 13, 2024 at 2:14 PM Andres Freund <
andres@anarazel.de> wrote:
> Random, mildly related thought: I wonder if it's time to, again, look at
> enabling -ftrapv in assert enabled builds.I had looked at that a few years
> back, and fixed a number of instances, but not all I think. But I think we are
> a lot closer to avoiding signed overflows everywhere, and it'd be nice to find
> overflow hazards more easily.
I agree that this would be very helpful.
> Many places are broken even with -fwrapv
> semantics (which we don't have on all compilers!). Trapping on such overflows
> makes it far easier to find problems with tools like sqlsmith.
Does this mean that some of our existing tests will panic when compiled
with -ftrapv or -fwrapv? If so I'd be interested in resolving the
remaining issues if you could point me in the right direction of how to
set the flag.
Thanks,
Joe Koshakow