Re: Assert Levels

greg

On 19 Sep 2008, at 20:13, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Greg Stark <greg(dot)stark(at)enterprisedb(dot)com> writes:
>> You'll also have to do enough empirical tests to convince people that
>> a --enable-cheap-casserts build really does perform the same as a
>> regular build.
>
> I don't think performance is even the main issue. We have never
> recommended having Asserts on in production because they can decrease
> system-wide reliability. As per the fine manual:

I think Simon's coming at this from a different angle - at least this
is what I took out of his suggestion: there Might be bugs or "can't
happen" events that can happen but only under heavy load. Currently
anyone running heavy workloads does it with assertions off so we
aren't testing that case.

There was a case of this that the HOT thesting turned up. There was a
"can't happen" assertion failure related to hint bits being lost that
was happening.

This is a good example of why running with assertions enabled on
production might not be a good idea. But it's also a good example of
why we should do our performance testing with assertions enabled if we
can do it without invalidating the results.