Re: [bug?] Missed parallel safety checks, and wrong parallel safety

On Sat, Apr 24, 2021 at 12:53 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Fri, Apr 23, 2021 at 6:45 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> >
> > Greg Nancarrow <gregn4422(at)gmail(dot)com> writes:
> > > I'm curious. The FmgrBuiltin struct includes the "strict" flag, so
> > > that would "lock down the value" of the strict flag, wouldn't it?
> >
> > It does, but that's much more directly a property of the function's
> > C code than parallel-safety is.
> >
>
> Isn't parallel safety also the C code property? I mean unless someone
> changes the built-in function code, changing that property would be
> dangerous. The other thing is even if a user is allowed to change one
> function's property, how will they know which other functions are
> called by that function and whether they are parallel-safe or not. For
> example, say if the user wants to change the parallel safe property of
> a built-in function brin_summarize_new_values, unless she changes its
> code and the functions called by it like brin_summarize_range, it
> would be dangerous. So, isn't it better to disallow changing parallel
> safety for built-in functions?
>
> Also, if the strict property of built-in functions is fixed
> internally, why we allow users to change it and is that of any help?
>

Yes, I'd like to know too.
I think it would make more sense to disallow changing properties like
strict/parallel-safety on built-in functions.
Also, with sufficient privileges, a built-in function can be
redefined, yet the original function (whose info is cached in
FmgrBuiltins[], from build-time) is always invoked, not the
newly-defined version.

Regards,
Greg Nancarrow
Fujitsu Australia