Hi,
On 2022-02-17 18:44:27 -0500, Tom Lane wrote:
> (that's visible now on buildfarm member caiman). We probably
> should take some thought for silencing this before it starts
> to be in people's faces during routine development.
Agreed.
One annoying thing I recently encountered, related to this, is that our stack
check fails to work with some -fsanitize* options because they end up using
multiple stacks (e.g. -fsanitize-address-use-after-scope). Not sure we can
really do anything about that...
> Fixing this is a bit harder than one could wish because we export
> set_stack_base() for use by PL/Java, so it would be better to not
> change that API. I ended up with the attached patch, which works
> to silence the warning so long as the new subroutine
> set_stack_base_from() is marked pg_noinline. I'm a little worried
> that in a year or two GCC will be smart enough to complain anyway.
> If that happens, we could probably silence the warning again by
> moving set_stack_base() to a different source file --- but at some
> point we might have to give up and change its API, I suppose.
We could try using __builtin_frame_address(0) when available, presumably gcc
won't warn about that...
Greetings,
Andres Freund