Michael Paquier <mich...@paquier.xyz> writes:
> On Fri, Jun 17, 2022 at 09:03:23PM +0200, Peter Eisentraut wrote:
>> I'm pretty sure PostgreSQL code already depends on this behavior anyway.
>> The code just isn't consistent about it.
> In the frontend, I'd like to think that you are right and that we have
> already some places doing that.
We do, indeed.
> The backend is a different story,
> like in GetMemoryChunkContext(). That should be easy enough to check
> with some LD_PRELOAD wizardry, at least.
Huh? The proposal is to accept the fact that free() tolerates NULL,
and then maybe make pfree() tolerate it as well. I don't think that
that needs to encompass any other functions.
regards, tom lane
