Re: Better sanity checking for message length words

Aleksander Alekseev <aleksander(at)timescale(dot)com> writes:
> I'm having slight issues applying your patch to the `master` branch.
> Is it the right target?

[ scratches head ... ] The patch still applies perfectly cleanly
for me, using either "patch" or "git apply".

> Regarding the idea, I think extra checks are a good thing and
> definitely something that can be introduced in the next major version.
> If we receive a complaint during beta-testing we can revert the patch
> or maybe increase the limit for small messages.

Actually, I did some more testing yesterday and found that
"make check-world" passes with PQ_SMALL_MESSAGE_LIMIT values
as small as 16. That may say more about our testing than
anything else --- for example, it implies we're not using long
statement or portal names anywhere. But still, it suggests
that 30000 is between one and two orders of magnitude too large.
I'm now thinking that 10000 would be a good conservative setting,
or we could try 1000 if we want to be a bit more aggressive.
As you say, beta-testing feedback could result in further
modifications.

regards, tom lane