Re: Reject negative max_retention_duration values

On Wed, Jun 10, 2026 at 05:12:26AM +0000, Hayato Kuroda (Fujitsu) wrote:
> Dear Chao,
> 
> > Yes, this patch rejects negative values at CREATE/ALTER SUBSCRIPTION time,
> > so in theory the if (MySubscription->maxretention <= 0) change is not strictly
> > necessary. I made that change for a few reasons (from strong do weak):
> 
> I personal preference is to use Assert() for detecting cannot-happen case,
> but it's not very strong opinion. Let's see how others say.

An assertion offers less protection than an elog(ERROR) if a value is
read from catalogs, which could be the case here?  Think for example
corrupted catalog data.  (I did not read the patch in details, so I
may have missed something, of course, but I was under the impression
that this could apply for this case with MySubscription.) 
--
Michael