Re: [HACKERS] MergeJoin and Mark/Restr scan positions (Re: [PORTS] Port Bug Report)

Bruce Momjian wrote:
>
> > BTW, elog(FATAL) from vacuum and btree were reported by 2-3
> > another ppl - shouldn't we ask users to re-compile server with
> > CASSERT in all "FATAL" cases ?
> > CASSERT assisted me very much to find real sources of this problem
> > and recently fixed vacuum bug.
>
> I am a little confused on when we should have them use this.  In most
> cases, turning on CASSERT doesn't help.  Should we suggest it in certain
> cases as we see it applies, or is there a rule we can give them to tell
> them when to try it?

For me, CASSERT is last chance to put me in right direction when
I can't reproduce reported bug (user can't provide test data etc).
But you're right - we shouldn't suggest anything generaly.

BTW, Alvin got ASSERTion for real data! We'll try to fix this...

Vadim