RE: 2.2.9 rel eng breaks limit checks (ITS#3085)

[ando@sys-net.it]

> More limits problems:
>
> Looking at limits.c lines 909-922 it appears that
> 	a) if the hard tlimit is set to -1 (unlimited)
> and	b) a search comes with tlimit set
> then the tlimit will be set to unlimited, instead of whatever value was
> in the search request. The else from 918-920 should leave op->ors_tlimit
> alone. Actually, the condition should just be combined into the test on
> line 922.

I see there is some issue left.  I'm fixing all of them,
and adding a statement about how things are supposed to
work from an implementor's viewpoint.  I regret to admit
this code changed too much and too qickly, so my main
concern now is to release something that work.  By designing
a comprehensive test I hope this will not happen again.

>
> I think we need a verbal recap of the whole limits design, because it's
> not obvious from looking at this code what is intended, or whether the
> code does what is intended. I also think this code got into RelEng too
> soon, it hasn't been shaken down at all yet.

Agree.

Ciao, Ando.

-- 
Pierangelo Masarati
mailto:pierangelo.masarati@sys-net.it