"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> I noticed that the standard seems (if I'm reading it correctly) to
> allow subtransactions to switch to more restrictive settings for
> both transaction isolation and read only status than the enclosing
> transaction, but not looser.
Yeah. My recollection is that we've discussed exactly this point with
respect to isolation level, and decided that we couldn't (or it wasn't
worthwhile to) allow serializable subxacts inside repeatable read.
I don't know whether your patch will change that tradeoff. But I don't
think it's really been discussed much with respect to read-only, perhaps
because nobody's paid all that much attention to read-only at all.
In any case, the behavior you state seems obviously correct, so let's
see what we can do about getting closer to that.
My guess is that a reasonable fix is to remember the read-only setting
as of snapshot lockdown, and thereafter to allow changing from
read-write to read-only but not vice versa. One thing to watch for
is allowing subxact exit to restore the previous read-write state.
(BTW it looks like assign_XactIsoLevel emits a rather useless debug
message in that case, so that code could stand some cleanup too. Also,
that code is set so that it will throw an error even if you're assigning
the currently active setting, which maybe is overly strict? Not sure.)
> I can fix up the patch if to support it again if you like. (I think
> it's just a matter of replacing a few lines that I replaced in the
> original patch.) If you'd rather do it, I'll stay out of your way.
Feel free to have a go at it.
regards, tom lane