> > So last I heard you and Jesse were planning on rebuilding
known_resource_types on each compile in Statler. Has that fallen by the
wayside?
>
> Jesse and I tried doing that in our first attempt to fix issue 4656. But
> you objected to the patch (off list apparently--I couldn't find it in an
> e-mail search),
>
IIRC that conversation ended (as we were heading up the stairs on our way
back to the office) with me conceding that 1) it should not be the
performance issue I'd been speculating (because the small amount of in-ram
work required would be swamped by the on-disk portion) and 2) if there
_were_ consequences of the sort I'd been sketching it was indicative of
deeper problems and 3) it wasn't reasonable to object to sound code on the
basis of hypothetical bugs elsewhere. Thus our "<1% of total compilation
time" bound in RFC.
> and the upshot of the discussion you and I had about it was the e-mail
> thread "PL RFC-1: Semantics of autoloaded classes". In that e-mail thread,
> we included the version that Jesse and I had tried as proposed solution I,
> and also included two other proposals. The community consensus was pretty
> much in agreement with proposed solution II, and I've come to agree that
> this is the right approach. So there are AFAIK no plans to rebuild
> known_resource_types on each compile.
>
I was remembering the consensus as:
implement option one as a "one major release" feature,
during which time an implementation of option two as a
warning rather than an error.
but skimming the thread again it looks as if you turned against the option
one part as the thread progressed, which I had missed/forgotten. No one
else seems to have had sustained objection to it, but I don't see any
persistent support either.
But if we're doing option two as the solution and not just a warn-only
prelude, that basically kicks a leg out from under my argument here.
I've got a residual thread-safety squeamishness but I think that's not
significant (basically, you could get a warning from each thread, in the
same sense that you could win several lotteries on the same day or could get
hit by lightning while choking on a lollipop with only one shoe on).
It'll come out once per environment in any case, but that's probably ok
(especially considering that the source may well differ between
environments).
I'm still not thrilled with having it on the type collection (which may just
be native cussedness at this point); it seems like that's the wrong place
for the flag, but I can't really come up with a better idea. The
environment itself maybe?
> I just went back to look at issue 4656 in the ticket database and it looks
> like its state was never updated to reflect the community consensus. I'll
> go update it now.
>
Thanks.
-- M
-----------------------------------------------------------
The power of accurate observation is
commonly called cynicism by those
who have not got it. ~George Bernard Shaw
------------------------------------------------------------
--
You received this message because you are subscribed to the Google Groups
"Puppet Developers" group.
To post to this group, send email to puppet-...@googlegroups.com.
To unsubscribe from this group, send email to
puppet-dev+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-dev?hl=en.
