Hi,
Just a quick update since I've been distracted by several other
things in the last day or so...
----- Original Message -----
> ----- Original Message -----
> > I would be surprised if my change caused this, unless you have an old
> > simple
> > pmda running (the binary would not match the library). More likely to be a
> > latent problem in the simple pmda that this change has exposed.
>
> I'm wondering if its the former issue - I'll do more tests to confirm, but I
> suspect some of the indeterminism I saw might be explained by rebuilding the
> PMDA in-between some runs (and thereafter it was fine).
>
> When I revert the change, the problem is definitely not reproducible anymore,
> so will go with that for now & put it back in dev shortly (post-release).
I've reverted the revert in dev, but the above statements are all still true -
not really understanding how, but some kind of ABI-breaking oddity seems to be
in play here. There's a comment in the pmda libdefs.h about how that structure
is used as "private per PMDA data when multiple DSO PMDAs are in use" - do you
know the background to that? That is our situation here, so that aspect might
be worth a deeper look.
cheers.
--
Nathan
|