Re: Weird irreproducible behaviors in plpython tests

On 2016-04-10 17:10:57 -0400, Tom Lane wrote:
> After failing at that, it occurred to me to try it under valgrind,
> and kaboom!  I found a *different* bug, which has apparently been
> there a long time.  (I say different because I don't see how this
> one could produce tick's symptoms; it's a reference to already-freed
> strings, but not an attempt to pfree one.)  I'll be fixing this one
> shortly, but now we have another puzzle: why isn't buildfarm member
> skink seeing the same failure?  It is running the plpython tests.
> Can anyone else reproduce a failure by valgrind'ing the plpython
> tests?  It looks here like

I possible that it's hidden by the broad python error suppressions I
added to valgrind.supp, and the reason they're showing for you is that
you don't appear to have python debugging symbols. But I don't exactly
see the error matching up...

- Andres