Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
> I came to the same conclusion (that the change in TuplesortMethod
> definiton is the culprit) a while ago and was about to push a fix that
> initialized it correctly in ExecSortInitializeDSM. But I agree reverting
> it back to the old definition is probably better.
Yeah, for the moment. James would like to not have
SORT_TYPE_STILL_IN_PROGRESS be part of the enum at all, I think,
and I can see his point --- but then we need some out-of-band
representation of "worker not done", so I'm not sure there'll be
any net reduction of cruft. Anyway that can be dealt with after
we have a stable buildfarm.
Note also that there's a separate comment-only patch in
<CAAaqYe9qzKbxCvSp3dfLkuS1v8KKnB7kW3z-hZ2jnAQaveSm8w@mail.gmail.com>
that shouldn't be forgotten about.
regards, tom lane