Amit Kapila <amit.kapila16@gmail.com> writes:
> Yes, it was failing only on windows. Today when I further tried to
> look into the issue, I found that if I rebuild plpgsql, it didn't occurred.
> So the conclusion is that it occurred due to build mismatch, however I
> am not sure why a rebuild of plpgsql is required for this patch.
> Sorry for the noise.
plpgsql contains references to PGC_SUSET and PGC_USERSET, whose values are
changed by this patch, so failure is unsurprising if you don't rebuild it.
(I saw failures in the regression tests even without Windows until I
updated plpgsql.)
Committed with minor cosmetic adjustments; mostly, rewriting some
comments.
regards, tom lane