Fix test_predtest's idea of what weak refutation means.
authorTom Lane <tgl@sss.pgh.pa.us>
Fri, 9 Mar 2018 00:44:14 +0000 (19:44 -0500)
committerTom Lane <tgl@sss.pgh.pa.us>
Fri, 9 Mar 2018 00:44:23 +0000 (19:44 -0500)
commita63c3274a68832182f84ca5d2b8cc5a7462bdacb
treea8682e07ebbcf53dea9171ef9096552dae3adb8c
parent960df2a9715c5c232b0f6f3effd40a3323eed6b0
Fix test_predtest's idea of what weak refutation means.

I'd initially supposed that predicate_refuted_by(..., true) ought to
say that "A refutes B" means "non-falsity of A implies non-truth of B".
But it seems better to define it as "truth of A implies non-truth of B".
This is more useful in the current system, slightly easier to prove,
and in closer correspondence to the existing code behavior.

With this change, test_predtest no longer claims that any existing
test cases show false proof reports, though there still are cases
where we could prove something and fail to.

Discussion: https://postgr.es/m/5983.1520487191@sss.pgh.pa.us
src/test/modules/test_predtest/expected/test_predtest.out
src/test/modules/test_predtest/test_predtest.c