Bruce Momjian <[EMAIL PROTECTED]> writes: > Uh, opinions on this? We don't normally test every feature of a > command, do we?
It seemed a tad excessive to me, at least for routine regression testing. What do you think of making it a separate test script and adding it to "make bigcheck", as we did once with the numeric_big tests? Also, the tests themselves are missing some bets. For instance, inserting identical rows into the different testschmtbl's means you couldn't easily tell if the SELECTs were returning rows from the wrong table. I'd be inclined to use visibly different data, say (1,2) in schm1.testschmtbl and (2,1) in public.testschmtbl. I'd prefer also that the tests for an extremely SQL-standard feature not rely on anything as not-standard as the contrib autoinc() trigger. The autoinc behavior exhibited by the test is a bug if you ask me, and it shouldn't be memorialized as correct behavior by a mainstream regression test ... especially not when the test doesn't make it clear that it's actually testing autoinc's misbehavior and not that of serial sequences. (I had to read it about three times before realizing that the results were not evidence of a serial-sequence problem...) regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 3: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly