Bryan Pendleton <[email protected]> writes: > I'm studying the DERBY-2487 patch proposal, and have encountered some > strange behavior that I don't understand. I'm hoping that somebody > on the list can make a suggestion or offer some ideas about where > I should look. > > Below are the results of several queries against the new system table > SYS.SYSXPLAIN_RESULTSETS which is created by the patch, after I've caused > a couple rows to be insert into the table using the new feature. The right > rows > appear to be getting stored in the table, but there's some sort of > subtle corruption occuring, which I don't understand. > > Note the odd behavior of the 'no_opens' column, which sometimes appears > to have the value NULL and sometimes has the value 1.
Perhaps you could try optimizer overrides to control whether the queries use a table scan or an index scan. If table scans and index scans produce different results, it could be something similar to DERBY-4032. -- Knut Anders
