After days and days of work building test cases and testing on database images I ran a "fix a screw up" script yesterday only to find out this morning that the "fix" wound up making the problem far, far, worse. Spent the entire day trying to fix that today.

The problem, it turns out, is that all of my test cases were created with the wrong understanding of the nature of what a particular table stored (I thought it was debit-only (i.e. a trace of invoices), turns out it is really a ledger-style table (i.e. values get entered that must balance to 0)). So though all the test cases I could create were passing, and every test was as rigorous as it could be, it was all just being absolutely sure it would be wrong.

Going to have more work on that tomorrow. Then probably have to crunch through much of the weekend to catch up on work that's been sidelined by this stuff. I'm really getting to the point of needing a vacation.

Oh, and I need to get some work done on PyOpenGL one of these days too...