Stefan Winter wrote:
FWIW, whatever the problem was I was having with 3.0.1 SEGVing on exit - it's now gone. 3.0.2 exits cleanly.
That's good. I couldn't reproduce the problem locally, but the delay in 3.0.2 was due to other issues I could reproduce. They probably had the same root cause as what you were seeing.
Which means I can finally start thinking about moving 3.0.x to production. :-)
Nice. We spent a *lot* of time hammering 3.0.2 in a wide variety of situations. More valgrind runs, coverity, clang analyzer, etc. They're not perfect, but they help. We've added more regression tests. So any changes to the core which introduce problems will be caught in the daily build / test cycle. That gives us confidence to move forward, and make some planned radical changes to the core. Alan DeKok.