No more zone just yet, wine’s just made me blabby.
The testing I’ve just been doing with my server is real black box stuff. The build process fires off a script that kicks off the server and then runs the test harness which connects to the server and does “good stuff”, on lots of threads, until it’s done. The script then runs a server shutdown app which asks the server to clean up and go away and, well, it does.
When you’re doing the consultant thing in banks and the like it’s strange how many servers you find that are really hard to shut down cleanly… If you can’t ask it to stop how can you ever get meaninful results out of profiling or memory leak tools? How can you ever run automated tests that start the server, run the test then stop the server? How can you ever tell if it goes blam when it shuts down (ah, I forgot, of course it never shuts down so it would never go blam anyway…).