The Dangers of the Bleeding Edge
Up until now, the segmentation fault has defied all my attempts to discover the cause. In fact, I’m at the point where I am preparing to write my own serialiser to avoid pickling (because it definitely occurs in the pickling process) or to move the server onto a Windows box (because it doesn’t crash on Windows).
I’ve asked a number of questions on #python about pickling, pickling and threads, pickling and c extensions and read the python docs on pickling a dozen times to try and work out how my c code might be causing the problem with no luck. This evening, though, one of the #python members — cygnus — suggested that I should move back to a 2.3 version of python. This involved a few changes to the code, but it did seem more stable, until just a few minutes ago when it also seg-faulted.
I’m going away tomorrow for the weekend, but when I get back, I’m just going to dump the map data in some other way than pickle and see if that “fixes” it. I’m not about to waste more time tracking the problem down and it’s highly likely that the problem will become obvious in developing the workaround.
