Pete's Log: poll results and sockets madness
Entry #584, (Life in General)(posted when I was 22 years old.)
woohoo, small gain in one of the polls: we're now #11 in the AP, #11 in the coaches.
So my great plans to circumvent the linux accept bug were temporarily defeated by what appeared to be another linux bug. This one, however, is specific to linuxppc. On linuxppc, select suffers from the same signal-caused segfault that accept does on ppc and x86. The good news for linux, i suppose, is that the accept segfault doesn't seem to happen under 2.4-test8. So now I've given up on developing this on realloc and am remotely working on it. Gotta love the fact that arla works well enough over a modem.
But the moral of the story is: I rule. If one of the players gets some kinda signal it'll do the right thing such that all the other players exit cleanly. Which took forever to get right, probably wasn't worth the time I put into it, and won't get me any extra credit. But it was satisfying, I learned a lot, and after a ton of frustration I finally got it right. I won.
Well, messing with this accept stuff some more, I've realized that I can't blame all this oddness on Linux. I'm using sigaction to set up my signal handlers, and I wasn't initalizing all of the members of my sigaction structure (I didn't think I needed to). I guess solaris handles this much more gracefully than Linux does. What a pain that was to debug, though...