debugging timing conflicts

i have a problem...or worse, ironically, i had a problem.

for the last two days, when my application opened, it's execution would simply stop. it was not caught in an infinite loop (CPU usage was 0) and no error was encountered.

i was in the process of trying to isolate the problematic bits when, as mysteriously as it appeared, it went away. it is definitely not something i changed, because a copy of the malfunctioning code now functions also.

i do not have a clue as to how to go about debugging something like this. any suggestions would be much appreciated.

You could keep a log file of timings with different parts of your code, especially around system calls and other calls to external resources.

I know just the thread for you. :) http://cboard.cprogramming.com/showthread.php?t=105440

why thank you good sir.

i actually found the source of the problem yesterday. the malfunctioning code promptly resumed malfunctioning repeatably after i posted that, and the version i was working on still didn't. so i ran a diff on all the files in the project and sure enough, i had actually fixed it.

i still don't understand why it was producing that behavior though. as best i can tell, the issue seems to have been that a thread was erroneously being executed in the main thread. this particular thread had an empty functor list, so it immediately suspended itself; which for some reason suspended execution of the main thread.

i do not get it. I called the method to execute method in it's own thread, not the main thread, so why this happened is truly a mystery to me.

You mean, why it sometimes crashed and sometimes didn't?

That's the nature of memory errors. Buffer overruns, invalid pointer accesses, double frees, etc. They can be very hard to track down. Valgrind's my favourite tool for doing so, or GDB if it actually segfaults. :)

Take buffer overruns, for example. Will it crash? It depends on what you're overwriting. If you're overwriting other variables in your program, your program might behave strangely. ("I entered 3 so why does it say 56892???") If it's overwriting into another process's memory, you'll probably get a segmentation fault, which you can at least debug.

But far worse things might happen. You might start overwriting your executable code and change the behaviour of your program, for example.

see edit...

:)

it never crashed...it hung, as though it were in an infinite loop, but CPU usage was 0%.

somehow the application's main thread was suspended instead of the intended thread; which implies that the method of interest was erroneously executed sequentially instead of in it's own thread.

I'm sorry, I forgot which thread I was replying to . . . my fault.

You could, perhaps, have run the program in a debugger, and interrupted it when it froze . . . if it froze while taking zero CPU, it's possible that the thread was just waiting for something (or two threads were waiting for each other . . .), and you'd at least be able to see where it was waiting.

Or you could just have examined your threading code very, very carefully, since it seems like it was a threading issue. :)

Anyway . . . you've fixed it now, haven't you?

yes, i did, but thanks for pointing me to that other thread; lots of good stuff in there.