Stepping Into Fiber

This has nothing to do with your problem, but I'm curious what the fibers are for? The MSDN docs seem to suggest that they're merely a convenience for porting apps that use user threads and schedule them manually, and should be avoided otherwise.

MSVC (both 6.0 and the .Net versions) do indeed support switching the currently debugged thread, although I've had mixed results with threads that are manually suspended and resumed at run time (as opposed to being directly blocked with synchronization calls, for instance). As for fibers... what Promit said.

Quote:

Original post by Promit This has nothing to do with your problem, but I'm curious what the fibers are for? The MSDN docs seem to suggest that they're merely a convenience for porting apps that use user threads and schedule them manually, and should be avoided otherwise.

You are correct about one thing, they are basically threads you can schedule yourself. If you need that ability then they are useful. I wouldn't say you should avoid them though. "Use the right tool for the job"

Quote:

Original post by duke Quote: Original post by Promit This has nothing to do with your problem, but I'm curious what the fibers are for? The MSDN docs seem to suggest that they're merely a convenience for porting apps that use user threads and schedule them manually, and should be avoided otherwise. You are correct about one thing, they are basically threads you can schedule yourself. If you need that ability then they are useful. I wouldn't say you should avoid them though. "Use the right tool for the job"

Well, the thing is, I'm skeptical of why you would ever need to (or want to) schedule threads yourself. Since I do know Shannon's a bright guy, I am curious about what's going on.

I ported an RTOS to Win32 :)

I initially tried to make each task its own thread but this worked out very poorly. I had a lot of trouble trying to control which task thread was allowed to run (using SuspendThread & ResumeThread) with race conditions. After three or four 'fixes' to try to make the switches atomic I remembered Fibers and that you controlled exactly when you switched between them (atomic task swap).

Everything runs smoothly now since there's no race conditions and everything is syncronous and very deterministic. Or at least it did until I tried to run code coverage... which died horribly and now I'm trying to debug it.