>> Just thinking out loud, but might it have something to do >> with threading concerns? If I hand out a reference to >> something on my stack to a strange API how do I know that >> it will finish with it before my stack changes.. > >But it isn't a strange API. After all, presumably the Managed DirectX >API designers understand how the DirectX API works. I'm no DirectX >programmer, but I assume that since it isn't documented, that the API >will not hold on the pointer.
I guess I meant 'strange' as in 'closed source, poorly documented, "I didn't write it", etc.' (don't know much about DirectX doco, so that's not a sledge as such). Perhaps the API designers are just trying to save programmers from themselves again..? Say for example I'm targeting the managed DirectX API, and my client code is multithreaded, and I do something horrible, passing structs allocated on my stack by ref to a 'helper' method that dereferences them and passes them by value to the underlying API for execution on another thread. I'd still be safe, because the dereferencing happened on my thread, however if I passed them by ref to the other thread, then by the time the second thread is executing my original thread has changed state, and literally anything might be on the stack. Is there any safeguard in place that could catch this type of thing? As I (vaguely) understand it, the difference between a class and struct is how much 'support' they get from the CLR. So a struct is more like a 'set of bits' and a class is more like a 'component' (it gets an object id, etc.). If you're just passing references around to a 'set of bits' what metadata do you have that says what is *really* at that set of bits..? I don't know a lot about the details, but it seems like this is why they may have decided to pass by value. It's just safer, and there is less to think about. The other argument might have been that the value is 'read' many more times that it is 'written', so the extra level of indirection (to dereference the pointer) ends up costing more than just passing a few extra bits up front. This argument seems to be shot to bits by the fact that the real work gets done on a pointer though.. John. =================================== This list is hosted by DevelopMentorŪ http://www.develop.com Some .NET courses you may be interested in: NEW! Guerrilla ASP.NET, 17 May 2004, in Los Angeles http://www.develop.com/courses/gaspdotnetls View archives and manage your subscription(s) at http://discuss.develop.com
