I tried that for a while, but it runs so slowly under gdb and I gave up waiting for it to segfault. Maybe more patience is all that is required.

I don't quite follow what that means. Starting the application and loading a coredump - doesn't that mean you have two instances of the app: the one you started and the segfaulted one in the coredump? Anyway, if you can explain how to do what you suggest, I'd be grateful.

I've never had issues like this with gdb before although I've used it for many years (but only on i386 not arm). Is there a reason why invoking it in one way would give you an accurate stack dump but in another way the stack appears corrupted?

As you might have guessed from my slow reply, there really is no hurry from my perspective. I'm away from my development box in any case, so can't run gdb. (Staying on my boat - the reason why I want to be able to use my N900 as a backup GPS for OpenCPN!) But I do have my N900 so I can still collect coredumps.

Cheers!