GDB caching question

[Andrew Cagney]

> When I set $fp or $pc varibles in gdb for i386, it sometime does not work. Does anyone else have done 
> something like this. In my case if I do a backgrace following by modification to $fp and $pc to change my stack, it does not work. However If I don't do a "backtrace" first, it works. I suspect some kind of caching of register value is cause of this behaviour.
> 
> Any insight here ?

GDB is _ment_ to flush all caches after any change to the target. 
Sounds like this isn't happening (there was another post related to this 
a week or so back).

Suggest setting break point on all the *changed* methods and see if/when 
they are called.

Andrew

PS: Note that the intent is to replace the lot by replacing a target 
changed event to the observer.