remote protocol question

remote protocol question

[Kumar Gala]

I was wondering if there is a way via the remote protocol for a stub to 
tell GDB to update (or invalidate) its copy of any state that might be 
cached?

thanks

- kumar

Re: remote protocol question

[Andrew Cagney]

> I was wondering if there is a way via the remote protocol for a stub to tell GDB to update (or invalidate) its copy of any state that might be cached?

Can you expand on the problem?

If GDB modifies, or resumes, the inferior, it should always invalidate 
its local cache (but no guarentee that it always gets it right :-).

Andrew

Re: remote protocol question

[Kumar Gala]

Andrew,

We are thinking of connecting GDB to a simulator over a socket.  The 
concern is that we stop GDB (via a breakpoint, control-C, etc.).  The 
simulator is able to take input from a separate console window.  It 
would be possible in that console window to then issue a set of 
commands to change the state of the simulation (NIA, GPRs, etc).  The 
question is if we can somehow ensure that when the user continues in 
GDB that the, GDB will have the correction notion of state.

Hopefully, that is clear, but maybe not ;)

thanks

- kumar


On Jul 14, 2004, at 12:29 PM, Andrew Cagney wrote:

>> I was wondering if there is a way via the remote protocol for a stub 
>> to tell GDB to update (or invalidate) its copy of any state that 
>> might be cached?
>
> Can you expand on the problem?
>
> If GDB modifies, or resumes, the inferior, it should always invalidate 
> its local cache (but no guarentee that it always gets it right :-).
>
> Andrew

Re: remote protocol question

[Andrew Cagney]

> Andrew,
> 
> We are thinking of connecting GDB to a simulator over a socket.  The concern is that we stop GDB (via a breakpoint, control-C, etc.).  The simulator is able to take input from a separate console window.  It would be possible in that console window to then issue a set of commands to change the state of the simulation (NIA, GPRs, etc).  The question is if we can somehow ensure that when the user continues in GDB that the, GDB will have the correction notion of state.
> 
> Hopefully, that is clear, but maybe not ;)
> 
> thanks

The remote protocol, which based on a simple master (GDB) / slave (STUB) 
relationship, wasn't designed for this.  I think it would be better 
solved with a properly asynchronous protocol.

Why not instead integrate the simulat and/or console into GDB?

Andrew