RE: [cdt-dev] target unexpectedly stops on a SIGINT while setting breakp

[Marc Khouzam <marc.khouzam@xxxxxxxxxxxx>]

> -----Original Message-----
> From: cdt-dev-bounces@xxxxxxxxxxx 
> [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Daniel Jacobowitz
> Sent: March-04-10 11:49 AM
> To: CDT General developers list.
> Subject: Re: [cdt-dev] target unexpectedly stops on a SIGINT 
> while setting breakpoint on running target
> 
> On Thu, Mar 04, 2010 at 11:15:40AM -0500, Marc Khouzam wrote:
> > We only use async mode when using non-stop.  I did consider 
> making the 
> > switch but two things stopped me:
> > 
> > 1- lots of regression testing since everything would be using asych 
> > GDB mode instead of the traditional synch mode.
> > 2- We still need to support older GDBs (we support 6.6 and up), so 
> > we'd still need to keep the code for the sync mode.
> > 
> > If it is a big enough problem, we could give it a try for 
> GDB 7.0 and 
> > fix the problem there only, but I'm not sure it is worth the effort.
> 
> Only my two cents, but I don't like thinking about the 
> situation this way.  If we want to give users the best 
> experience, we need to be aggressive about taking advantage 
> of newer features in GDB.
> 
> Not to mention packages that include one particular version 
> of GDB - like Wascana will, and like any vendor's product.

That's very true.
We do have different code for GDB 7.0 and older GDBs so this would not
be a first.  The code change may actually be small.

What is the fundamental change in GDB behavior when using async mode?
Is it simply that we still have a prompt that will accept -exec-interrupt
while the target is running?
For example, I would have thought we could give breakpoint commands in
this case, but it turns out they still fail if the target is running.
So, if we can get a feel of what GDB will do different, we could find
that it is not a big deal to make the switch.

Marc