Re: SIGSEGV on gdb 6.7*

[Michael Snyder <msnyder at specifix dot com>]

On Tue, 2008-02-26 at 19:43 -0500, Daniel Jacobowitz wrote:
> On Tue, Feb 26, 2008 at 04:37:33PM -0800, Michael Snyder wrote:
> > On Tue, 2008-02-26 at 19:26 -0500, Daniel Jacobowitz wrote:
> > > On Tue, Feb 26, 2008 at 02:54:47PM -0800, Michael Snyder wrote:
> > > > That said -- I agree with Daniel.  I can see where 
> > > > flushing the register cache and flushing the frame cache
> > > > are two things that should probably always be done at
> > > > the same time -- but I'm worried about the extra overhead
> > > > that this patch introduces.  We call registers_changed
> > > > A LOT, and in doing so we assume that it has a very
> > > > low overhead.
> > > 
> > > If the registers have changed, how can the frame cache still possibly
> > > be valid?
> > 
> > No argument -- it can't.
> > 
> > Are you swinging around toward wanting to accept this patch?
> > ;-)
> 
> I'm talking about Greg's version, which calls it from registers_changed.

Yeah, me too.  This is that thread.  ;-)

> What do you think of that one?

I think it's probably the right thing to do, I'm must
a little concerned about the overhead.  I wouldn't oppose
putting it in, I just thought we could benefit from a little
breathing room to think about it.

The only case where we know it will crash is reg_flush_command, 
and in that case, flushing the frames is unquestionably desirable.
So if we fix it there, we cover the known crasher, and then we 
have breathing space to contemplate the more general fix.

Or, we can just accept Greg's patch.

Back to you... what's your inclination?   ;-)