Re: [PATCH] gdb/stack.c: avoid stale pointers when printing frame arguments

[Andrew Burgess <aburgess@redhat.com>]

* Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org> [2022-04-05 11:47:20 -0300]:

> On 4/5/22 10:58, Tom Tromey wrote:
> > > > > > > "Bruno" == Bruno Larsen <blarsen@redhat.com> writes:
> > 
> > Bruno> This sounds like a good idea. I am just not sure if you are
> > Bruno> suggesting it as a fix instead of what I proposed, or to
> > Bruno> implement later, can you clarify it please?
> > 
> > You don't have to do it.
> > 
> > Bruno> +  if (frame_cache_count < get_frame_cache_generation ())
> > Bruno> +    reinit_frame_cache ();
> > 
> > > > I don't think I understand this bit.  If the generation changes,
> > > > hasn't
> > > > the cache already been cleared?
> > 
> > Bruno> If the cache has been cleared by printing a frame, it was done because
> > Bruno> a function was called manually (probably). If it did happen, the cache
> > Bruno> may have been invalidated and it is safer to rebuild
> > Bruno> everything. print_frame_cache by itself doesn't reinitialize the frame
> > Bruno> cache.
> > 
> > My understanding is that the generation only changes when
> > reinit_frame_cache is called.  So if that's the case, why does it need
> > to be reinitalized a second time?  That's the part I don't understand.
> 
> It has been reinitialized to call the function by hand (if you set
> debug frame 1 and call something by hand, you'll see a few
> refreshes), but it is not been refreshed after finishing the
> handmade call. We might still have dummy frames or incorrect
> information leftover from the function call.

If the handmade (inferior) call can leave invalid frames in the frame
cache, then the correct thing (I would think) is to flush the frame
cache after making the inferior call.

Thanks,
Andrew