Re: [PATCH v4 5/5] gdb/frame: Add reinflation method for frame_info_ptr

[Bruno Larsen <blarsen@redhat.com>]

On 07/10/2022 21:34, Tom Tromey wrote:
>>>>>> "Bruno" == Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org> writes:
> ...
>
> Bruno> diff --git a/gdb/infcmd.c b/gdb/infcmd.c
> Bruno> index d292a0e55bf..bc171d8818a 100644
> Bruno> --- a/gdb/infcmd.c
> Bruno> +++ b/gdb/infcmd.c
> Bruno> @@ -1814,6 +1814,7 @@ finish_command (const char *arg, int from_tty)
> Bruno>    frame = get_prev_frame (get_selected_frame (_("No selected frame.")));
> Bruno>    if (frame == 0)
> Bruno>      error (_("\"finish\" not meaningful in the outermost frame."));
> Bruno> +  frame.prepare_reinflate ();
>   
> Bruno>    clear_proceed_status (0);
>   
> Bruno> @@ -1872,6 +1873,7 @@ finish_command (const char *arg, int from_tty)
>   
> Bruno>        print_stack_frame (get_selected_frame (NULL), 1, LOCATION, 0);
> Bruno>      }
> Bruno> +  frame.reinflate ();
>   
> I'm curious about this change.  Did you add this to avoid a crash?
> I.e., it's a latent bug of some kind?

This is one of the triggers for bug python/28856 
(https://sourceware.org/bugzilla/show_bug.cgi?id=28856), but a quick 
recap: If one of the arguments of the function has a pretty printer that 
calls a function from the inferior, frame becomes stale and finish breaks.

That bug was the original reason I started this series, lol.

Cheers,
Bruno

>
> Bruno>    if (execution_direction == EXEC_REVERSE)
> Bruno>      finish_backward (sm);
> Bruno> diff --git a/gdb/mi/mi-cmd-stack.c b/gdb/mi/mi-cmd-stack.c
> Bruno> index f23559f886c..2c5bd2db96d 100644
> Bruno> --- a/gdb/mi/mi-cmd-stack.c
> Bruno> +++ b/gdb/mi/mi-cmd-stack.c
> Bruno> @@ -39,6 +39,7 @@
> Bruno>  #include "inferior.h"
> Bruno>  #include "observable.h"
>   
> Bruno> +
> Bruno>  enum what_to_list { locals, arguments, all };
>
> Spurious change.
>
> Tom
>