Re: [PATCH 1/2] gdb/infrun: make fetch_inferior_event restore thread if exited or signalled

[Pedro Alves <pedro@palves.net>]

On 2022-04-24 04:20, Simon Marchi via Gdb-patches wrote:

> Change-Id: Ibc7df543e2c46aad5f3b9250b28c3fb5912be4e8
> ---
>  gdb/infrun.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/gdb/infrun.c b/gdb/infrun.c
> index 4e7ca803c798..6dccb7a2ff2a 100644

Looks fine.

> --- a/gdb/infrun.c
> +++ b/gdb/infrun.c
> @@ -4052,12 +4052,19 @@ reinstall_readline_callback_handler_cleanup ()
>  static void
>  clean_up_just_stopped_threads_fsms (struct execution_control_state *ecs)
>  {
> +  /* The first clean_up call below assumes the event thread is the current
> +     one.  */
> +  if (ecs->event_thread != nullptr)
> +    gdb_assert (ecs->event_thread == inferior_thread ());
> +
>    if (ecs->event_thread != nullptr
>        && ecs->event_thread->thread_fsm () != nullptr)
>      ecs->event_thread->thread_fsm ()->clean_up (ecs->event_thread);
>  
>    if (!non_stop)
>      {
> +      scoped_restore_current_thread restore_thread;
> +
>        for (thread_info *thr : all_non_exited_threads ())
>  	{
>  	  if (thr->thread_fsm () == nullptr)
> @@ -4068,9 +4075,6 @@ clean_up_just_stopped_threads_fsms (struct execution_control_state *ecs)
>  	  switch_to_thread (thr);
>  	  thr->thread_fsm ()->clean_up (thr);

We could also look at eliminating the need to switch the thread
when we're about to call a function that takes an explicit thread_info pointer,
but I'm fine with what you have, especially since in practice that would just probably
mean a scoped_restore_current_thread per clean_up call instead of a just a single one for
all threads...