This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix for GDB failing to interrupt after run when no PID issued by stub


Omair Javaid <omair.javaid@linaro.org> writes:

> Heres the gdb log with run (no other command b/w connection and run):
> https://paste.ubuntu.com/26496015/
>

One thing suspicious to me is an empty reply to qXfer:threads:read

w $qXfer:threads:read::0,fff#03
r $l<?xml version="1.0"?>\n<threads>\n</threads>\n#02

OpenOCD claims support "qXfer:threads:read+;", but such reply above may
confuse GDB.  In OpenOCD source, I find OpenOCD supports qXfer:threads
unconditionally, but I think it should be changed to support it when
"target->rtos != NULL".  This may don't matter.

> Heres the gdb log with continue + run (after we issue a
> continue/interrupt cycle where inferior PID is set and
> inferior->control.stop_soon = NO_STOP_QUIETLY;)
> https://paste.ubuntu.com/26496048/
>

>
> The situation with OpenOCD happens because we ran and set inferior pid
> to null hoping to update it. Later in extended_remote_create_inferior
> called buy run command thread pid gets assigned but we are checking
> inferior->control.stop_soon just before we update the pid and thus it
> remains to STOP_QUIETLY for lifetime of the inferior as this flag is
> only altered by attach command handlers or during call of
> clear_proceed_status 

I still don't understand why inferior->control.stop_soon remains
"STOP_QUIETLY", inferior::control::stop_soon is initialized to
NO_STOP_QUIETLY.  I even don't see where it is changed to STOP_QUIETLY.

I can understand that GDB check inferior_ptid (try to set
inferior->control.stop_soon) first [1], and then update inferior_ptid
later [2],

  if (!have_inferiors ())
    {
      /* Clean up from the last time we ran, before we mark the target
	 running again.  This will mark breakpoints uninserted, and
	 get_offsets may insert breakpoints.  */
      init_thread_list ();
      init_wait_for_inferior ();  <-- [1]
    }

  /* vRun's success return is a stop reply.  */
  stop_reply = run_worked ? rs->buf : NULL;
  add_current_inferior_and_thread (stop_reply);  <-- [2]

However, current_inferior_ isn't changed at all, and
current_inferior_->control.stop_soon is not changed either
(NO_STOP_QUIETLY).  Could you show me where inf->control.stop_soon is
changed from NO_STOP_QUIETLY?

-- 
Yao (齐尧)


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]