"No registers" error on connecting to MIPS LSI PMON target

"No registers" error on connecting to MIPS LSI PMON target

[Atsushi Nemoto]

I tried to use gdb-6.0 for remote-debugging with MIPS LSI PMON.
I got a message "No registers." on "target lsi /dev/ttyS0" command.

I think this is because get_selected_frame() is called BEFORE the
correct target_ops is installed by push_target().

This is a temporary patch.  It seems to work, but I can not tell the
situation mentioned in the "FIXME" comment is really solved or not.


--- gdb-6.0/gdb/remote-mips.c.org	Wed Jun 11 22:16:28 2003
+++ gdb-6.0/gdb/remote-mips.c	Wed Oct  8 16:55:19 2003
@@ -1494,10 +1494,12 @@
      the request itself succeeds or fails.  */
 
   mips_request ('r', 0, 0, &err, mips_receive_wait, NULL);
+#if 0
   /* FIXME: cagney/2002-11-29: Force the update of selected frame.
      This shouldn't be necessary, only many many places still refer to
      selected_frame directly (instead of using get_selected_frame().  */
   get_selected_frame (); /* Hack!!!  */
+#endif
 }
 
 /* Open a connection to the remote board.  */
---
Atsushi Nemoto

Re: "No registers" error on connecting to MIPS LSI PMON target

[Daniel Jacobowitz]

On Wed, Oct 08, 2003 at 07:47:59PM +0900, Atsushi Nemoto wrote:
> I tried to use gdb-6.0 for remote-debugging with MIPS LSI PMON.
> I got a message "No registers." on "target lsi /dev/ttyS0" command.
> 
> I think this is because get_selected_frame() is called BEFORE the
> correct target_ops is installed by push_target().
> 
> This is a temporary patch.  It seems to work, but I can not tell the
> situation mentioned in the "FIXME" comment is really solved or not.
> 
> 
> --- gdb-6.0/gdb/remote-mips.c.org	Wed Jun 11 22:16:28 2003
> +++ gdb-6.0/gdb/remote-mips.c	Wed Oct  8 16:55:19 2003
> @@ -1494,10 +1494,12 @@
>       the request itself succeeds or fails.  */
>  
>    mips_request ('r', 0, 0, &err, mips_receive_wait, NULL);
> +#if 0
>    /* FIXME: cagney/2002-11-29: Force the update of selected frame.
>       This shouldn't be necessary, only many many places still refer to
>       selected_frame directly (instead of using get_selected_frame().  */
>    get_selected_frame (); /* Hack!!!  */
> +#endif
>  }
>  
>  /* Open a connection to the remote board.  */

Hmm, I've got one of these too and see the same error.  This looks to
predate the deprecation of selected_frame, so I think it can probably
go away now.  Andrew would have a better idea...

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: "No registers" error on connecting to MIPS LSI PMON target

[Andrew Cagney]

>  
>>    mips_request ('r', 0, 0, &err, mips_receive_wait, NULL);
>> +#if 0
>>    /* FIXME: cagney/2002-11-29: Force the update of selected frame.
>>       This shouldn't be necessary, only many many places still refer to
>>       selected_frame directly (instead of using get_selected_frame().  */
>>    get_selected_frame (); /* Hack!!!  */
>> +#endif
>>  }
>>  
>>  /* Open a connection to the remote board.  */
> 
> 
> Hmm, I've got one of these too and see the same error.  This looks to
> predate the deprecation of selected_frame, so I think it can probably
> go away now.  Andrew would have a better idea...

Actually, I don't.  The old code looked like:

   set_current_frame (create_new_frame (read_fp (), read_pc ()));
   select_frame (get_current_frame (), 0);

and the above is "functionaly" equivalent.  My best guess is that 
read_fp() and read_pc() never thought to check if there were registers 
and that let the old code work around a race condition :-(

Also, yes, get_selected_frame() and get_current_frame() being on-demand 
should also have eliminated the need for this method (well, in theory at 
least :-).

So, "if it works for both of you", yank the code.  6.0 as well.

Can you also bug-report the need to audit GDB for this?  I'm pretty sure 
that the code is this code was cloned into several remote-*.c files :-/

Andrew

Re: "No registers" error on connecting to MIPS LSI PMON target

[Daniel Jacobowitz]

On Wed, Oct 08, 2003 at 03:14:07PM -0400, Andrew Cagney wrote:
> > 
> >>   mips_request ('r', 0, 0, &err, mips_receive_wait, NULL);
> >>+#if 0
> >>   /* FIXME: cagney/2002-11-29: Force the update of selected frame.
> >>      This shouldn't be necessary, only many many places still refer to
> >>      selected_frame directly (instead of using get_selected_frame().  */
> >>   get_selected_frame (); /* Hack!!!  */
> >>+#endif
> >> }
> >> 
> >> /* Open a connection to the remote board.  */
> >
> >
> >Hmm, I've got one of these too and see the same error.  This looks to
> >predate the deprecation of selected_frame, so I think it can probably
> >go away now.  Andrew would have a better idea...
> 
> Actually, I don't.  The old code looked like:
> 
>   set_current_frame (create_new_frame (read_fp (), read_pc ()));
>   select_frame (get_current_frame (), 0);
> 
> and the above is "functionaly" equivalent.  My best guess is that 
> read_fp() and read_pc() never thought to check if there were registers 
> and that let the old code work around a race condition :-(
> 
> Also, yes, get_selected_frame() and get_current_frame() being on-demand 
> should also have eliminated the need for this method (well, in theory at 
> least :-).
> 
> So, "if it works for both of you", yank the code.  6.0 as well.

Done, attached.

> Can you also bug-report the need to audit GDB for this?  I'm pretty sure 
> that the code is this code was cloned into several remote-*.c files :-/

Also done.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

2003-10-09  Daniel Jacobowitz  <drow@mvista.com>

	* remote-mips.c (mips_initialize): Remove unneeded call to
	get_selected_frame.  Suggested by Atsushi Nemoto <anemo@mba.ocn.ne.jp>.

Index: remote-mips.c
===================================================================
RCS file: /cvs/src/src/gdb/remote-mips.c,v
retrieving revision 1.40
diff -u -p -r1.40 remote-mips.c
--- remote-mips.c	2 Oct 2003 20:28:30 -0000	1.40
+++ remote-mips.c	9 Oct 2003 14:31:44 -0000
@@ -1494,10 +1494,6 @@ mips_initialize (void)
      the request itself succeeds or fails.  */
 
   mips_request ('r', 0, 0, &err, mips_receive_wait, NULL);
-  /* FIXME: cagney/2002-11-29: Force the update of selected frame.
-     This shouldn't be necessary, only many many places still refer to
-     selected_frame directly (instead of using get_selected_frame().  */
-  get_selected_frame (); /* Hack!!!  */
 }
 
 /* Open a connection to the remote board.  */

Re: "No registers" error on connecting to MIPS LSI PMON target

[Atsushi Nemoto]

>>>>> On Wed, 08 Oct 2003 15:14:07 -0400, Andrew Cagney <ac131313@redhat.com> said:

ac131313> So, "if it works for both of you", yank the code.  6.0 as
ac131313> well.

ac131313> Can you also bug-report the need to audit GDB for this?  I'm
ac131313> pretty sure that the code is this code was cloned into
ac131313> several remote-*.c files :-/

I saw the fix was committed and PR was sent.  Thank you all.
---
Atsushi Nemoto