Problem with setting register value using MI interface

Problem with setting register value using MI interface

[Grzegorz Cieslewski]

Hi,

I am trying to write a value to a register using MI2 interface.
Unfortunately sometimes th -var-assign command fails and gives me
error.  Does any one know why is that?

GDB<< -var-create r1 * $r1
GDB>> ^done,name="r1",numchild="0",value="2140372880",type="uint32_t"
GDB>> (gdb)
GDB<< -var-info-type r1
GDB>> ^done,type="uint32_t"
GDB>> (gdb)
GDB<< -var-set-format r1 hexadecimal
GDB>> ^done,format="hexadecimal"
GDB>> (gdb)
GDB<< -var-evaluate-expression r1
GDB>> ^done,value="0x7f937f90"
GDB>> (gdb)
GDB<< -var-info-type r1
GDB>> ^done,type="uint32_t"
GDB>> (gdb)
GDB<< -var-set-format r1 hexadecimal
GDB>> ^done,format="hexadecimal"
GDB>> (gdb)
GDB<< -var-assign r1 1
GDB>> &"mi_cmd_var_assign: Could not assign expression to variable object\n"
GDB>> ^error,msg="mi_cmd_var_assign: Could not assign expression to
variable object"
GDB>> (gdb)
GDB<< set $r1=1
GDB>> &"set $r1=1\n"
GDB>> &"Cannot access memory at address 0x1\n"
GDB>> ^error,msg="Cannot access memory at address 0x1"
GDB>> (gdb)


-- 
=====================================================
Grzegorz Cieslewski
Research Assistant
High-performance Computing & Simulation (HCS) Research Laboratory
University of Florida, Dept. of Electrical and Computer Engineering
330 Benton Hall, Gainesville, FL, 32611-6200
Phone: (352) 392-9041
Email: cieslewski@hcs.ufl.edu
Web: www.hcs.ufl.edu
=====================================================

Re: Problem with setting register value using MI interface

[Vladimir Prus]

Grzegorz Cieslewski wrote:

> Hi,
> 
> I am trying to write a value to a register using MI2 interface.
> Unfortunately sometimes th -var-assign command fails and gives me
> error.  Does any one know why is that?
> 
> GDB<< -var-create r1 * $r1
> GDB>> ^done,name="r1",numchild="0",value="2140372880",type="uint32_t"
> GDB>> (gdb)
> GDB<< -var-info-type r1
> GDB>> ^done,type="uint32_t"
> GDB>> (gdb)
> GDB<< -var-set-format r1 hexadecimal
> GDB>> ^done,format="hexadecimal"
> GDB>> (gdb)
> GDB<< -var-evaluate-expression r1
> GDB>> ^done,value="0x7f937f90"
> GDB>> (gdb)
> GDB<< -var-info-type r1
> GDB>> ^done,type="uint32_t"
> GDB>> (gdb)
> GDB<< -var-set-format r1 hexadecimal
> GDB>> ^done,format="hexadecimal"
> GDB>> (gdb)
> GDB<< -var-assign r1 1
> GDB>> &"mi_cmd_var_assign: Could not assign expression to variable
> object\n" GDB>> ^error,msg="mi_cmd_var_assign: Could not assign expression
> to variable object"

Is this a complete output, with no commands in between? Can you reproduce
this on x86, and if so, can you post a complete reproduction recipe?

The only 'error' I immediately see on that code path is this:
        
        if (!frame)
          error (_("Value being assigned to is no longer active."));

in value_assign, but I don't think it can ever fire unless you change frames above.

- Volodya

Re: Problem with setting register value using MI interface

[Daniel Jacobowitz]

On Thu, Oct 25, 2007 at 10:00:11PM +0400, Vladimir Prus wrote:
> Is this a complete output, with no commands in between? Can you reproduce
> this on x86, and if so, can you post a complete reproduction recipe?
> 
> The only 'error' I immediately see on that code path is this:
>         
>         if (!frame)
>           error (_("Value being assigned to is no longer active."));
> 
> in value_assign, but I don't think it can ever fire unless you change frames above.

It can if you're in the outermost frame (i.e. can not backtrace).  I
believe there's a PR about this; I know there were a couple of
discussions about it, and there's a nasty hack in our local source
tree for it too.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Problem with setting register value using MI interface

[Vladimir Prus]

On Thursday 25 October 2007 22:06:38 Daniel Jacobowitz wrote:
> On Thu, Oct 25, 2007 at 10:00:11PM +0400, Vladimir Prus wrote:
> > Is this a complete output, with no commands in between? Can you reproduce
> > this on x86, and if so, can you post a complete reproduction recipe?
> > 
> > The only 'error' I immediately see on that code path is this:
> >         
> >         if (!frame)
> >           error (_("Value being assigned to is no longer active."));
> > 
> > in value_assign, but I don't think it can ever fire unless you change frames above.
> 
> It can if you're in the outermost frame (i.e. can not backtrace).  I
> believe there's a PR about this; I know there were a couple of
> discussions about it, and there's a nasty hack in our local source
> tree for it too.

Yeah, I misremembered when this issue arises. Anyway, I'd rather have a complete
reproduction recipe so that we can be sure it's this bug.

- Volodya

Re: Problem with setting register value using MI interface

[Grzegorz Cieslewski]

Yes, that is complete output with no commands in between.
I was able to replicate it by hand (without my front end)

-break-insert *0x102cea88
^done,bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x102cea88",at="<ATL_slaswp+184>",times="0"}
(gdb)
-exec-run
^running
(gdb)
*stopped,reason="breakpoint-hit",bkptno="1",thread-id="0",frame={addr="0x102cea88",func="ATL_slaswp",args=[]}
(gdb)
-var-create r1 * $r1
^done,name="r1",numchild="0",value="2146255728",type="uint32_t"
(gdb)
-var-assign r1 1
&"mi_cmd_var_assign: Could not assign expression to variable object\n"
^error,msg="mi_cmd_var_assign: Could not assign expression to variable object"
(gdb)

I am currently running on powerpc.  I did not get a chance to try this
on x86, any ideas how can we debug without porting to x86?



On 10/25/07, Vladimir Prus <ghost@cs.msu.su> wrote:
> Grzegorz Cieslewski wrote:
>
> > Hi,
> >
> > I am trying to write a value to a register using MI2 interface.
> > Unfortunately sometimes th -var-assign command fails and gives me
> > error.  Does any one know why is that?
> >
> > GDB<< -var-create r1 * $r1
> > GDB>> ^done,name="r1",numchild="0",value="2140372880",type="uint32_t"
> > GDB>> (gdb)
> > GDB<< -var-info-type r1
> > GDB>> ^done,type="uint32_t"
> > GDB>> (gdb)
> > GDB<< -var-set-format r1 hexadecimal
> > GDB>> ^done,format="hexadecimal"
> > GDB>> (gdb)
> > GDB<< -var-evaluate-expression r1
> > GDB>> ^done,value="0x7f937f90"
> > GDB>> (gdb)
> > GDB<< -var-info-type r1
> > GDB>> ^done,type="uint32_t"
> > GDB>> (gdb)
> > GDB<< -var-set-format r1 hexadecimal
> > GDB>> ^done,format="hexadecimal"
> > GDB>> (gdb)
> > GDB<< -var-assign r1 1
> > GDB>> &"mi_cmd_var_assign: Could not assign expression to variable
> > object\n" GDB>> ^error,msg="mi_cmd_var_assign: Could not assign expression
> > to variable object"
>
> Is this a complete output, with no commands in between? Can you reproduce
> this on x86, and if so, can you post a complete reproduction recipe?
>
> The only 'error' I immediately see on that code path is this:
>
>         if (!frame)
>           error (_("Value being assigned to is no longer active."));
>
> in value_assign, but I don't think it can ever fire unless you change frames above.
>
> - Volodya
>
>
>


-- 
=====================================================
Grzegorz Cieslewski
Research Assistant
High-performance Computing & Simulation (HCS) Research Laboratory
University of Florida, Dept. of Electrical and Computer Engineering
330 Benton Hall, Gainesville, FL, 32611-6200
Phone: (352) 392-9041
Email: cieslewski@hcs.ufl.edu
Web: www.hcs.ufl.edu
=====================================================

Re: Problem with setting register value using MI interface

[Grzegorz Cieslewski]

When I do similar sequence of commands using regular interface I get
weird warning/error "Cannot access memory at address 0x1" but when I
check the value of the r1 register it is written correctly?  Does
anyone know the reason for that?

(gdb) b *0x102cea88
Breakpoint 1 at 0x102cea88
(gdb) r
Starting program: /hcs/home/cieslewski/projects/FaultInjection/gfi_svn/bin/LU

Breakpoint 1, 0x102cea88 in ATL_slaswp ()
(gdb) info reg r1
r1             0x7fe21370       2145522544
(gdb) set $r1=1
Cannot access memory at address 0x1
(gdb) info reg r1
r1             0x1      1
(gdb)





On 10/26/07, Grzegorz Cieslewski <cieslewski@hcs.ufl.edu> wrote:
> Yes, that is complete output with no commands in between.
> I was able to replicate it by hand (without my front end)
>
> -break-insert *0x102cea88
> ^done,bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x102cea88",at="<ATL_slaswp+184>",times="0"}
> (gdb)
> -exec-run
> ^running
> (gdb)
> *stopped,reason="breakpoint-hit",bkptno="1",thread-id="0",frame={addr="0x102cea88",func="ATL_slaswp",args=[]}
> (gdb)
> -var-create r1 * $r1
> ^done,name="r1",numchild="0",value="2146255728",type="uint32_t"
> (gdb)
> -var-assign r1 1
> &"mi_cmd_var_assign: Could not assign expression to variable object\n"
> ^error,msg="mi_cmd_var_assign: Could not assign expression to variable object"
> (gdb)
>
> I am currently running on powerpc.  I did not get a chance to try this
> on x86, any ideas how can we debug without porting to x86?
>
>
>
> On 10/25/07, Vladimir Prus <ghost@cs.msu.su> wrote:
> > Grzegorz Cieslewski wrote:
> >
> > > Hi,
> > >
> > > I am trying to write a value to a register using MI2 interface.
> > > Unfortunately sometimes th -var-assign command fails and gives me
> > > error.  Does any one know why is that?
> > >
> > > GDB<< -var-create r1 * $r1
> > > GDB>> ^done,name="r1",numchild="0",value="2140372880",type="uint32_t"
> > > GDB>> (gdb)
> > > GDB<< -var-info-type r1
> > > GDB>> ^done,type="uint32_t"
> > > GDB>> (gdb)
> > > GDB<< -var-set-format r1 hexadecimal
> > > GDB>> ^done,format="hexadecimal"
> > > GDB>> (gdb)
> > > GDB<< -var-evaluate-expression r1
> > > GDB>> ^done,value="0x7f937f90"
> > > GDB>> (gdb)
> > > GDB<< -var-info-type r1
> > > GDB>> ^done,type="uint32_t"
> > > GDB>> (gdb)
> > > GDB<< -var-set-format r1 hexadecimal
> > > GDB>> ^done,format="hexadecimal"
> > > GDB>> (gdb)
> > > GDB<< -var-assign r1 1
> > > GDB>> &"mi_cmd_var_assign: Could not assign expression to variable
> > > object\n" GDB>> ^error,msg="mi_cmd_var_assign: Could not assign expression
> > > to variable object"
> >
> > Is this a complete output, with no commands in between? Can you reproduce
> > this on x86, and if so, can you post a complete reproduction recipe?
> >
> > The only 'error' I immediately see on that code path is this:
> >
> >         if (!frame)
> >           error (_("Value being assigned to is no longer active."));
> >
> > in value_assign, but I don't think it can ever fire unless you change frames above.
> >
> > - Volodya
> >
> >
> >
>
>
> --
> =====================================================
> Grzegorz Cieslewski
> Research Assistant
> High-performance Computing & Simulation (HCS) Research Laboratory
> University of Florida, Dept. of Electrical and Computer Engineering
> 330 Benton Hall, Gainesville, FL, 32611-6200
> Phone: (352) 392-9041
> Email: cieslewski@hcs.ufl.edu
> Web: www.hcs.ufl.edu
> =====================================================
>


-- 
=====================================================
Grzegorz Cieslewski
Research Assistant
High-performance Computing & Simulation (HCS) Research Laboratory
University of Florida, Dept. of Electrical and Computer Engineering
330 Benton Hall, Gainesville, FL, 32611-6200
Phone: (352) 392-9041
Email: cieslewski@hcs.ufl.edu
Web: www.hcs.ufl.edu
=====================================================