public inbox for ecos-discuss@sourceware.org
 help / color / mirror / Atom feed
* [ECOS] ARM Debuggers backtrace
@ 2001-09-24 15:56 Fabrice Gautier
  2001-09-25  7:34 ` Grant Edwards
  0 siblings, 1 reply; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-24 15:56 UTC (permalink / raw)
  To: Ecos-List (E-mail)

Hi,

I'm not sure if gdb is abble to display call stack but ARMs debugger is not
able to do so, at least on a default build.

I tried to add the -apcs-frame option to gdb, but it seems it doesn't help.

Does anyone now if there is a way to display call stack properly on ARM?

Thanks

-- 
Fabrice Gautier
Software Engineer, Sigma Designs
Fabrice_Gautier@sdesigns.com

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [ECOS] ARM Debuggers backtrace
  2001-09-24 15:56 [ECOS] ARM Debuggers backtrace Fabrice Gautier
@ 2001-09-25  7:34 ` Grant Edwards
  0 siblings, 0 replies; 11+ messages in thread
From: Grant Edwards @ 2001-09-25  7:34 UTC (permalink / raw)
  To: Fabrice Gautier; +Cc: Ecos-List (E-mail)

On Mon, Sep 24, 2001 at 04:04:57PM -0700, Fabrice Gautier wrote:

> I'm not sure if gdb is abble to display call stack but ARMs
> debugger is not able to do so, at least on a default build.

You're joking!  A debugger that can't display a call stack?

> I tried to add the -apcs-frame option to gdb, but it seems it
> doesn't help.
> 
> Does anyone now if there is a way to display call stack
> properly on ARM?

I'm using GDB 5.0, and the "bt" command works fine: it displays
the call stack complete with parameter values.

-- 
Grant Edwards
grante@visi.com

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
  2001-09-27 20:28 Fabrice Gautier
@ 2001-09-28  2:49 ` Philippe Robin
  0 siblings, 0 replies; 11+ messages in thread
From: Philippe Robin @ 2001-09-28  2:49 UTC (permalink / raw)
  To: Fabrice Gautier; +Cc: Ecos-List (E-mail)

Hi Fabrice,

Yes we aware of this, see issue number 000519.1 at 
http://www.eagercon.com/dwarf/dwarf2std.htm . This is what made the wording 
in the DWARF3 (was DWARF2.1) draft specification to be changed to be clear 
that the offset should not be factored; and also added some new signed 
frame operators.

Regards,

Philippe

At 08:37 PM 9/27/01 -0700, Fabrice Gautier wrote:
>Hi,
>
>Well, gcc from cvs now generate the .debug_frame section used by the arm
>debuggers.
>But now there it's seems there's a bug in the Arm debugger....
>
>(see http://gcc.gnu.org/ml/gcc/2001-09/msg01172.html and thread for details)
>
>Philippe, since you work at ARM, I figured this could be a good idea if you
>could tell the people in charge of the Arm debuggers (maybe in particular
>the ones working on the dwarf2 stuff) that it seems they got something
>wrong.
>
>Thanks
>
>
> > -----Original Message-----
> > From: Fabrice Gautier [ mailto:Fabrice_Gautier@sdesigns.com ]
> > Subject: RE: [ECOS] ARM Debuggers backtrace
> >
>
> > And all that to see that gcc 3 doesn't seem to handle ARM
> > call stack better
> > that 2.95 ... It looks like the ARM debugger needs debug
> > symbol rather than
> > frame pointers. When I "objdump" a image compiled with armcc
> > there is an
> > debug_frame section, can't have it with gcc.... (see
> > http://www.arm.com/support.ns4/html/sdt_debug?OpenDocument&Exp
>andSection=8#_
>Section8 about this "virtual stack frame" stuff )
>
>
>         grmffff..
>
>--
>Fabrice Gautier
>Software Engineer, Sigma Designs
>Fabrice_Gautier@sdesigns.com

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
@ 2001-09-27 20:28 Fabrice Gautier
  2001-09-28  2:49 ` Philippe Robin
  0 siblings, 1 reply; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-27 20:28 UTC (permalink / raw)
  To: Fabrice Gautier, 'Jonathan Larmour'
  Cc: Ecos-List (E-mail), 'Philippe Robin'

Hi,

Well, gcc from cvs now generate the .debug_frame section used by the arm
debuggers.
But now there it's seems there's a bug in the Arm debugger.... 

(see http://gcc.gnu.org/ml/gcc/2001-09/msg01172.html and thread for details)

Philippe, since you work at ARM, I figured this could be a good idea if you
could tell the people in charge of the Arm debuggers (maybe in particular
the ones working on the dwarf2 stuff) that it seems they got something
wrong. 

Thanks


> -----Original Message-----
> From: Fabrice Gautier [ mailto:Fabrice_Gautier@sdesigns.com ]
> Subject: RE: [ECOS] ARM Debuggers backtrace
> 

> And all that to see that gcc 3 doesn't seem to handle ARM 
> call stack better
> that 2.95 ... It looks like the ARM debugger needs debug 
> symbol rather than
> frame pointers. When I "objdump" a image compiled with armcc 
> there is an
> debug_frame section, can't have it with gcc.... (see
> http://www.arm.com/support.ns4/html/sdt_debug?OpenDocument&Exp
andSection=8#_
Section8 about this "virtual stack frame" stuff )


	grmffff..

-- 
Fabrice Gautier
Software Engineer, Sigma Designs
Fabrice_Gautier@sdesigns.com

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
@ 2001-09-25 16:32 Fabrice Gautier
  0 siblings, 0 replies; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-25 16:32 UTC (permalink / raw)
  To: 'Jonathan Larmour'; +Cc: Ecos-List (E-mail)

> -----Original Message-----
> From: Jonathan Larmour [ mailto:jlarmour@redhat.com ]
> Sent: Tuesday, September 25, 2001 2:56 PM
> Subject: Re: [ECOS] ARM Debuggers backtrace
> 
> Fabrice Gautier wrote:
> > 
> > I tried the command by hand, using -v for verbose, and it 
> > showed that it used as to compile not arm-elf-as as it should.
> > 
> > I'n gonna triple-check everything but i'm a bit dismayed...
> 
> It really should work if you're using the same prefix. That's 
> what I did...

Indeed, so my dismay...

Trust me multiple check is not enough 

One's have to try to reproduce manually all the gory commands, cpp and as,
to find his mistake... 

That was just a typo "H-i866-pc-cygwin" instead of "H-i686-pc-cygwin" (or
how to loose a few hours)

Whatever....


And all that to see that gcc 3 doesn't seem to handle ARM call stack better
that 2.95 ... It looks like the ARM debugger needs debug symbol rather than
frame pointers. When I "objdump" a image compiled with armcc there is an
debug_frame section, can't have it with gcc.... (see
http://www.arm.com/support.ns4/html/sdt_debug?OpenDocument&ExpandSection=8#_
Section8 about this "virtual stack frame" stuff )


	grmffff..

-- 
Fabrice Gautier
Software Engineer, Sigma Designs
Fabrice_Gautier@sdesigns.com

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [ECOS] ARM Debuggers backtrace
  2001-09-25 14:11 Fabrice Gautier
@ 2001-09-25 14:56 ` Jonathan Larmour
  0 siblings, 0 replies; 11+ messages in thread
From: Jonathan Larmour @ 2001-09-25 14:56 UTC (permalink / raw)
  To: Fabrice Gautier; +Cc: Ecos-List (E-mail)

Fabrice Gautier wrote:
> 
> I tried the command by hand, using -v for verbose, and it showed that it
> used as to compile not arm-elf-as as it should.
> 
> I'n gonna triple-check everything but i'm a bit dismayed...

It really should work if you're using the same prefix. That's what I did...

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
@ 2001-09-25 14:11 Fabrice Gautier
  2001-09-25 14:56 ` Jonathan Larmour
  0 siblings, 1 reply; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-25 14:11 UTC (permalink / raw)
  To: 'Jonathan Larmour'; +Cc: Ecos-List (E-mail)

> -----Original Message-----
> From: Jonathan Larmour [ mailto:jlarmour@redhat.com ]
> Sent: Tuesday, September 25, 2001 1:02 PM
> To: Fabrice Gautier
> Cc: Ecos-List (E-mail)
> Subject: Re: [ECOS] ARM Debuggers backtrace
> 
> 
> Fabrice Gautier wrote:
> > 
> > But i'm now trying to see if GCC 3 solved the problem 
> (there is a new ARM
> > back end in Gcc 3 so maybe...) but i've got some 
> difficulties building gcc
> > 3...
> 
> I've done it. What's the problem?
> 

Strange problem when building libgcc: xgcc keeps using as instead of
arm-elf-as... First I though it was because i was not building gcc with the
same "--prefix" that I used for binutils. But then i rebuilded binutils and
it's the same problem again...

First error message is :

/d/gnu-packages/gcc-3-build/gcc/xgcc -B/d/gnu-packages/gcc-3-build/gcc/
-B/d/gnu
-tools-3/H-i866-pc-cygwin/arm-elf/bin/
-B/d/gnu-tools-3/H-i866-pc-cygwin/arm-elf
/lib/ -isystem /d/gnu-tools-3/H-i866-pc-cygwin/arm-elf/include -O2
-DCROSS_COMP
ILE -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes
-Wmissing-prototype
s -isystem ./include  -Dinhibit_libc -fno-inline -g1  -DIN_LIBGCC2
-D__GCC_FLOAT
_NOT_NEEDED -Dinhibit_libc -I. -I. -I../../gcc-3.0.1/gcc
-I../../gcc-3.0.1/gcc/.
 -I../../gcc-3.0.1/gcc/config -I../../gcc-3.0.1/gcc/../include -DL_udivsi3
-xass
embler-with-cpp -c ../../gcc-3.0.1/gcc/config/arm/lib1funcs.asm -o
libgcc/./_udi
vsi3.o
/tmp/ccLRgTpx.s: Assembler messages:
/tmp/ccLRgTpx.s:2: Warning: rest of line ignored; first ignored character is
`@'


I tried the command by hand, using -v for verbose, and it showed that it
used as to compile not arm-elf-as as it should.


I'n gonna triple-check everything but i'm a bit dismayed...


	

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [ECOS] ARM Debuggers backtrace
  2001-09-25 12:52 Fabrice Gautier
@ 2001-09-25 13:02 ` Jonathan Larmour
  0 siblings, 0 replies; 11+ messages in thread
From: Jonathan Larmour @ 2001-09-25 13:02 UTC (permalink / raw)
  To: Fabrice Gautier; +Cc: Ecos-List (E-mail)

Fabrice Gautier wrote:
> 
> But i'm now trying to see if GCC 3 solved the problem (there is a new ARM
> back end in Gcc 3 so maybe...) but i've got some difficulties building gcc
> 3...

I've done it. What's the problem?

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
@ 2001-09-25 12:52 Fabrice Gautier
  2001-09-25 13:02 ` Jonathan Larmour
  0 siblings, 1 reply; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-25 12:52 UTC (permalink / raw)
  To: 'Jonathan Larmour'; +Cc: Ecos-List (E-mail)

> -----Original Message-----
> From: Jonathan Larmour [ mailto:jlarmour@redhat.com ]
> Sent: Tuesday, September 25, 2001 12:40 PM
> To: Fabrice Gautier
> Cc: Ecos-List (E-mail)
> Subject: Re: [ECOS] ARM Debuggers backtrace
> 
> 
> Fabrice Gautier wrote:
> > >
> > > > I'm not sure if gdb is abble to display call stack but ARMs
> > > > debugger is not able to do so, at least on a default build.
> > >
> [snip] 
> > So gcc + gdb works and armcc + ARM debugger works but not gcc + ARM
> > debugger... too bad...going to ask some gcc/gdb mailing lists.
> 
> Try compiling with -mapcs-frame if you haven't already.

Already tried (correct --apcs-frame to -mapcs-frame in my first mail) 

This is apparently a known problem, saw something about gcc-2.95 not
generating correct stack frames for Arm debugger. A look at the assembly
shows that the generated code is actualy quite different.

But i'm now trying to see if GCC 3 solved the problem (there is a new ARM
back end in Gcc 3 so maybe...) but i've got some difficulties building gcc
3... 

Thanks

-- 
Fabrice Gautier
Software Engineer, Sigma Designs
Fabrice_Gautier@sdesigns.com


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [ECOS] ARM Debuggers backtrace
  2001-09-25 10:10 Fabrice Gautier
@ 2001-09-25 12:39 ` Jonathan Larmour
  0 siblings, 0 replies; 11+ messages in thread
From: Jonathan Larmour @ 2001-09-25 12:39 UTC (permalink / raw)
  To: Fabrice Gautier; +Cc: Ecos-List (E-mail)

Fabrice Gautier wrote:
> >
> > > I'm not sure if gdb is abble to display call stack but ARMs
> > > debugger is not able to do so, at least on a default build.
> >
[snip] 
> So gcc + gdb works and armcc + ARM debugger works but not gcc + ARM
> debugger... too bad...going to ask some gcc/gdb mailing lists.

Try compiling with -mapcs-frame if you haven't already.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

^ permalink raw reply	[flat|nested] 11+ messages in thread

* RE: [ECOS] ARM Debuggers backtrace
@ 2001-09-25 10:10 Fabrice Gautier
  2001-09-25 12:39 ` Jonathan Larmour
  0 siblings, 1 reply; 11+ messages in thread
From: Fabrice Gautier @ 2001-09-25 10:10 UTC (permalink / raw)
  To: 'Grant Edwards'; +Cc: Ecos-List (E-mail)

> -----Original Message-----
> From: Grant Edwards [ mailto:grante@visi.com ]
> Sent: Tuesday, September 25, 2001 7:37 AM
> To: Fabrice Gautier
> Cc: Ecos-List (E-mail)
> Subject: Re: [ECOS] ARM Debuggers backtrace
> 
> 
> On Mon, Sep 24, 2001 at 04:04:57PM -0700, Fabrice Gautier wrote:
> 
> > I'm not sure if gdb is abble to display call stack but ARMs
> > debugger is not able to do so, at least on a default build.
> 
> You're joking!  A debugger that can't display a call stack?

Actually it can if I use their compiler...

 
> I'm using GDB 5.0, and the "bt" command works fine: it displays
> the call stack complete with parameter values.

So gcc + gdb works and armcc + ARM debugger works but not gcc + ARM
debugger... too bad...going to ask some gcc/gdb mailing lists.

Thanks,

  Fabrice

^ permalink raw reply	[flat|nested] 11+ messages in thread

end of thread, other threads:[~2001-09-28  2:49 UTC | newest]

Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-09-24 15:56 [ECOS] ARM Debuggers backtrace Fabrice Gautier
2001-09-25  7:34 ` Grant Edwards
2001-09-25 10:10 Fabrice Gautier
2001-09-25 12:39 ` Jonathan Larmour
2001-09-25 12:52 Fabrice Gautier
2001-09-25 13:02 ` Jonathan Larmour
2001-09-25 14:11 Fabrice Gautier
2001-09-25 14:56 ` Jonathan Larmour
2001-09-25 16:32 Fabrice Gautier
2001-09-27 20:28 Fabrice Gautier
2001-09-28  2:49 ` Philippe Robin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).