Print 64bit address from gdb

Print 64bit address from gdb

[H . J . Lu]

I am looking at the sign extended vma bug. The gdb output doesn't make
any sesnes to me:

During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).

There is

struct complaint innerblock_anon_complaint =
{"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};

I don't think it works with the sign extended vma from the 64bit bfd.
Am I right? Are we going to fix it? I guess we should pass

{"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};

if the address is long long.

H.J.

Re: Print 64bit address from gdb

[Daniel Jacobowitz]

On Thu, Aug 09, 2001 at 04:11:53PM -0700, H . J . Lu wrote:
> I am looking at the sign extended vma bug. The gdb output doesn't make
> any sesnes to me:
> 
> During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).
> 
> There is
> 
> struct complaint innerblock_anon_complaint =
> {"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};
> 
> I don't think it works with the sign extended vma from the 64bit bfd.
> Am I right? Are we going to fix it? I guess we should pass
> 
> {"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};
> 
> if the address is long long.

Yes, that's a problem.  Perhaps you can use your conveniently introduced
*printf_vma functions for this?

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: Print 64bit address from gdb

[H . J . Lu]

On Thu, Aug 09, 2001 at 04:31:41PM -0700, Daniel Jacobowitz wrote:
> On Thu, Aug 09, 2001 at 04:11:53PM -0700, H . J . Lu wrote:
> > I am looking at the sign extended vma bug. The gdb output doesn't make
> > any sesnes to me:
> > 
> > During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).
> > 
> > There is
> > 
> > struct complaint innerblock_anon_complaint =
> > {"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};
> > 
> > I don't think it works with the sign extended vma from the 64bit bfd.
> > Am I right? Are we going to fix it? I guess we should pass
> > 
> > {"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};
> > 
> > if the address is long long.
> 
> Yes, that's a problem.  Perhaps you can use your conveniently introduced
> *printf_vma functions for this?

It should use 

{"inner block (%s-%s) not inside outer block (%s-%s)", 0, 0};

and use

sprintf_vma/bfd_sprintf_vma. bfd_sprintf_vma is better for user and
sprintf_vma may be better for developer. I prefer bfd_sprintf_vma.


H.J.

Re: Print 64bit address from gdb

[Daniel Jacobowitz]

On Fri, Aug 10, 2001 at 09:41:10AM -0700, H . J . Lu wrote:
> On Thu, Aug 09, 2001 at 04:31:41PM -0700, Daniel Jacobowitz wrote:
> > On Thu, Aug 09, 2001 at 04:11:53PM -0700, H . J . Lu wrote:
> > > I am looking at the sign extended vma bug. The gdb output doesn't make
> > > any sesnes to me:
> > > 
> > > During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).
> > > 
> > > There is
> > > 
> > > struct complaint innerblock_anon_complaint =
> > > {"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};
> > > 
> > > I don't think it works with the sign extended vma from the 64bit bfd.
> > > Am I right? Are we going to fix it? I guess we should pass
> > > 
> > > {"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};
> > > 
> > > if the address is long long.
> > 
> > Yes, that's a problem.  Perhaps you can use your conveniently introduced
> > *printf_vma functions for this?
> 
> It should use 
> 
> {"inner block (%s-%s) not inside outer block (%s-%s)", 0, 0};
> 
> and use
> 
> sprintf_vma/bfd_sprintf_vma. bfd_sprintf_vma is better for user and
> sprintf_vma may be better for developer. I prefer bfd_sprintf_vma.

I assume that bfd_sprintf_vma is the one which will truncate to 32bit
in a 32bit BFD?

If so, please use sprintf_vma instead.  These are debugging error
messages; they are primarily of interest to the developers.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: Print 64bit address from gdb

[H . J . Lu]

On Fri, Aug 10, 2001 at 10:41:31AM -0700, Daniel Jacobowitz wrote:
> > > 
> > > Yes, that's a problem.  Perhaps you can use your conveniently introduced
> > > *printf_vma functions for this?
> > 
> > It should use 
> > 
> > {"inner block (%s-%s) not inside outer block (%s-%s)", 0, 0};
> > 
> > and use
> > 
> > sprintf_vma/bfd_sprintf_vma. bfd_sprintf_vma is better for user and
> > sprintf_vma may be better for developer. I prefer bfd_sprintf_vma.
> 
> I assume that bfd_sprintf_vma is the one which will truncate to 32bit
> in a 32bit BFD?

No, it truncates to 32bit for 32bit ELF.

> 
> If so, please use sprintf_vma instead.  These are debugging error
> messages; they are primarily of interest to the developers.

Either way is fine with me.


H.J.

Re: Print 64bit address from gdb

[Elena Zannoni]

H . J . Lu writes:
 > On Thu, Aug 09, 2001 at 04:31:41PM -0700, Daniel Jacobowitz wrote:
 > > On Thu, Aug 09, 2001 at 04:11:53PM -0700, H . J . Lu wrote:
 > > > I am looking at the sign extended vma bug. The gdb output doesn't make
 > > > any sesnes to me:
 > > > 
 > > > During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).
 > > > 
 > > > There is
 > > > 
 > > > struct complaint innerblock_anon_complaint =
 > > > {"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};
 > > > 
 > > > I don't think it works with the sign extended vma from the 64bit bfd.
 > > > Am I right? Are we going to fix it? I guess we should pass
 > > > 
 > > > {"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};
 > > > 
 > > > if the address is long long.
 > > 
 > > Yes, that's a problem.  Perhaps you can use your conveniently introduced
 > > *printf_vma functions for this?
 > 
 > It should use 
 > 
 > {"inner block (%s-%s) not inside outer block (%s-%s)", 0, 0};
 > 
 > and use
 > 
 > sprintf_vma/bfd_sprintf_vma. bfd_sprintf_vma is better for user and
 > sprintf_vma may be better for developer. I prefer bfd_sprintf_vma.
 > 
 > 
 > H.J.


H.J. how do you get the complaints to show up?
The stop_whining variable would seem to stop them, unless you used the 
set complaints command.
Just curious.

I would prefer to use the paddr functions, from utils.c. All the
complaints that print addresses should be updated, then.....


Elena

Re: Print 64bit address from gdb

[H . J . Lu]

On Wed, Aug 15, 2001 at 12:46:43AM -0400, Elena Zannoni wrote:
> 
> 
> H.J. how do you get the complaints to show up?

I was working on the sign extended vma bug in the stabs reader. See

http://sources.redhat.com/ml/gdb-patches/2001-08/msg00115.html


H.J.

Re: Print 64bit address from gdb

[Elena Zannoni]

H . J . Lu writes:
 > On Wed, Aug 15, 2001 at 12:46:43AM -0400, Elena Zannoni wrote:
 > > 
 > > 
 > > H.J. how do you get the complaints to show up?
 > 
 > I was working on the sign extended vma bug in the stabs reader. See
 > 
 > http://sources.redhat.com/ml/gdb-patches/2001-08/msg00115.html
 > 
 > 

Yes, did you turn the complaints on?

(gdb) help set complain
Set max number of complaints about incorrect symbols.
(gdb) show complain
Max number of complaints about incorrect symbols is 0.
(gdb) 

Elena


 > H.J.

Re: Print 64bit address from gdb

[H . J . Lu]

On Wed, Aug 15, 2001 at 01:11:48AM -0400, Elena Zannoni wrote:
> H . J . Lu writes:
>  > On Wed, Aug 15, 2001 at 12:46:43AM -0400, Elena Zannoni wrote:
>  > > 
>  > > 
>  > > H.J. how do you get the complaints to show up?
>  > 
>  > I was working on the sign extended vma bug in the stabs reader. See
>  > 
>  > http://sources.redhat.com/ml/gdb-patches/2001-08/msg00115.html
>  > 
>  > 
> 
> Yes, did you turn the complaints on?
> 
> (gdb) help set complain
> Set max number of complaints about incorrect symbols.
> (gdb) show complain
> Max number of complaints about incorrect symbols is 0.
> (gdb) 

(gdb) show complain
Max number of complaints about incorrect symbols is 0.


H.J.

Re: Print 64bit address from gdb

[Elena Zannoni]

H . J . Lu writes:
 > On Wed, Aug 15, 2001 at 01:11:48AM -0400, Elena Zannoni wrote:
 > > H . J . Lu writes:
 > >  > On Wed, Aug 15, 2001 at 12:46:43AM -0400, Elena Zannoni wrote:
 > >  > > 
 > >  > > 
 > >  > > H.J. how do you get the complaints to show up?
 > >  > 
 > >  > I was working on the sign extended vma bug in the stabs reader. See
 > >  > 
 > >  > http://sources.redhat.com/ml/gdb-patches/2001-08/msg00115.html
 > >  > 
 > >  > 
 > > 
 > > Yes, did you turn the complaints on?
 > > 
 > > (gdb) help set complain
 > > Set max number of complaints about incorrect symbols.
 > > (gdb) show complain
 > > Max number of complaints about incorrect symbols is 0.
 > > (gdb) 
 > 
 > (gdb) show complain
 > Max number of complaints about incorrect symbols is 0.
 > 

Hmm, so the complaint shows up once (at least) even if it the max
allowed is set to zero?

Something to look at in the morning.

Elena


 > 
 > H.J.

Re: Print 64bit address from gdb

[H . J . Lu]

On Wed, Aug 15, 2001 at 01:22:04AM -0400, Elena Zannoni wrote:
>  > > (gdb) 
>  > 
>  > (gdb) show complain
>  > Max number of complaints about incorrect symbols is 0.
>  > 
> 
> Hmm, so the complaint shows up once (at least) even if it the max
> allowed is set to zero?

Yes. I think it is normal since it is an internal gdb bug.


H.J.

Re: Print 64bit address from gdb

[Andrew Cagney]

> I am looking at the sign extended vma bug. The gdb output doesn't make
> any sesnes to me:
> 
> During symbol reading, inner block (0x802ac9d4-0xffffffff) not inside outer block (0x802aca18-0xffffffff).


Mutter something about varargs being dangerous ....  This was on a 
little endian host right?


> There is
> 
> struct complaint innerblock_anon_complaint =
> {"inner block (0x%lx-0x%lx) not inside outer block (0x%lx-0x%lx)", 0, 0};
> 
> I don't think it works with the sign extended vma from the 64bit bfd.
> Am I right? Are we going to fix it? I guess we should pass
> 
> {"inner block (0x%llx-0x%llx) not inside outer block (0x%llx-0x%llx)", 0, 0};
> 
> if the address is long long.


``long long''? ``%ll''?

There are two possible fixes.  Use paddr()/paddr_nz() when the message 
is for debugging or is considered language independant raw value.  Use 
GDB's print formatting mechanism when the context is language dependant.

Looking at complaints.[hc] other than changing things to functions, is 
there a robust (i.e. verifiable at complile time) way of fixing this?

	Andrew