Stabs: constant string, constant boolean

Stabs: constant string, constant boolean

[Daniël Mantione]

Hi,

Could someone enlighten me on the support of the constant string and
constant boolean for stabs debug information.

I tried several versions of gdb (not yet 6.0), but all I get out of them
is "error type". I try to generate debug info exactly as described on
http://source.redhat.com/gdb/current/onlinedocs/stabs_3.html#SEC16

I.e. I try to add a boolean like this:

        .stabs "BOOL1:c=b0;",36,0,5,0

This results in "error type". However, if I try:

        .stabs "BOOL1:c=i0;",36,0,5,0

... gdb treats it as integer fine.

Any ideas which versions of gdb support these well?

Greetings,

Daniël Mantione

P.s. I'm not on this mailinglist, please cc to me.

Re: Stabs: constant string, constant boolean

[Ian Lance Taylor]

Daniël Mantione <daniel@deadlock.et.tudelft.nl> writes:

> Could someone enlighten me on the support of the constant string and
> constant boolean for stabs debug information.
> 
> I tried several versions of gdb (not yet 6.0), but all I get out of them
> is "error type". I try to generate debug info exactly as described on
> http://source.redhat.com/gdb/current/onlinedocs/stabs_3.html#SEC16
> 
> I.e. I try to add a boolean like this:
> 
>         .stabs "BOOL1:c=b0;",36,0,5,0
> 
> This results in "error type". However, if I try:
> 
>         .stabs "BOOL1:c=i0;",36,0,5,0
> 
> ... gdb treats it as integer fine.

Looking at the code, I see that gdb only supports the 'r', 'i', and
'e' types.  The 'b', 'c', 's', and 'S' types are documented, but not
supported.

I note this in the docs:

    The boolean, character, string, and set constants are not
    supported by GDB 4.9, but it ignores them.

This is not entirely unreasonable, as the 'b', 'c', 's', and 'S' types
are short on type information.  I think it would normally be better to
use 'e' instead of 'b' or 'c'.  I suppose that 's' and 'S' aren't seen
as an issue since they won't arise in C.

Ian

Re: Stabs: constant string, constant boolean

[Daniël Mantione]

On 26 Jan 2004, Ian Lance Taylor wrote:

> Looking at the code, I see that gdb only supports the 'r', 'i', and
> 'e' types.  The 'b', 'c', 's', and 'S' types are documented, but not
> supported.
>
> I note this in the docs:
>
>     The boolean, character, string, and set constants are not
>     supported by GDB 4.9, but it ignores them.
>
> This is not entirely unreasonable, as the 'b', 'c', 's', and 'S' types
> are short on type information.  I think it would normally be better to
> use 'e' instead of 'b' or 'c'.  I suppose that 's' and 'S' aren't seen
> as an issue since they won't arise in C.

Thanks for the info, even though that is pretty bad news :(
Constant sets & strings do exist in Pascal, i.e.:

const str1='blablabla';
      set1=['b','l','a'];

I know our former team member Pierre Muller contributed some patches to
add Pascal support to gdb, but apparently not to support these...
Alas, Pierre has stopped with Free Pascal development; I'm currently
trying to give some life-support to the debug-info generation he wrote.

Greetings,

Daniël

Re: Stabs: constant string, constant boolean

[Daniel Jacobowitz]

On Tue, Jan 27, 2004 at 09:39:43AM +0100, Daniël Mantione wrote:
> 
> On 26 Jan 2004, Ian Lance Taylor wrote:
> 
> > Looking at the code, I see that gdb only supports the 'r', 'i', and
> > 'e' types.  The 'b', 'c', 's', and 'S' types are documented, but not
> > supported.
> >
> > I note this in the docs:
> >
> >     The boolean, character, string, and set constants are not
> >     supported by GDB 4.9, but it ignores them.
> >
> > This is not entirely unreasonable, as the 'b', 'c', 's', and 'S' types
> > are short on type information.  I think it would normally be better to
> > use 'e' instead of 'b' or 'c'.  I suppose that 's' and 'S' aren't seen
> > as an issue since they won't arise in C.
> 
> Thanks for the info, even though that is pretty bad news :(
> Constant sets & strings do exist in Pascal, i.e.:
> 
> const str1='blablabla';
>       set1=['b','l','a'];
> 
> I know our former team member Pierre Muller contributed some patches to
> add Pascal support to gdb, but apparently not to support these...
> Alas, Pierre has stopped with Free Pascal development; I'm currently
> trying to give some life-support to the debug-info generation he wrote.

<plug>

If you're working on new code, why not use dwarf2 instead?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: Stabs: constant string, constant boolean

[Daniël Mantione]

On Tue, 27 Jan 2004, Daniel Jacobowitz wrote:

> <plug>
>
> If you're working on new code, why not use dwarf2 instead?

Does dwarf2 support passing these things to gdb then?

I could start on dwarf2 support in compiler, it's on the roadmap, and if
it would give us better support its even a good idea (but it doesn't help
that our gdb expert left the project).

On the other hand we're approaching 2.0, and that's the reason I'm working on the
stabs code; sort out issues with it.

Daniël

Re: Stabs: constant string, constant boolean

[Daniel Jacobowitz]

On Tue, Jan 27, 2004 at 04:26:31PM +0100, Daniël Mantione wrote:
> On Tue, 27 Jan 2004, Daniel Jacobowitz wrote:
> 
> > <plug>
> >
> > If you're working on new code, why not use dwarf2 instead?
> 
> Does dwarf2 support passing these things to gdb then?

Offhand, I don't know.  But it defines them in a much clearer way than
stabs, and adding them to GDB would be easier than adding the
corresponding stabs support.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer