dumb question; mips multi-abi

dumb question; mips multi-abi

[Michael Snyder]

I confess I have not been following the development of ABI support
very closely.  I have a user who wants to debug an application that
uses a mix of (essentially) o32 and o64 code.  I think he has some
sort of calling translation layer between the two.

Is that at all possible?  Can gdb switch back and forth on the fly,
while debugging a single app?  Maybe manually?  Maybe without expecting
to be able to bridge the backtrace?

Re: dumb question; mips multi-abi

[Daniel Jacobowitz]

On Thu, Mar 17, 2005 at 11:40:27AM -0800, Michael Snyder wrote:
> I confess I have not been following the development of ABI support
> very closely.  I have a user who wants to debug an application that
> uses a mix of (essentially) o32 and o64 code.  I think he has some
> sort of calling translation layer between the two.
> 
> Is that at all possible?  Can gdb switch back and forth on the fly,
> while debugging a single app?  Maybe manually?  Maybe without expecting
> to be able to bridge the backtrace?

Manually should work fine - set mips abi.  In practice in HEAD I think
you get an internal error from a missing call to flush_cached_frames();
that's on my list to push out this month, I hope...

Automatically, I have no idea.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: dumb question; mips multi-abi

[Joel Brobecker]

> I confess I have not been following the development of ABI support
> very closely.  I have a user who wants to debug an application that
> uses a mix of (essentially) o32 and o64 code.  I think he has some
> sort of calling translation layer between the two.
> 
> Is that at all possible?  Can gdb switch back and forth on the fly,
> while debugging a single app?  Maybe manually?  Maybe without expecting
> to be able to bridge the backtrace?

Hmmm, I am a bit pessimistic about this, unfortunately. I can't tell
you for sure, because I don't know this area of mips-tdep very much,
but my recent reading of this file seems to indicate that the ABI
is set in the gdbarch tdep structure when the executable is loaded.

I also noticed recently that you could force the ABI using:

      add_setshow_enum_cmd ("abi", class_obscure, mips_abi_strings,
                            &mips_abi_string, _("\
    Set the MIPS ABI used by this program."), _("\
    Show the MIPS ABI used by this program."), _("\
    This option can be set to one of:\n\
      auto  - the default ABI associated with the current binary\n\
      o32\n\
      o64\n\
      n32\n\
      n64\n\
      eabi32\n\
      eabi64"),
                            mips_abi_update,
                            show_mips_abi,
                        &setmipscmdlist, &showmipscmdlist);

But then again, this setting is only used when initializing the
gdbarch structure...

It looks like you'll have to choose on ABI between the two... :-/

-- 
Joel

Re: dumb question; mips multi-abi

[Daniel Jacobowitz]

On Thu, Mar 17, 2005 at 11:52:06AM -0800, Joel Brobecker wrote:
> But then again, this setting is only used when initializing the
> gdbarch structure...

But setting it switches the gdbarch.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: dumb question; mips multi-abi

[Joel Brobecker]

> > But then again, this setting is only used when initializing the
> > gdbarch structure...
> 
> But setting it switches the gdbarch.

Aaaahhhhh, this is so nice. I missed the "mips_abi_update" part.
Really nice!

-- 
Joel