Sparc V9 disassembly?

Sparc V9 disassembly?

[Raymond Toy]

I've just joined this list so if this question has been asked, forgive
me.  I looked at the recent mailing list archives and didn't see
anything obvious.

I'm trying to debug some Sparc v9 code and gdb's disassembler doesn't
want to print out the new v9 instructions.  Before trying the
disassembly, I gave the following commands to gdb (version 4.18):

info archi
set archi sparc:v9a
disassemble <addr1> <addr2>

It seems the bug is in gdb_print_insn_sparc.  It sets info->mach to
TM_PRINT_INSN_MACH, which is the sparc/sparclite.  If I use gdb to
change this to 8 (sparc v9?), the disassembler recognizes the v9
instructions.  

Is this correct behavior?

Ray

P.S.  This is a Ultra 30 (Ultrasparc IIi), solaris 2.7 machine.  gdb
configured using default options (except --prefix), from a stock
gdb-4.18 tarball.

Re: Sparc V9 disassembly?

[Andrew Cagney]

Raymond Toy wrote:
> 
> I've just joined this list so if this question has been asked, forgive
> me.  I looked at the recent mailing list archives and didn't see
> anything obvious.
> 
> I'm trying to debug some Sparc v9 code and gdb's disassembler doesn't
> want to print out the new v9 instructions.  Before trying the
> disassembly, I gave the following commands to gdb (version 4.18):
> 
> info archi
> set archi sparc:v9a
> disassemble <addr1> <addr2>
> 
> It seems the bug is in gdb_print_insn_sparc.  It sets info->mach to
> TM_PRINT_INSN_MACH, which is the sparc/sparclite.  If I use gdb to
> change this to 8 (sparc v9?), the disassembler recognizes the v9
> instructions.
> 
> Is this correct behavior?

I'd try it with the current GDB however, I suspect that SPARC GDB
currently ignores requests to change the architecture via ``set
architecture''.

	Andrew

Re: Sparc V9 disassembly?

[Raymond Toy]

>>>>> "Andrew" == Andrew Cagney <ac131313@cygnus.com> writes:

    >> 
    >> info archi
    >> set archi sparc:v9a
    >> disassemble <addr1> <addr2>
    >> 
    >> It seems the bug is in gdb_print_insn_sparc.  It sets info->mach to
    >> TM_PRINT_INSN_MACH, which is the sparc/sparclite.  If I use gdb to
    >> change this to 8 (sparc v9?), the disassembler recognizes the v9
    >> instructions.
    >> 
    >> Is this correct behavior?

    Andrew> I'd try it with the current GDB however, I suspect that SPARC GDB
    Andrew> currently ignores requests to change the architecture via ``set
    Andrew> architecture''.

I commented out the offending line in gdb_print_insn_sparc, and
recompiled.  Everything seems to work according to the way I want, so
set architecture does work.  A peek at the code seems to indicate that
as well.

I will, however, try the most recent snapshot.

Since the code has made a conscious effort to set info->mach, I wasn't
really sure if everything else would work.

Ray