Re: [PATCH 4/4] gdb/riscv: disassemble an attribute-less object by the description ISA

[Kevin Buettner <kevinb@redhat.com>]

Hi Farid,

A nit...

On 10/22/26 8:44 AM, Farid Khaidari wrote:

>  static int
> @@ -4212,6 +4251,37 @@ riscv_print_insn (bfd_vma addr, struct disassemble_info *info)
>    if (s != nullptr)
>      info->section = s->the_bfd_section;
>  
> +  /* An object that records its own ISA is disassembled according to
> +     that ISA.  One that does not, such as a core file loaded without an
> +     executable, is disassembled according to the target description;
> +     libopcodes would otherwise fall back to plain rv64gc.  ARCH_OPTION
> +     must outlive the default_print_insn call below, which parses it.
> +
> +     The user options are read from the gdbarch and appended here.
> +     libopcodes clears INFO->DISASSEMBLER_OPTIONS once it has parsed
> +     them, so the string has to be supplied again for the next
> +     instruction; each parse first resets every option it does not see
> +     back to its default, which is why the user ones have to be
> +     repeated rather than parsed once.  They come last and so take
> +     precedence.  */
> +  std::string arch_option;
> +  auto *di = static_cast<gdb_disassemble_info *> (info->application_data);
> +  if (di != nullptr && !riscv_disasm_bfd_has_isa_attribute (info))
> +    {
> +      gdbarch *arch = di->arch ();
> +      riscv_gdbarch_tdep *tdep = gdbarch_tdep<riscv_gdbarch_tdep> (arch);
> +
> +      arch_option = tdep->disassembler_options_arch;
> +
> +      const char *user_options = get_disassembler_options (arch);
> +      if (user_options != nullptr && *user_options != '\0')
> +	{
> +	  arch_option += ',';
> +	  arch_option += user_options;
> +	}
> +      info->disassembler_options = arch_option.c_str ();

This points info->disassembler_options at storage owned by a local
std::string, which is destroyed when riscv_print_insn returns.  It
only works because print_insn_riscv() sets disassembler_options to
NULL after parsing them.  (There is an early return before that, when
riscv_init_disasm_info fails, but that function currently always
returns true.)  Relying on that libopcodes behaviour seems fragile:
if print_insn_riscv ever stops clearing the field, or returns early
before doing so, the next instruction will see a dangling pointer.

Kevin