Re: [AArch64][3/6] DWARF unwinder support for signed return address

[Jiong Wang <jiong dot wang at foss dot arm dot com>]

On 10/08/17 12:08, Pedro Alves wrote:
> On 08/09/2017 01:17 PM, Jiong Wang wrote:
> > +
> > +  /* Only DW_CFA_GNU_window_save is expected on AArch64.  */
> > +  if (op != DW_CFA_GNU_window_save)
> > +    return false;
> > +
> Was there any progress in giving this an Aarch64-specific name
> to avoid confusion?  FYI, I was confused by this again for a couple
> minutes when I read it, until I remembered the previous discussions.

Hi Pedro,

  Yes, We can add DW_CFA_AARCH64_negate_ra_state to dwarf2.def, with the same
  number of DW_CFA_GNU_window_save, then need to introduce a new DW_CFA_DUP
  to avoid duplicated case value trouble when auto-generate switch/case for
  returning CFA name.  Will post patch to GCC then can be synced to binutils-gdb
  onced approved.

> > +/* Implement the execute_dwarf_cfa_vendor_op method.  */
> > +
> > +static bool
> > +aarch64_execute_dwarf_cfa_vendor_op (struct gdbarch *gdbarch, gdb_byte op,
> > +				     struct dwarf2_frame_state *fs)
> > +{
> > +  struct dwarf2_frame_state_reg *ra_state_column;
> > +  static unsigned char exp_0 = 0x30; /* DW_OP_lit0.  */
> > +  static unsigned char exp_1 = 0x31; /* DW_OP_lit1.  */
> Can these be static const?  (likewise elsewhere in the patch.)
>
> Also, gdb_byte.

Will fix in the next update.


Thanks.

Regards,
Jiong