[Bug tdep/29409] [gdb, tdep/aarch64] FAIL: gdb.opt/inline-small-func.exp: info breakpoints

["vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29409

--- Comment #9 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #7)
> Interestingly, that alone was not enough, and I also needed this bit:
> ...
> @@ -302,6 +303,8 @@ aarch64_analyze_prologue (struct gdbarch *gdbarch,
>      regs[i] = pv_register (i, 0);
>    pv_area stack (AARCH64_SP_REGNUM, gdbarch_addr_bit (gdbarch));
>  
> +  CORE_ADDR orig_start = start;
> +
>    for (; start < limit; start += 4)
>      {
>        uint32_t insn;
> @@ -544,7 +547,7 @@ aarch64_analyze_prologue (struct gdbarch *gdbarch,
>      }
>  
>    if (cache == NULL)
> -    return start;
> +    return seen_stack_set ? start : orig_start;
>  
>    if (pv_is_register (regs[AARCH64_FP_REGNUM], AARCH64_SP_REGNUM))
>      {
> @@ -583,7 +586,7 @@ aarch64_analyze_prologue (struct gdbarch *gdbarch,
>         cache->saved_regs[i + regnum + AARCH64_D0_REGNUM].set_addr (offset);
>      }
>  
> -  return start;
> +  return seen_stack_set ? start : orig_start;
>  }
>  
>  static CORE_ADDR
> ...
> which is roughly equivalent with this bit for amd64:
> ...
>   if (cache.frameless_p)
>     return start_pc;
> ...

Hmm, this seems to break (in test-case gdb.base/call-rt-st.exp) returning
0x4008e0 for:
...
00000000004008dc <loop_count>: 
  4008dc:       d10043ff        sub     sp, sp, #0x10
  4008e0:       b9000fff        str     wzr, [sp, #12]
  4008e4:       14000004        b       4008f4 <loop_count+0x18>
...
and this fixes it:
...
@@ -336,7 +339,8 @@ aarch64_analyze_prologue (struct gdbarch *gdbarch,
            }                                                                   

          /* Did we move SP to FP?  */                                          
-         if (rn == AARCH64_SP_REGNUM && rd == AARCH64_FP_REGNUM)               
+         if (rn == AARCH64_SP_REGNUM                                           
+             && (rd == AARCH64_FP_REGNUM || rd == AARCH64_SP_REGNUM))          
            seen_stack_set = true;                                              
        }                                                                       
       else if (inst.opcode->iclass == pcreladdr                                
...
but I'm not sure about this.  Maybe !frameless_p and seen_stack_set are
different concepts and need to be tracked separately?

-- 
You are receiving this mail because:
You are on the CC list for the bug.