Re: [PATCH v8 5/7] Support software single step on ARM in GDBServer

[Antoine Tremblay <antoine.tremblay@ericsson.com>]

On 01/13/2016 11:13 AM, Yao Qi wrote:
> Antoine Tremblay <antoine.tremblay@ericsson.com> writes:
>
>> +
>> +/* Get the raw next possible addresses.  PC in next_pcs is the current program
>> +   counter, which is assumed to be executing in ARM mode.
>> +
>> +   The values returned have the execution state of the next instruction
>> +   encoded in it.  Use IS_THUMB_ADDR () to see whether the instruction is
>> +   in Thumb-State, and gdbarch_addr_bits_remove () to get the plain memory
>> +   address in GDB and arm_addr_bits_remove in GDBServer.  */
>> +
>> +VEC (CORE_ADDR) *
>> +arm_get_next_pcs_raw (struct arm_get_next_pcs *self,
>> +		      CORE_ADDR pc)
>> +{
>> +  int byte_order = self->byte_order;
>> +  unsigned long pc_val;
>> +  unsigned long this_instr = 0;
>> +  unsigned long status;
>> +  CORE_ADDR nextpc;
>> +  struct regcache *regcache = self->regcache;
>> +  VEC (CORE_ADDR) *next_pcs = NULL;
>> +
>> +  pc_val = (unsigned long) pc;
>> +  this_instr = self->ops->read_mem_uint (pc, 4, byte_order);
>                                                     ^^^^^^^^^^
>
>> -
>> -/* Get the raw next address.  PC is the current program counter, in
>> -   FRAME, which is assumed to be executing in ARM mode.
>> -
>> -   The value returned has the execution state of the next instruction
>> -   encoded in it.  Use IS_THUMB_ADDR () to see whether the instruction is
>> -   in Thumb-State, and gdbarch_addr_bits_remove () to get the plain memory
>> -   address.  */
>> -
>> -static CORE_ADDR
>> -arm_get_next_pc_raw (struct regcache *regcache, CORE_ADDR pc)
>> -{
>> -  struct gdbarch *gdbarch = get_regcache_arch (regcache);
>> -  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
>> -  enum bfd_endian byte_order_for_code = gdbarch_byte_order_for_code (gdbarch);
>> -  unsigned long pc_val;
>> -  unsigned long this_instr;
>> -  unsigned long status;
>> -  CORE_ADDR nextpc;
>> -
>> -  pc_val = (unsigned long) pc;
>> -  this_instr = read_memory_unsigned_integer (pc, 4, byte_order_for_code);
>                                                         ^^^^^^^^^^^^^^^^^^^
>
> The code change above introduces a bug as I see.  The original code uses
> byte_order_for_code which is right to me, but it becomes byte_order
> after the change.  Patch below fixes it.
>

Thank you!

Antoine