Re: [PATCH] arm reversible : <phase_2_complete>

["Petr Hluzín" <petr.hluzin@gmail.com>]

2011/12/4 oza Pawandeep <oza.pawandeep@gmail.com>:
> Hi,
>
> Updated patch contains all the 3 review comments fixed (Tom, Petr and Yao).
> Change log is elaborated; I am looking forwarding to adding more
> details if need. seeking for your feedback on the same.
>
> diff -urN arm_orig/ChangeLog arm_new/ChangeLog
> --- arm_orig/ChangeLog  2011-12-03 18:05:04.000000000 +0530
> +++ arm_new/ChangeLog   2011-12-04 16:45:00.000000000 +0530
> @@ -1,3 +1,65 @@
> +2011-12-03  Oza Pawandeep   <oza.pawandeep@gmail.com>
> +
> +       * arm-linux-tdep.c: arm_linux_init_abi modified to include
> +          arm reversible debugging feature.
> +          registered arm_process_record to gdb_arch
> +          syscall pointer initialization.
> +
> +       * arm-tdep.c: arm-reversible-debugging implementation.
> +         newly added functions are as follows.
> +
> +          > arm_process_record: handles basic initialization and record
> +             summarisation, decodes basic insn ids, on which it hands over
> +             controls to decode_insn.
> +          > deallocate_reg_mem : clean up function
> +          > decode_insn: Decodes arm/thumb insn and calls appropriate
> +             decoding routine to record the change.
> +          > thumb_record_branch: branch insn reording (thumb)
> +          > thumb_record_ldm_stm_swi: load, store and sycall insn
> +             recoding  (thumb)
> +          > thumb_record_misc: misc insn recording  (thumb)
> +          > thumb_record_ld_st_stack: store and stack insn recording  (thumb)
> +          > thumb_record_ld_st_imm_offset: load, store with immediate offset
> +             insn recording  (thumb)
> +          > thumb_record_ld_st_reg_offset: load, store with register offset
> +             recording  (thumb)
> +          > thumb_record_add_sub_cmp_mov: addition, subtractation, compare
> +             and move insn recording  (thumb)
> +          > thumb_record_shift_add_sub: shift, add and sub insn recording
> +             (thumb)
> +          > arm_record_coproc_data_proc: coprocessor and data processing
> +             recording (partially implemented) (arm)
> +          > arm_record_coproc: coprocessor insn recording
> +             (partially implemented) (arm)
> +          > arm_record_b_bl: branch insn recording (arm)
> +          > arm_record_ld_st_multiple: load and store multiple insn recording
> +             (arm)
> +          > arm_record_ld_st_reg_offset: load and store reg offset recording
> +             (arm)
> +          > arm_record_ld_st_imm_offset: load and store immediate offset
> +             recording (arm)
> +          > arm_record_data_proc_imm: data processing insn recording  (arm)
> +          > arm_record_data_proc_misc_ld_str: data processing, misc, load and
> +             store insn recording  (arm)
> +          > arm_record_extension_space:arm extension space insn recording
> +             (arm)
> +          > arm_record_strx: str(X) type insn recording  (arm)
> +          > sbo_sbz: checks for mendatory sbo and sbz fields in insn,
> +
> +          added new data structures:
> +          > insn_decode_record_t: local record structure which contains insn's
> +          record, which includes both reg and memory.
> +
> +          REG_ALLOC and MEM_ALLOC macros takes care of actual memory allocation
> +          in local record which is finally processed by arm_rocess_record.
> +
> +       * arm-tdep.h: arm-reversible data structures
> +
> +         > modified gdbarch_tdep: added member (function pointer) arm_swi_record
> +            which is supposed to be recording system calls
> +         > arm_process_record externed.
> +
> +

WOW, thats a lot of text. Other people's changelog entries are more
terse. For example the GNU guidelines say "For example, “New function”
is enough for the change log when you add a function, because there
should be a comment before the function definition to explain what it
does."

In function decode_insn():
> +
> +  if (extract_arm_insn (arm_record, insn_size))
> +    {
> +      if (record_debug)
> +        {
> +          printf_unfiltered (_("Process record: error reading memory at "
> +                              "addr %s len = %d.\n"),
> +          paddress (arm_record->gdbarch, arm_record->this_addr), insn_size);
> +          return -1;
> +        }
> +    }

If the extract_arm_insn() call fails then the function returns only if
`record_debug' is true. This is strange. Enabling debugging traces
usually expected to only affect printing, here it affects behavior of
the decode_insn() function. Plus the rest of the function seems to
misbehave if it is passed failed result in `arm_record'. I guess the
`return -1;' statement should be moved outside of body of condition
`if (record_debug)'. Unortunatelly I did not notice that earlier.

In arm_process_record():
> +
> +  if (extract_arm_insn (&arm_record, 2))
> +    {
> +      if (record_debug)
> +        {
> +          printf_unfiltered (_("Process record: error reading memory at "
> +                             "addr %s len = %d.\n"),
> +                             paddress (arm_record.gdbarch,
> +                             arm_record.this_addr), 2);
> +          return -1;
> +        }
> +    }

The same applies here.

Except the one last thing the patch looks good to me now. My previous
suggestions have been resolved.
As always, I did not check whitespace, ARM semantics and Changelog entries.

-- 
Petr Hluzin