This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Fix alignment of disassemble /r
- From: Daniel Gutson <daniel dot gutson at tallertechnologies dot com>
- To: Yao Qi <yao at codesourcery dot com>
- Cc: gdb-patches <gdb-patches at sourceware dot org>
- Date: Thu, 17 Apr 2014 11:37:07 -0300
- Subject: Re: [PATCH] Fix alignment of disassemble /r
- Authentication-results: sourceware.org; auth=none
- References: <CAF5HaEUBbK4d2reSxUpUQwp7Zt_KjcpNC_nbUrMHRb0yNo9wtw at mail dot gmail dot com> <534F294D dot 4050907 at codesourcery dot com>
On Wed, Apr 16, 2014 at 10:07 PM, Yao Qi <yao@codesourcery.com> wrote:
> On 04/12/2014 06:24 AM, Daniel Gutson wrote:
>> when disassembling in raw mode (/r) in a variable-length insn
>> architecture (i.e. x86),
>> the output can be completely messed since no alignment takes place.
>
> The /r output is messed, but not completely :).
>
>>
>> I am aware of the uiout->table stuff, but it seems an overkill since I
>> should change
>> the current_uiout when disassembling in this mode (and I didn't find
>> any actual use of this
>> machinery at least for x86).
>> Therefore, I added a hack in the dump_insns when the /r flag is specified.
>> This clearly isn't the cutiest thing in the world, and I specified a
>> hardcoded maximum number
>> of opcode bytes to align (currently 8) though it is easily changeable.
>
> Hard-coded opcode length will affect other targets. For example, without
> your patch, the disassembly for c6x is like,
> (gdb) disassemble /r main
> Dump of assembler code for function main:
> 0x00000860 <+0>: c2 1b be 07 subah .D2 b15,16,b15
> 0x00000864 <+4>: f6 02 3d 07 stw .D2T2 b14,*+b15(32)
>
> with your patch applied, it becomes
>
> (gdb) disassemble /r main
> Dump of assembler code for function main:
> 0x00000860 <+0>: c2 1b be 07 subah .D2 b15,16,b15
> 0x00000864 <+4>: f6 02 3d 07 stw .D2T2 b14,*+b15(32)
>
> gdbarch_max_insn_length can tell us the max instruction length, but if we
> align the instruction to the max length (it is 16 on x86), the text
> instruction will be far behind the hex code, which is a little ugly.
Hi Yao, thanks for answering.
So what if I add a new configuration variable, such as
set disassemble-raw-alignment
with "off" as default, and if set to on, pad to gdbarch_max_insn_length ?
>
> --
> Yao (éå)
--
Daniel F. Gutson
Chief Engineering Officer, SPD
San Lorenzo 47, 3rd Floor, Office 5
CÃrdoba, Argentina
Phone: +54 351 4217888 / +54 351 4218211
Skype: dgutson