Re: [PATCH] gdb/microblaze-tdep.c: Check whether less than zero in conditional expression

[Chen Gang <gang.chen.5i5j@gmail.com>]

On 07/24/2014 10:24 AM, Michael Eager wrote:
> On 07/23/14 18:58, Chen Gang wrote:
>>
>>
>> On 07/24/2014 06:20 AM, Chen Gang wrote:
>>> On 07/24/2014 06:17 AM, Michael Eager wrote:
>>>> On 07/23/14 15:06, Chen Gang wrote:
>>>>> On 07/24/2014 04:04 AM, Michael Eager wrote:
>>>>>> On 07/20/14 07:03, Chen Gang wrote:
>>>>>>> Use typecast 'size_t' on 'reg', not only avoid the related
>>>>>>> warning, but
>>>>>>> also check whether less than zero -- for 'reg' is type 'int', and
>>>>>>> sizeof
>>>>>>> (dwarf2_to_reg_map) is less than 0x7fff.
>>>>>>>
>>>>>>> It is quoted in gdb_assert(), so need check 'reg' whether less
>>>>>>> than zero.
>>>>>>> And the related warning (with '-W'):
>>>>>>>
>>>>>>>      ../../binutils-gdb/gdb/microblaze-tdep.c:667:3: error:
>>>>>>> comparison between signed and unsigned integer expressions
>>>>>>> [-Werror=sign-compare]
>>>>>>>
>>>>>>> ChangeLog:
>>>>>>>
>>>>>>>     * microblaze-tdep.c (microblaze_dwarf2_reg_to_regnum): Check
>>>>>>> whether
>>>>>>>     less tha zero in conditional expression.
>>>>>>>
>>>>>>>
>>>>>>> Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
>>>>>>> ---
>>>>>>>     gdb/microblaze-tdep.c | 2 +-
>>>>>>>     1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>>>
>>>>>>> diff --git a/gdb/microblaze-tdep.c b/gdb/microblaze-tdep.c
>>>>>>> index 7e89241..9bec260 100644
>>>>>>> --- a/gdb/microblaze-tdep.c
>>>>>>> +++ b/gdb/microblaze-tdep.c
>>>>>>> @@ -664,7 +664,7 @@ static int dwarf2_to_reg_map[78] =
>>>>>>>     static int
>>>>>>>     microblaze_dwarf2_reg_to_regnum (struct gdbarch *gdbarch, int
>>>>>>> reg)
>>>>>>>     {
>>>>>>> -  gdb_assert (reg < sizeof (dwarf2_to_reg_map));
>>>>>>> +  gdb_assert ((size_t) reg < sizeof (dwarf2_to_reg_map));
>>>>>>>       return dwarf2_to_reg_map[reg];
>>>>>>>     }
>>>>>>
>>>>>> I don't see anything in the patch which does what you describe,
>>>>>> checking whether reg is less than zero.  Converting a signed
>>>>>> integer to an unsigned integer is not a way to check whether
>>>>>> it is less than zero.  This is better:
>>>>>>
>>>>>> + gdb_assert (reg >= 0 && (size_t) reg < sizeof (dwarf2_to_reg_map));
>>>>>>
>>>>>
>>>>> Yeah, it is common statement. It is also OK to me, although after type
>>>>> cast, 'reg >=0' can be omited (it can let code simpler, but let code
>>>>> not quit easy understanding).
>>>>
>>>> No, if you want to verify that the value is greater than zero,
>>>> this cannot be omitted.  A negative value would converted to
>>>> a positive value by the cast.  There no reason to believe that
>>>> this would cause the other half of the test to fail.
>>>>
>>>
>>> When an 'int' negative value converted to a positive value, it will be
>>> larger than 0x7fff which must be larget than 'sizeof
>>> (dwarf2_to_reg_map)'.
>>>
>>
>> If what I said is correct, your idea/suggestions is still OK to me: easy
>> understanding has higher priority than keeping source code simple.
> 
> Yes, you are correct.  Took me a moment to think through.
> 

OK, thank you for your work.

Thanks.
-- 
Chen Gang

Open, share, and attitude like air, water, and life which God blessed