[PATCH] x86: Remove duplicated I386_PCREL_TYPE_P/X86_64_PCREL_TYPE_P

[Jan Beulich]

On 05.01.2023 17:55, H.J. Lu wrote:
> On Thu, Jan 5, 2023 at 8:52 AM Jan Beulich <jbeulich@suse.com> wrote:
>> On 05.01.2023 17:50, H.J. Lu wrote:
>>> On Wed, Jan 4, 2023 at 11:42 PM Jan Beulich <jbeulich@suse.com> wrote:
>>>> On 04.01.2023 20:14, H.J. Lu via Binutils wrote:
>>>>> I386_PCREL_TYPE_P and X86_64_PCREL_TYPE_P are defined twice.  Remove
>>>>> the duplications.
>>>>
>>>> I recall noticing this as well, quite some time back, but I didn't feel
>>>> like touching it because I was puzzled by ...
>>>>
>>>>> --- a/bfd/elfxx-x86.h
>>>>> +++ b/bfd/elfxx-x86.h
>>>>> @@ -97,13 +97,6 @@
>>>>>  #define PLT_FDE_START_OFFSET 4 + PLT_CIE_LENGTH + 8
>>>>>  #define PLT_FDE_LEN_OFFSET   4 + PLT_CIE_LENGTH + 12
>>>>>
>>>>> -#define I386_PCREL_TYPE_P(TYPE) ((TYPE) == R_386_PC32)
>>>>
>>>> ... this not including PC8 and PC16 when ...
>>>
>>> This is I386_PCREL_TYPE_P.
>>>
>>>>> -#define X86_64_PCREL_TYPE_P(TYPE) \
>>>>> -  ((TYPE) == R_X86_64_PC8 \
>>>>> -   || (TYPE) == R_X86_64_PC16 \
>>>>> -   || (TYPE) == R_X86_64_PC32 \
>>>>> -   || (TYPE) == R_X86_64_PC64)
>>>>
>>>> ... this does.
>>>
>>> This is X86_64_PCREL_TYPE_P, not I386_PCREL_TYPE_P.
>>>
>>>> Jan
>>>
>>> The current ones have
>>>
>>> #define X86_64_PCREL_TYPE_P(TYPE) \
>>>   ((TYPE) == R_X86_64_PC8 \
>>>    || (TYPE) == R_X86_64_PC16 \
>>>    || (TYPE) == R_X86_64_PC32 \
>>>    || (TYPE) == R_X86_64_PC64)
>>> #define I386_PCREL_TYPE_P(TYPE) ((TYPE) == R_386_PC32)
>>>
>>> and the ones I removed are
>>>
>>> -#define I386_PCREL_TYPE_P(TYPE) ((TYPE) == R_386_PC32)
>>> -#define X86_64_PCREL_TYPE_P(TYPE) \
>>> -  ((TYPE) == R_X86_64_PC8 \
>>> -   || (TYPE) == R_X86_64_PC16 \
>>> -   || (TYPE) == R_X86_64_PC32 \
>>> -   || (TYPE) == R_X86_64_PC64)
>>>
>>> They are identical.
>>
>> That wasn't the question, though. I really did ask about the 32-bit vs
>> 64-bit difference, which looks suspect to me.
>>
> 
> R_386_PC8 and R_386_PC16 were never handled by linker.

May I then ask why that is (or, worded differently, why the two respective
types are handled for x86-64)? Is this just one of the many inconsistencies
that we have?

Jan