This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: Possible gas regression
On Sun, Sep 5, 2010 at 11:28 AM, Kai Tietz <ktietz70@googlemail.com> wrote:
> 2010/9/5 NightStrike <nightstrike@gmail.com>:
>> On Sun, Sep 5, 2010 at 1:42 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Sun, Sep 5, 2010 at 10:19 AM, NightStrike <nightstrike@gmail.com> wrote:
>>>> We noticed while compiling mingw-w64 using binutils CVS HEAD this
>>>> morning the following failure:
>>>>
>>>> x86_64-w64-mingw32-gcc -DHAVE_CONFIG_H -I.
>>>> -I../../../build/mingw/mingw-w64-crt ?-m64
>>>> -I../../../build/mingw/mingw-w64-crt/include -D_CRTBLD
>>>> -I/buildbot/pac0602-cygwin/cygwin-x86-x86_64/build/build/root/x86_64-w64-mingw32/include
>>>> ?-g -O2 -MT math/lib64_libmingwex_a-scalbnl.o -MD -MP -MF
>>>> math/.deps/lib64_libmingwex_a-scalbnl.Tpo -c -o
>>>> math/lib64_libmingwex_a-scalbnl.o `test -f 'math/scalbnl.S' || echo
>>>> '../../../build/mingw/mingw-w64-crt/'`math/scalbnl.S
>>>> scalbnl.S: Assembler messages:
>>>> scalbnl.S:20: Error: operand type mismatch for `and'
>>>>
>>>> scalbnl.S is:
>>>>
>>> ...
>>>> ? ? ? ?andq ? ?$0xffffffff, %r8
>>>
>>> AND takes an immediate of [-0x80,0x7f] or [-0x80000000,0x7fffffff].
>>> 0xffffffff is out of range.
>>>
>>> --
>>> H.J.
>>>
>>
>> Using H.J's suggested fix of "andq $-1, %r8". ?Thanks, HJ!
>>
>
> Hmm, how can be 0xffffffff out of range for an andq? I understand that
> an operand 0xffffffffffffffff (what is -1) would be out of range.
>
>
Please read Intel64 SDM vol 2a.
--
H.J.