This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [Fwd: [Bug bootstrap/27644] [4.1 regression] Bootstrap failure on native ARM targets]
Alan Modra wrote:
> On Thu, May 18, 2006 at 12:41:49AM -0700, Mark Mitchell wrote:
>> Andrew Pinski wrote:
>>> On May 17, 2006, at 11:44 PM, Mark Mitchell wrote:
>>>> How does PowerPC communicate 32-bit vs. 64-bit mode to the assembler and
>>>> other tools?
>>> well for one, it creates two different versions of ASM_SPEC.
>>> One for defaulting to 32bit and one for defaulting to 64bit.
>>> Which is what x86 really should be doing anyways.
>> Then the assembler has to use the same default as the compiler, which is
>> not good.
>
> No, that isn't the case. The compiler always passes -a32 or -a64 to the
> assembler, even when the compiler itself is producing its default
> output.
I see. That's neat.
I still think HJ's idea is sound, though; if one option cancels another,
then we might as well let the driver do it, and simplify the lives of
specs writers. Your solution might be a good fix for the problem on the
4.1 branch, if we're afraid of HJ's changes there.
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713