> Alan Hourihane<alanh@...> writes:
>> On 19/01/12 14:38, Andreas Schwab wrote:
>>> Alan Hourihane<alanh@...> writes:
>>>> Fixed the test, as it was broken.
>>> No, it isn't.
>> Then please explain what is ?
> It is the resposibility of ffi_call to put the correct value in the
> return value argument. That's the whole point of the test.
But it does. Just with -O0 it works, and with -O2 it doesn't because
GCC's optimizer steps in.