Re: RE: Segfault with Faster MDCT patch and -fPIC

[Dave Hooper]

Right exactly (and changing pic register or changing one hardcoded register
for another definitely hacks). I'll try and take a look later on but should
just be a case of stopping trying to define/reuse these register variables
between macros and have clearly defined scopes that gcc can understand in
the usual way.
>
>
>> I suspect gcc is failing to stack it because it thinks I want to modify
>> it. The explicit register("r10") stuff is a bit of a hack really, and I
>> could easily see that gcc interprets my intentions differently than I
>> had originally assumed.
>>
>
>
> IMO hacks aren't the best idea since I want to get this into libtremor,
which means it has to compile with more then just gcc.