[erlang-questions] Erlang RB12-1 Mac OS X Leopard problems

[Vlad Balin]

Thanks a lot. This file cannot be compiled on Leopard/PPC. May be I'm doing
something wrong, but:


static void *fpu_fpe_enable(void *arg)

{

    thread_t t = *(thread_t*)arg;

    struct ppc_thread_state state;

    unsigned int state_size = PPC_THREAD_STATE_COUNT;

error: 'struct __darwin_mcontext' has no member named 'ss'


...


#elif defined(__i386__)

    mcontext_t mc = uc->uc_mcontext;

    struct __darwin_i386_float_state *fpstate = &mc->__fs;

    fpstate->__fpu_mxcsr = 0x1F80;

    *(unsigned short *)&fpstate->__fpu_fsw &= ~0xFF;

#elif defined(__ppc__)

    mcontext_t mc = uc->uc_mcontext;

    mc->ss.srr0 += 4;

error: 'struct __darwin_mcontext' has no member named 'ss'

    mc->fs.fpscr = 0x80|0x40|0x10;

error: 'struct __darwin_mcontext' has no member named 'ss'



#endif

2008/2/28, Mikael Pettersson <mikpe@REDACTED>:
>
> Vlad Balin writes:
>   > Hello,
>   > I've got a bit different problem with last OTP release on MacOS 10.5Leopard
>   > (iMac G5, ppc, 32-bit mode).
>   >
>   > I can't build HiPE for some reason. Configure script says that there
>   > floating point exceptions are unreliable and this is the reason to
> disable
>   > HiPE on ppc. However, I've got no problems with HiPE on MacOS 10.4 on
> the
>   > same machine. Is there any specific reasons that HiPE is disabled, or
> this
>   > is bug?
>
>
> The configure script reporting unreliable floating-point exceptions
> most probably means that the FPE test program couldn't be compiled
> on OSX 10.5, which in turn is probably because Apple has seen fit to
> make incompatible changes between 10.4 and 10.5.
>
> Someone with access to 10.5 will need to update the FPE test program
> embedded in erts/configure.in to work also on 10.5.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20080228/adbc9b50/attachment.htm>