Finally: Mac Intel patches

Mikael Pettersson <>
Thu Aug 24 15:53:20 CEST 2006


Joel Reymont writes:
 > 
 > On Aug 24, 2006, at 9:51 AM, Mikael Pettersson wrote:
 > 
 > > First run 'nm' on sys_float.o: Is erts_restore_fpu defined there?
 > > And if so, is the name prefixed by '_' or not.
 > >
 > > If it isn't defined there then that's the problem because lots
 > > of places in the runtime system (not just HiPE) reference it.
 > 
 > nm ./obj/i386-apple-darwin8.7.1/opt/hybrid/sys_float.o
 >           U ___error
 > 00000000 T _erts_sys_init_float
 > 0000001c T _erts_thread_init_float
 >           U _finite
 > 000001a6 T _matherr
 > 00000010 C _saved_program_buf
 >           U _sprintf
 >           U _strtod
 > 00000088 T _sys_chars_to_double
 > 00000021 T _sys_double_to_chars
 >           U _sys_sigset

It looks like NO_FPE_SIGNALS are being set, which would happen
if erts/configure failed to determine that they're working.
(I changed the Darwin/x86 part of erts/configure and it's
entirely possible that it got broken.)

What does `fgrep NO_FPE_SIGNALS erts/*/config.h' return?

If you run `./configure --enable-hipe 2>&1 | tee foo' and
then `fgrep floating foo', what does it say in the check
for unreliable floating point exceptions: reliable or not?



More information about the erlang-questions mailing list