I am trying to fix it, but the core may be too far into advanced C++ features for the old GCC 4.0.1 compiler to handle.

OS X Intel is fine.

I confirm that. This time the bad beasts seems to be nycaptor.c and rendersw.c
With NOWERR=1 option I am been able to skip the "nycaptor.c" error but went stuck at "rendersw.c" part.
I cross my fingers hoping the bugs will be ironed out... :-)

Yeah, nycaptor's already fixed. rendersw looks much harder

Good news, I found the problem and u8 will again compile on PPC OS X, with the usual tms57002 workaround required.

GREAT news! :-) Compliments, really... uhm... am I the only MAME-addicted PPC Mac user left? I hope the other users will take a moment for dropping a line here and say a big THANK YOU to this man for his efforts in supporting our little community! ;-)

Haven't tried building on PS3 Linux lately (also PPC) but may give it a shot this weekend. Thanks RB!

I'm trying to hack something in for u7 so it automatically compiles tms57002 -O0 on non-x86/x64, but it looks like GCC didn't support #pragma optimize until 4.4 which makes that a lot harder.

The following should do the job (sdl.mak)

u8 builds OK on Mac PPC but only with option NOWERROR=1 because of this warning during compile:

src/emu/ioport.c: In member function ‘void natural_keyboard::build_codes(ioport_manager&)’:
src/emu/ioport.c:1237: warning: ‘newcode$field$1’ may be used uninitialized in this function

And, of course, the option OPTIMIZE=0 when the TMS57002 part is reached.

Whoops. I somehow completely missed "vetting" Aaron's input changes. I'll get that fixed and try out couriersud's tms57002 thing for next time.