tjt@lll-tis.arpa (Tim Tessin) (08/27/87)
Has anyone ported 1.2 to a MIPS? I am trying to bring it up and it seems
to work (sort of). I keep getting nasty messges from cfront about
complex expressions in a ostream construct. To wit:
cerr << "blah blah" << sizeof (struct hdr) << "blah blah\n";
cfront tells me (paraphrase):
"nasty Internal error at line xxx, blah blah, core dump"
replacing the sizeof expression with a simple variable allows cfront
to continue. (The (bool?expr:expr) construct also screws up likewise.)
Problems I found in MIPS environment to watch for: The mips compiler
doesn't define the appropriate flags for machine dependent stuff in the
cpp ( #define unix, #define mips, etc.). They do it in the cc program.
Of course, all include files dependent on the mips architecture fail
using a simple CC script. I used cc -E instead of cpp in the CC script and
improved things alot! Don't use .C extensions for C++ files. The compiler
only defines the flag "LANGUAGE_C" when the extension is .c. "LANGUAGE_C"
is used in "signal.h" no less, and it fails to provide signal definitions if
you simply compile a .C file including "signal.h".
Size of various word/float boundaries is almost the same as the vax,
I don't think this is where the problems are.
I have some other flakey problems with c++ applications programs
(like segmentation violations) but I don't have enough info yet to
determine whether cfront is at fault.
Tim Tessin - Lawrence Livermore National Laboratory
Phone: (415) 423-4560 / 422-8971
ARPA: tjt@lll-tis-b.ARPA
UUCP: {ihnp4,dual,sun}!lll-lcc!lll-tis!tjt