Re: -m64 for darwin

[Mike Stump <mrs at apple dot com>]

On Mar 22, 2005, at 9:22 AM, Geoffrey Keating wrote:
> Mike Stump <mrs@apple.com> writes:
> > Here is additional -m64 multilib support for darwin8.  Without these,
> > libffi and boehm-gc just don't build with -m64.

> > Ok?
>
> This is OK, assuminng you ran the java and libffi testsuites.

:-(  I'll need clarification then...

libjava doesn't yet build:

g++ -m64 -DHAVE_CONFIG_H -I. -I../../../../gcc/libjava -I./include - 
I./gcj -I../../../../gcc/libjava -Iinclude -I../../../../gcc/libjava/ 
include -I../../../../gcc/libjava/../boehm-gc/include -I../boehm-gc/ 
include -I../../../../gcc/libjava/libltdl -I../../../../gcc/libjava/ 
libltdl -I../../../../gcc/libjava/.././libjava/../gcc -I../../../../ 
gcc/libjava/../zlib -I../../../../gcc/libjava/../libffi/include -I../ 
libffi/include -fno-rtti -fnon-call-exceptions -fdollars-in- 
identifiers -Wswitch-enum -D_FILE_OFFSET_BITS=64 -Wextra -Wall - 
D_GNU_SOURCE -DPREFIX=\"/Volumes/mrs3/Packages/gcc-20050128\" - 
DLIBDIR=\"/Volumes/mrs3/Packages/gcc-20050128/lib\" -DBOOT_CLASS_PATH= 
\"/Volumes/mrs3/Packages/gcc-20050128/share/java/libgcj-4.1.0.jar\" - 
DJAVA_EXT_DIRS=\"/Volumes/mrs3/Packages/gcc-20050128/share/java/ext\"  
-DLIBGCJ_DEFAULT_DATABASE=\"/Volumes/mrs3/Packages/gcc-20050128/lib/ 
gcj-4.1.0/classmap.db\" -g -O2 -m64 -MT stacktrace.lo -MD -MP - 
MF .deps/stacktrace.Tpo -c ../../../../gcc/libjava/stacktrace.cc  - 
fno-common -DPIC -o .libs/stacktrace.o
../../../../gcc/libjava/stacktrace.cc:39: error: base  
'_Jv_UnwindState' with only non-default constructor in class without  
a constructor
make[4]: *** [stacktrace.lo] Error 1


For the parts that do build, it is a monotonic improvement.

For the libffi testsuite, I get:

                === libffi Summary ===

# of unexpected failures        80
# of unsupported tests          2

which mirrors what I got before.  Just in case you're curious what's  
in the log:

Running target unix
Using /Volumes/mrs3/net/gcc/libffi/testsuite/config/default.exp as  
tool-and-target-specific interfa
ce file.
Running /Volumes/mrs3/net/gcc/libffi/testsuite/libffi.call/call.exp ...
set_ld_library_path_env_vars: ld_library_path=.:/Volumes/mrs3/net/gcc- 
darwin/gcc:/Volumes/mrs3/net/
gcc-darwin/powerpc-apple-darwin8.0.0/./libffi/.libs:/Volumes/mrs3/net/ 
gcc-darwin/powerpc-apple-darw
in8.0.0/./libstdc++-v3/src/.libs
FAIL: libffi.call/closure_fn0.c (test for excess errors)
Excess errors:
default_target_compile: No compiler to compile with
WARNING: libffi.call/closure_fn0.c compilation failed to produce  
executable
FAIL: libffi.call/closure_fn1.c (test for excess errors)
Excess errors:
default_target_compile: No compiler to compile with
WARNING: libffi.call/closure_fn1.c compilation failed to produce  
executable
FAIL: libffi.call/closure_fn2.c (test for excess errors)

which, it has been doing for the past 9 months or so.

So, if you had asked about no regressions on darwin8, that much is  
true; though, the testsuite status just sucks currently.

If you want me to fix all problems before fixing one problem, there  
seems to be at least some people that prefer things to be really  
broken, let me know, and I'll give up now.  My hope is that we can  
improve it, one bug at a time, until it works again, as well as it  
used to.  These patches don't take us all the way there, only part way.

Ok to check in?