GCC 6.2 toolchain for AmigaOS 3

[Marlon_]

Originally Posted by emufan

that's pretty cool. this may help with other project depending on cmake.

were you able to finish the whole yaml-cpp build, since it doesnt find includes in:
/opt/m68k-amigaos/lib/gcc/m68k-amigaos/6.4.1b/include/c++/tr1/wctype.h
wctype.h / wchar.h
I thought those paths are somehow gcc internal, do we have to included it in CXXFLAGS?

http://eab.abime.net/showpost.php?p=...&postcount=718

[emufan]

Originally Posted by Marlon_

http://eab.abime.net/showpost.php?p=...&postcount=718

#1) I saw the answers added above

is there a cmake option to list all those defintions ( -D .... ) for the given CMakeLists.txt?

[Marlon_]

Originally Posted by emufan

#1) I saw the answers added above

is there a cmake option to list all those defintions ( -D .... ) for the given CMakeLists.txt?

Lets move this discussion here: http://eab.abime.net/showthread.php?t=90570&page=4

[wawa]

@bebbo

btw, by chance, you didnt made any attempt to make gcc deal with motorola asm syntax so far?

[bebbo]

Originally Posted by emufan

I guess not, everything removed before I build new toolchain.
maybe something with my selfmade Makefile, since m68k-amigaos-ranlib.exe works with cmake build lib,
which uses the same gcc toolchain binaries.

note that you must not append trash data to a hunk library file - some people used to append a version string to the lib...

[bebbo]

btw: amiga-gcc has now support for the sdks.

[Marlon_]

Originally Posted by bebbo

btw: amiga-gcc has now support for the sdks.

sweet, there's no need for the old toolchain then!

[emufan]

Originally Posted by bebbo

note that you must not append trash data to a hunk library file - some people used to append a version string to the lib...

cannot say, if I did so, here is my selfmade makefile for yaml-cpp:

CXX=m68k-amigaos-g++.exe
AR=m68k-amigaos-ar.exe
RANLIB=m68k-amigaos-ranlib.exe
CXXFLAGS=-noixemul -c -Wall -Iinclude
LDFLAGS=
SOURCES=src/binary.cpp \
	src/convert.cpp \
	src/directives.cpp \
	src/emit.cpp \
	src/emitfromevents.cpp \
	src/emitter.cpp \
	src/emitterstate.cpp \
	src/emitterutils.cpp \
	src/exceptions.cpp \
	src/exp.cpp \
	src/memory.cpp \
	src/node.cpp \
	src/node_data.cpp \
	src/nodebuilder.cpp \
	src/nodeevents.cpp \
	src/null.cpp \
	src/ostream_wrapper.cpp \
	src/parse.cpp \
	src/parser.cpp \
	src/regex_yaml.cpp \
	src/scanner.cpp \
	src/scanscalar.cpp \
	src/scantag.cpp \
	src/scantoken.cpp \
	src/simplekey.cpp \
	src/singledocparser.cpp \
	src/stream.cpp \
	src/tag.cpp \
	src/contrib/graphbuilder.cpp \
	src/contrib/graphbuilderadapter.cpp

OBJECTS=$(SOURCES:.cpp=.o)
EXECUTABLE=yaml-cpp
LIBRARY=libyaml-cpp.a

# all: $(SOURCES) $(EXECUTABLE) $(LIBRARY)
all: $(SOURCES) $(LIBRARY)

$(EXECUTABLE): $(OBJECTS)
	$(CXX) $(LDFLAGS) $(OBJECTS) -o $@

$(LIBRARY): $(OBJECTS)
	$(AR) cru libyaml-cpp.a  src/binary.o src/convert.o src/directives.o src/emit.o src/emitfromevents.o src/emitter.o src/emitterstate.o src/emitterutils.o src/exceptions.o src/exp.o src/memory.o src/node.o src/node_data.o src/nodebuilder.o src/nodeevents.o src/null.o src/ostream_wrapper.o src/parse.o src/parser.o src/regex_yaml.o src/scanner.o src/scanscalar.o src/scantag.o src/scantoken.o src/simplekey.o src/singledocparser.o src/stream.o src/tag.o src/contrib/graphbuilder.o src/contrib/graphbuilderadapter.o
	$(RANLIB) libyaml-cpp.a

cpp.o:
	$(CXX) $(CXXFLAGS) $< -o $@

clean :
	-rm -r src/*.o libyaml-cpp.a

Originally Posted by bebbo

btw: amiga-gcc has now support for the sdks.

yep, good news. will try this version next time

[bebbo]

Originally Posted by wawa

@bebbo

btw, by chance, you didnt made any attempt to make gcc deal with motorola asm syntax so far?

gcc *is* creating motorola asm code
as *does* assemble motorola asm code

Only objdump does not yet produce morotola asm code. And it is dumb work with little value for me...

[bebbo]

Originally Posted by emufan

cannot say, if I did so, here is the makefile:

...

that's looking ok to me - no idea.

note: ranlib can actually be omitted...

[emufan]

note: ranlib can actually be omitted...

ok, good to know.

Originally Posted by bebbo

that's looking ok to me - no idea.

I just extracted the yaml-cpp source somewhere else, put my makefile in that folder.
now it does build without ranlib error.

$ m68k-amigaos-ar.exe t libyaml-cpp.a
binary.o
convert.o
....
graphbuilderadapter.o

maybe there was some cygwin stuff left from my previous tests *argh*

[bebbo]

Originally Posted by emufan

ok, good to know.

I just extracted the yaml-cpp source somewhere else, put my makefile in that folder.
now it does build without ranlib error.
maybe there was some cygwin stuff left from my previous tests *argh*

nice

(maybe it was an arch object file? I think you can add it to a hunk lib but the hunk ranlib can't parse it)

I used cygwin too, but now I'm often using a Linux/VM since it's way faster.

[emufan]

Originally Posted by bebbo

nice

yep, good to see things are working as they should

(maybe it was an arch object file? I think you can add it to a hunk lib but the hunk ranlib can't parse it)

it was build with the very same binaries and makefile.

I used cygwin too, but now I'm often using a Linux/VM since it's way faster.

I should consider that too, with all the bloatware on that windows box, it slows down things alot,
and does crash whole windows, using "-j4" "--threads 4" ...
have to reboot to windows safe mode (f8) - 30 mins vs xx hours :/

[wawa]

Originally Posted by bebbo

gcc *is* creating motorola asm code
as *does* assemble motorola asm code

Only objdump does not yet produce morotola asm code. And it is dumb work with little value for me...

just tested it and apparently your gcc undestands motorola syntax better then the aros one. id love to borrow the code in question for aros68k if i may, i was looking for it in binutils gas (im diffing it with aros sources), but cannot locate the responsible section. any hint where to find it?

btw. apparently it still doesnt parse semicolon as a comment token.

[bebbo]

Originally Posted by wawa

just tested it and apparently your gcc undestands motorola syntax better then the aros one. id love to borrow the code in question for aros68k if i may, i was looking for it in binutils gas (im diffing it with aros sources), but cannot locate the responsible section. any hint where to find it?

btw. apparently it still doesnt parse semicolon as a comment token.

the semicolon is often used to separate multiple insns in same line

=> can't change it without hitting to much

And I guess, you simply have to define

#define REGISTER_PREFIX_OPTIONAL 1

EDIT:

objdump displays now MOTOROLA syntax.

> m68k-amigaos-objdump.exe  -dA ins.o
ins.o:     file format amiga

Disassembly of section .text:

LLL:
        asl.w #2,d2
        moveq #1,d0
        move.l d0,(a0)+
        move.l d0,-(a0)
        movea.w d0,a0
        move.l d0,-123(a0)
        move.l d0,123(a0)
        move.l d0,(42,a0,d1.w)
        move.l d0,(42,a0,d1.w*4)
        move.l d0,([42,a0],d1.w*4,7)
        move.l d0,([42,a0,d1.w*4],7)
        lea LLL+0xa7(pc),a0
        bne.s LLL+0x44
        move.l (LLL+0x5c,pc,d1.w),d0
        move.l ([LLL+0x60,pc],d1.w*4,3),d0
        move.l ([LLL+0x68,pc,d1.w*4],3),d0
        rts

[wawa]

Originally Posted by bebbo

the semicolon is often used to separate multiple insns in same line

=> can't change it without hitting to much

but at the beginning of the line semicolon is clearly meant as a comment, i believe? parsing this shouldnt break anything i quess.

[wawa]

btw:

error: no such remote ref f494ffb0f9f668a9111d7ad3afa2d7a50a348a53
Fetched in submodule path 'submodules/binutils-2.14', but it did not contain f494ffb0f9f668a9111d7ad3afa2d7a50a348a53. Direct fetching of that commit failed.
ERROR: command "git submodule update" failed with 1

any idea what may be wrong? i tried to pull and clean rebuild every way i could think of..

[Marlon_]

Originally Posted by wawa

btw:

error: no such remote ref f494ffb0f9f668a9111d7ad3afa2d7a50a348a53
Fetched in submodule path 'submodules/binutils-2.14', but it did not contain f494ffb0f9f668a9111d7ad3afa2d7a50a348a53. Direct fetching of that commit failed.
ERROR: command "git submodule update" failed with 1

any idea what may be wrong? i tried to pull and clean rebuild every way i could think of..

Are you using this: https://github.com/bebbo/amigaos-cross-toolchain

Or are you using this: https://github.com/bebbo/amiga-gcc

?

[wawa]

amigaos-cross-toolchain afa i see..

[Marlon_]

Originally Posted by wawa

amigaos-cross-toolchain afa i see..

I recommend using amiga-gcc as it's essentially the same, but compiles much faster as it supports parallell compilation. I'm compiling it all with "make -j8" and it takes no time to compile it all. Cygwin might have an issue with it, but as long as you're compiling native on Linux or Mac, there shouldn't be an issue.