[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

hello world for ada compilation (again ...)

Kris Hermans wrote:
> 
> Hello,
> 
> I installed the crosscompiler for gnat/rtems and even succeeded in
> compiling the kernel.
> My system is debian linux, GNAT/RTEMS with target = pc386.

I attached Makefile.pc386.  It worked at one point. :)

What you are trying to do is make gnatmake link applications the
same way the make-exe rule does in make/custom/pc386.cfg.

> Sadly, in the sample app (hello_world_ada), there is no makefile included
> for my set-up.  While trying to figure out one for myself, I have a few
> questions:
> 
> * in the file MakeFile.erc32 there is the line:
>  CARGS=-B${rtemsdir}/lib/ -specs bsp_specs -qrtems -mcypress  I guess this
> defines some flags to pass to the binder and the linker.  However, I don't
> get the actual meaning of it (what is bsp_specs, qrtems, mcypress and how
> do I have to modify it for a pc386 ?)

TO help clarify things...

Everything from the -B ... -qrtems just enables the BSP specific "specs"
file and tells it to use the rules for linking with RTEMS.  bsp_specs
is an extension to the internal gcc rules for building things.

-mcypress is VERY sparc specific -mcpu=i486 would be a comparable 
example from the i386 world.

>
> However, make complains:
> 
> make -f Makefile.pc386.old
> /home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/bin/i386-rtems-gcc -O4 -g  -Wall
> -ansi -fasm -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/
> -specs bsp_specs -qrtems -mcypress -c init.c
> cc1: Invalid option `cypress'
> make: *** [init.o] Error 1

Exactly this specifies a sparc specific option.
 
> * So, I leave out the invalid option (without knowing what it means ...).
> I get:
> 
>  make -f Makefile.pc386.old
> /home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/bin/i386-rtems-gcc -O4 -g  -Wall
> -ansi -fasm -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/
> -specs bsp_specs -qrtems -c init.c
> /home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/bin/i386-rtems-gnatmake -v -O
> -gnata -gnatE -gnato hello -g \
>     -bargs -r \
>     -cargs -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/
> -specs bsp_specs -qrtems \
>     -largs -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/
> -specs bsp_specs -qrtems init.o


> 
> GNATMAKE 3.11p  (981118) Copyright 1995-1998 Free Software Foundation,
> Inc.
>   "hello.ali" being checked ...
>   -> "hello.ali" missing.
> i386-rtems-gcc -c -O -gnata -gnatE -gnato -g
> -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/ -specs
> bsp_specs -qrtems hello.adb
> hello.adb:7:11: warning: file name does not match unit name, should be
> "main.adb"
> i386-rtems-gnatbind -aO./ -r -I- -x hello.ali
> i386-rtems-gnatlink -g
> -B/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/ -specs
> bsp_specs -qrtems init.o hello.ali
> i386-rtems-gcc: file path prefix
> `/home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/rtems/pc386/lib/' never used
> /home/pcuser8/RTEMS/gnat-3.11p/i386-rtems/bin/i386-rtems-size hello
>    text    data     bss     dec     hex filename
>  154464    6076   42188  202728   317e8 hello
> 
> What is the meaning of this all ?  Did I succeed ? Why isn 't there a file
> with a .exe extension (I have just plain "hello", can I use this for
> creating a bootable disk with grub ?).

It looks like you got a hello executable.  Unfortunately, the rule
is not putting the end result in hello.exe as expected.  

DO a "file" on hello.  It appears to be an executable.  If it is the 
same format as the C sample executables, then you should be OK that
much.

Be warned,  the makefile I sent you forces some addresses on the link.
I don't know if your modifications produced the same result.

> Sorry for all the (perhaps silly) qustions, but I really want to be sure
> that my compile set-up is ok, before proceeding with actually executing
> the apps.

These aren't silly.  It is difficult to explain how to make gnatmake
produce executable for RTEMS.

> Thanks for the answers,
> 
> Kris (a RTEMS newbie, but I guess you figured this out already ;-))

You are doing fine.

Building gnat is not for the faint of heart.  Using gnatmake to produce
applications requires careful comparison of what make-exe is doing and
what you can gnatmake is doing.

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985
-------------- next part --------------
#
#  Makefile for hello world example
#

MAIN=hello

# Tool paths
target=i386-rtems
tooldir=/home/joel/gnat-3.10p/${target}/
rtemsdir=${tooldir}/rtems/pc386

# Tool names
GCC=${tooldir}/bin/${target}-gcc
GNATMAKE=${tooldir}/bin/${target}-gnatmake
SIZE=${tooldir}/bin/${target}-size
OBJCOPY=${tooldir}/bin/${target}-objcopy
NM=${tooldir}/bin/${target}-nm

HEADERADDR=0x00097E00
START16FILE=$(rtemsdir)/lib/start16.bin
START16ADDR=0x00097C00
RELOCADDR=0x00100000

CARGS=-B${rtemsdir}/lib/ -specs bsp_specs -qrtems \
   -Wl,-Ttext,$(RELOCADDR) -Wl,--oformat,elf32-i386

all: init.o $(MAIN)

$(MAIN): $(MAIN).exe
	ln $(MAIN).exe $(MAIN)

$(MAIN).exe: init.o
	$(GNATMAKE) -v -O -gnata -gnatE -gnato $(MAIN) -g -o $(MAIN).obj \
	    -bargs -r \
	    -cargs $(CARGS) \
	    -largs $(CARGS) init.o
	$(SIZE) $(MAIN).obj
	$(NM) -g -n $(MAIN).obj >$(MAIN).num
	$(OBJCOPY) -O a.out-i386 \
            --remove-section=.rodata \
            --remove-section=.comment \
            --remove-section=.note \
            --strip-unneeded $(MAIN).obj $@
	$(OBJCOPY) -O binary $(MAIN).obj $(MAIN).bin
	$(rtemsdir)/build-tools/bin2boot -v $(MAIN).bt $(HEADERADDR) \
          $(START16FILE) $(START16ADDR) 0 $(MAIN).bin $(RELOCADDR) 0


init.o: init.c
	$(GCC) -O4 -g  -Wall -ansi -fasm $(CARGS) \
	    -DGNAT_MAIN_STACKSPACE=16 -c init.c

clean:
	rm -f b_$(MAIN).c b_$(MAIN).o *.o *.ali $(MAIN)