Sunfish with GCC
John.Tytgat at aaug.net
Mon Jul 4 19:04:08 PDT 2005
In message <4e014e844d.Jo at hobbes.bass-software.com>
John Tytgat <John.Tytgat at aaug.net> wrote:
> I would do a release after the following points have been handled:
> 1. Creation of a document outlining how to use GCC for module generation.
!gcc.docs.Intro tells a little bit but a few more words couldn't do any
> 2. Investigate whether StubsG and the CMunge program can be used for
> GCC module generation. If this really works, then non-Norcroft owners
> can generate C based modules.
Did more than this suggestion : I've now added CMunge v0.52 (which is
more recent than what's to be found in the StubsG v0.04 release) in our
GCCSDK build system (with some fixes). This means, not only do we have now
a full solution for GCC module builders on RISC OS, but also the
cross-compiling case is covered.
I did some some testing with cross-compiling Sunfish and that seems to
work. Typically you will have in the makefile :
CMHG = /home/riscos/cross/bin/cmunge
CMHGFLAGS = -throwback -32bit -tgcc
moduledefs.h modulehdr.o: modulehdr.cmhg
$(CMHG) $(CMHGFLAGS) modulehdr.cmhg -o modulehdr.o -d moduledefs.h
[ "-throwback" when cross-compiling still means outputing via stderr ]
If some more people could smoke out this cmunge release both on RISC OS
and cross-compiling (I had to made some changes to the official version),
that would be great.
> 3. Investigate why RemovePrinterFS doesn't work when using GCC. When we
> know the reason we can still decide whether we fix it for the release 3
> or not.
That was the vsnprintf issue and now resolved (see point 5).
> 4. Smoke test with Memphis, CryptRandom and EtherY module src code.
Still to be done.
> 5. Investigate if we can't get rid of the "#undef vsnprintf" workaround.
> 6. Write ",ffa" extended module files instead of ",ff8" when cross-compiling
> (feedback from John B).
And this too.
John Tytgat, in his comfy chair at home BASS
John.Tytgat at aaug.net ARM powered, RISC OS driven
More information about the gcc