Re: mes wip-arm state

[Jan Nieuwenhuizen]

Danny Milosavljevic writes:

Hello Danny,

> On Mon, 01 Jun 2020 08:33:30 +0200
> Jan Nieuwenhuizen <janneke@gnu.org> wrote:
>
>> I checked and I don't believe they do [force usage of signed char].
>
> That's too bad because mescc ARM uses signed char (I haven't changed
> it, so it should interpret "char" as signed char like it always did).

Oops...does that mean we (probably) have to implement -funsigned-char
(and make that the default for ARM?).  Of course, we can see how far
we get with tcc first.

>> We can also make tcc work before merging, but we could also start a new
>> wip-tcc / wip-arm-tcc later if the need arises.  Whatever works for you!
>
> I'd definitely have the current state marked and stored somewhere--at least
> tagged or something.
>
> If it's not too much work, let's merge it into master.  I think I haven't
> changed the compiler frontend much (or at all), so the risk of affecting x86
> should be minimal.
>
> If the changeset would touch something it shouldn't, let's stop merging then.

Sure, that makes sense.

> (The syscall stuff was technically changed--but only minimally by splitting
> it into two source files so I don't have to carry errno everywhere.  Effect
> for callers of _syscall* should be unchanged)
>
> Let's definitely test it on x86, though :)

Yeah ;)

>> As you propbably have seen, in the bootstrap we use a forked tcc:
>> somewhat older (0.9.27), patched.
>> 
>>     https://gitlab.com/janneke/tinycc/-/tree/mes-0.21
>
> I've checked it out now.
>
>> ba1c47d * lotsa debug printing.
>
> ... and readded that for me.
>
> Thanks!
>
> How do I build tcc using mescc?

I created some scripts to do that, it's less than great.  My excuse is
that mescc only recently grew a POSIX command line, and we don't have
"make" during the guix bootstrap anyway.  Then, I also keep changing
the mes directory layout and build system...Maybe we can give this
a facelift.

Anyway...I went to gnu/packages/commencement.scm to check...I saw
build.sh and cc.sh but those are not used atm.

Also, "bootstrap.sh" has some commented code 

        cp -f $MES_PREFIX/lib/crt$i.c .
        ##cp -f $MES_PREFIX/gcc-lib/x86-mes/crt$i.c .

that you may want to flip to

        ##cp -f $MES_PREFIX/lib/crt$i.c .
        cp -f $MES_PREFIX/gcc-lib/x86-mes/crt$i.c .

and you'll have look into setting environment variables (MES_PREFIX,
GUILE_LOAD_PATH for Nyacc).

As a summary: we build "mes-tcc" first, then using the "boot.sh" script,
we build several new iterations (bootX-tcc) until fixpoint.

You'll probably want to first just build mes-tcc and do some tests with
that.  I suspect you'll hit some hurdles (undefined symbols, or even
unsupported C constructs...Good luck!  :-)

Greetings,
Janneke

-- 
Jan Nieuwenhuizen <janneke@gnu.org> | GNU LilyPond http://lilypond.org
Freelance IT http://JoyofSource.com | Avatar® http://AvatarAcademy.com