loader and ficl/Forth help

[Jeremy Chadwick]

On Sun, Feb 24, 2008 at 06:38:38PM +0100, Oliver Fromme wrote:
> There's a debugging aid called "testmain" so you can run
> Forth code interactively in multi-user mode for testing
> purposes.  For details please read the thread starting
> here:
> 
> http://lists.freebsd.org/pipermail/freebsd-stable/2005-May/015387.html

Thanks!  This helps a lot.  I'm able to reproduce the sh_el not found
condition with testmain.

> Personally I use qemu for testing loader things.  I wrote
> a trivial script that updates a small disk image and then
> boot its in qemu, which only takes two or three seconds,
> so testing cycles are pretty short.

Hmm, interesting software.  I'm used to VMware myself, but this seems a
bit less bloated.  I'd have to run it on Windows, since I don't use
FreeBSD as a workstation (no SDL, no VNC/X, etc.).  The FreeBSD box in
my home runs Samba, so I should be able to set up a method of creating
ISO images of FreeBSD, which I can then test via qemu on Windows via
CIFS/SMB share...

> I suggest you switch to ASCII characters automatically if
> the loader is running on comconsole (i.e. serial), and use
> CP437 it it is running on vidconsole (i.e. VGA).  Then
> there will be no need to manually switch a variable.

Great idea; I think I'll go with that.

> I think the stack layout is not correct, particularly the
> drops aren't quite right in some branches of the "if"
> statements, I think.  It's best to make a stack diagram
> on paper and go through all possible branches with it.

I have a lot more reading to do on the subject of Forth.  Stacks are
hardly a new concept to me, but the conditions where Forth puts
something onto the stack are.  I've never worked with a language like
this before (at least RPN doesn't frustrate me!).  I've already found
some reading material which should help.

> s" loader_logo_lines" getenv
> dup -1 = if
>         drop
>         s" arch-pc98" environment? if
>                 drop
>                 pc98_lines
>         else
>                 cp437_lines
>         then
> else
>         s" ascii" compare-insensitive 0= if
>                 ascii_lines
>         else
>                 cp437_lines
>         then
> then

Again, thanks -- I don't quite understand the changes yet, but will
after I read aforementioned documents.

Sadly, the "sh_el not found" error still continues with this code -- but
it does fix the crashing I was seeing.  I believe the problem is that
constants cannot be put inside of a function (functions being runtime
things, not compile-time things).  Since the original code used
compile-time conditions ([if], etc.), it meant that only one declaration
for sh_el, sv_el, etc. was made.

Since the *_lines functions are doing the assignments, I had to change
them all to runtime, e.g.:

    149 sh_el !
    150 sv_el !

And so on, ditto with adding "variable sh_el", "variable sv_el" and
others to the very top of frames.4th.  This actually fixed the "not
found" error, and now I'm stuck with "Error: compile only!" being
induced by the "dup -1 = if" line.  The same happens when running this
natively (outside of testmain).  I'll have to get the qemu stuff set up
so I can test this more easily.

-- 
| Jeremy Chadwick                                    jdc at parodius.com |
| Parodius Networking                           http://www.parodius.com/ |
| UNIX Systems Administrator                      Mountain View, CA, USA |
| Making life hard for others since 1977.                  PGP: 4BD6C0CB |