When writing the xcomp documentation, I realized that with careful
threading and by accepting a bit of code duplication in the xcomp
toolset, I could de-stabilize a couple of words.
(n) and EXIT are a bit trickier, but I think it can be done. It
would be nice to get rid of stable wordrefs...
The 1 byte limitation has been effective for a while now, but I
hadn't made the move yet, I wanted to see if the limitation would
cause me problems. It doesn't.
Doing this now slightly facilitates the IY->BC move in z80.
Bootstrapping: if you try to recreate the CVM binary from the
previous commit with this code, you'll have bootstrapping problems.
The first bootstrap will compile a binary with 2-bytes wide cells
but branching conditionals that yields 1-byte cells. That's bad.
I got around the issue by temporarily inserting a "397 399 LOADR"
instruction in cvm/xcomp.fs, right before the xcomp overrides. This
way, I force 1-byte cells everywhere on the first compiliation,
which then allows me to apply the logic change in cvm/vm.c and have
a properly running binary.
There is now no more actual code in stable ABI, only references.
This makes refactoring of this code much easier. For example,
changing IY to BC as the IP register.
Previously, it was impossible to cross-compile Collapse OS from a
binary-offsetted Collapse OS because stable ABI wordrefs would have
a wrongly offsetted address.
This solves the problem by replacing those wordrefs by direct,
hardcoded stable ABI offset references.
I'm planning on going back to 8-bit branching. 16-bit br cells incur
a non-negligible penalty and, while at first 64 words (128 bytes
forward or backward) seemed a bit limiting, I now don't see why one
would ever construct such a big branch. It would be un-forthy.
Also, I looked at using BC instead of IY to hold IP and the transition
would be a lot easier with 8-bit branching.
In this commit, all I do is add overflow checks in IF. The mechanic
below doesn't change. I'll give myself some time to think it over so
that I avoid yet another back and forth.
This is done by adding a "C<* override" layer that is reset to 0 on
ABORT.
The problem was that when ABORT happened during a LOAD, we had a
resetted RSP and started from a clean INTERPRET, but LOAD didn't
have the opportunity to restore C<*, which caused it to continue
interpreting from the faulty BLK.
With a C<* override, we don't need to *restore* C<*, we just need to
clear the override.
This is more than cosmetic, it's also highly usable. The presence
or absence of the "ok" message allows us to know whether the command
aborted. Previously, the "> " prompt appeared when the system expected
a prompt in the INTERPRET context, whether the previous command aborted
or not.
Also, this allows us to get rid of that ugly FLAGS global variable.
