I've mistakenly broken it in previous commit.

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...

Also, fix the SPI relay driver to properly AND-away the result of
the CTL read.

Tested with a real prototype, works fine.

By default, it changes nothing, but it allows interesting
setups, such as using AT28! for directly uploading to EEPROM.

I've also updated the EEPROM recipe to upload directly to 0x2000.
I'm not sure what has changed, but it's working fine now.

Previously, communications with my RC2014 would always be rocky and
fail most of the time. By sending an CR at init, things go much
smoother.

Working on programming AVR chips exposes a glaring omission in my
first design of the SPI Relay: not allowing multiple devices make
this task hard. I constantly have to unplug my SD card before, plug
the AVR chip holder, then play a bit, then unplug the AVR holder,
then replug the SD card...

My prototype for a SPI relay design is built, but I haven't tested
it yet. I need to adapt the code first, which is what I do here.

When the prototype is tested, I'll update the SDC recipe with a new
schema.

The batch mode thing seems to have been caused by bad timing on my
prototype. Now that I've corrected it, live interaction seems to
work fine.

After having done my initial handshaking, I thought the rest of the
job was only a matter of implementing the protocol, but I was wrong.
There were many issues to fix before I could reliably communicate
with my 328P, mostly timing.

Now, I seem to be able to reliably extract fuse information, but
only in batch mode (that is, run "asp$ aspfl@" directly, then
running "(spid)" before running the next command). If I try to
interact with the chip in a single asp$ session, I sometimes get
wrong values (but no sync error! that's worrying...).

chkPS, used to be only for words that pushed back to PS, but I've
recently removed underflow checks from next and I forgot to add
missing chkPS, calls that go with this change.

AVR chips often run at less than z80's system clock. This means that
our SPI relay needs to have its own clock to properly communicate
with it. This means that the delay between OUT and IN can't be
hardcoded to 2 NOPs anymore. It needs to be configurable.

my previous commit was only tested in emulation, I had init wrong.

Always returning zero caused test failures not to be properly
detected.

Although the SPI Relay driver is RC2014-specific, the SD Card driver
is generic enough to be a subsystem. That's the second subsystem we
add and this warrants, I think, the formalization of a new concept:
protocols.

It gives a better idea of where stuff can be moved/added to.

Among random "better safe than sorry" changes, the real fix is in
changing "4" for "5" above _find declaration. This off-by-one error
had that word, which is the root word in z80, have a 0x01 prev field
instead of a 0x00 one.

When all memory was initialized to zero, it didn't matter, we ended
up hitting 0 prev and considered ourselves properly at the end of
dict.

When memory wasn't initialized, however, we would end up jumping at
all kinds of places, leading to random behavior.

This should help spot bugs due to bad initialization.

The previous approach of maintaining R> and W> pointers was
conceptually simple, but made INT handler code actually quite
complex.

Now, we maintain indexes instead. It's much easier to perform
bounds checks and to compare for equality, something we have to
do quick in the INT handler.

it seems I left my asm argument harmonization half done here.

Instructions list at B208 doesn't correspond to many actual
mnemonics. This mnemonic here was the worst offender.

I've changed my mind about having documentation in-system. It doesn't
serve much of a purpose and make blkfs significantly heavier.

This commit is the first step in writing a documentation outside of
the blkfs.

Also, remove RSP underflow check from next for the same reason as
in the previous commit in z80.

with uflw checks not being in next anymore, we can save ourselves
a forward jump.

Also, remove underflow checks from next: it's taxing and useless
because underflow check already takes place in each relevant native
word.

They were a bit terse.

Under OpenBSD, stdin is already used by the device itself because of
the whole "stty has no memory" situation.

Previously, it could never write more than a few bytes before pingpong
getting a mismatch error. Now, I can pingpong Collapse OS binary
without a mismatch.