1
0
mirror of https://github.com/hsoft/collapseos.git synced 2024-11-27 12:58:09 +11:00
collapseos/doc/selfhost.txt

99 lines
3.7 KiB
Plaintext
Raw Normal View History

# Assembling Collapse OS from within it
This is where we tie lose ends, complete the circle, loop the
loop: we assemble a new Collapse OS *entirely* from within
Collapse OS.
Build Collapse OS' from within Collapse OS is very similar to
how we do it from the makefiles in /arch. If you take
the time to look one, you'll see something that look like "cat
xcomp.fs | $(STAGE)". That's the thing. Open "xcomp.fs" in a
text editor and take a look at it. Some xcomp units are simple
proxy to a block, which you'll find in the blk/ subfolder for
this recipe.
To assemble Collapse OS from within it, all you need to do is
execute the content of this unit, minus the 2 lines of "spitting
to port 2" at the end of the unit, which is a special signal for
the stage binary.
The rest can be executed on any Collapse OS system with enough
memory. It will yield a binary in memory. To know the start/end
offset of the binary, you'll type the same two commands at the
end of xcomp.fs, but replace the "/MOD 2 PC! 2 PC!" words with
".X". Then, write that binary between those offsets on your
target media. That binary should be the exact same as what you
get in "os.bin" when you run "make". You now have a new Collapse
OS deployment.
Is that it? Yes. But for your own enlightenment, let's look at
the xcomp unit in more details. Full details are at
doc/bootstrap.txt.
The first part is configuration of your new system. When RAM
starts, where RSP and PSP start, what ports to use for what
device, etc. These configuration declarations are expected in
the boot code and driver code.
Then, we load the proper assembler and the cross compiler (xcomp
for short), which we'll of course need for the task ahead.
Then come xcomp overrides, which are needed for xcomp to be
effective.
At this point, we're about to begin spitting binary content,
this will be our starting offset. "ORG" will soon be set to your
current "H@".
Then, we assemble the boot binary, drivers' native words, then
inner core, close the binary with a hook word. We're finished
with cross-compiling.
We're at the offset that will be "CURRENT" on boot, so we update
"LATEST".
Then, we spit the init source code that will be interpreted on
boot. And... that's it!
# What to do on SDerr?
If you self host from a machine with a SD card and you get
"SDerr" in the middle of a LOAD operation, something went wrong
with the SD card. The bad news is that it left your xcomp
operation in an inconsistent state. The easiest thing to do it
to restart the operation from scratch. Those error are not
frequent unless hardware is faulty.
# Cross-compiling directly to EEPROM
If your target media is a RAM mappable media, you can save prec-
ious RAM by cross-compiling Collapse OS directly to it. It req-
uires special handling.
You can begin the process in a regular manner, but right before
you're about to assemble the boot code, take a pause.
Up until now, you've been loading your cross compiling tools in
RAM, now, you're about to write Collapse OS. So what you need
to do is change HERE to the address of your EEPROM. Example:
0x2000 HERE !
If you need to activate special system overrides such as the
one described in doc/hw/at28.txt, now is the time.
Then, you can continue the process normally.
# Verifying
You can use "/tools/memdump" to dump the memory between your
begin/end offsets so that you can compare against your reference
stage 1. Before you do, you have to take yourself out of xcomp
mode. First, run XCOFF to go back to your regular dict. Then,
run "FORGET CODE" to undo the xcomp overrides you've added
before. That will rewind HERE. You don't want that. Put HERE
back to after your ending offset so that you don't overwrite
your binary.
Then, you can run "/tools/memdump".