- hosted (v4/tools/hosted.c): a line longer than the buffer is read to its
end and refused as too long. Before, fgets handed it over in 1025-byte
pieces and the last piece ran as a line of its own.
- bare metal (kernel/src/v4/sk_v4.c): the line editor takes 1024
characters, as the boot does, and a line typed past that is refused as
too long instead of being cut at 255 and run. V4_BOOT_LINE_TOO_LONG is
now reported at the prompt and while a line waits.
- engine (v4/src/exec.c): a store the node itself refuses (to "any port",
"which port", How or Have) left the opcode running on: !+ advanced A,
against node.h's FAULTS, and !p advanced P past the handler, into the
next fault kind's entry. A refused store now does nothing more; a store
that raises an error (D-18) still counts as made, but P stays the
handler's.
- test_port.c: both cases; they fail on the old exec.c (A 1017, P 901).
Engine-level tests (all v4/tests/test_*.c but test_host_*) pass at 32 and
64 bits; test_port also under ASan/UBSan. Not yet run: make -C v4 test,
test-32, hosted-check, and the three QEMU acceptance boots.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GE3SDvQn5EoxdGufk84zz2
sk_v4_identity.c: a USB drive is this system's stick if its GPT has both StarshipOS partitions and
it is the disk the loader booted from (BootDisk's DiskGUID), or, when the boot was not from a
stick, the first such disk while none is attached. Then: its Artemis partition must carry a good
artemis_sig ('Artemis: the stick's signature is bad -- not used' otherwise), is attached first --
the first disk slot -- ('Artemis: the stick's partition attached'), Zuse's key is read from its
genesis marker, and its identity partition is checked and attached as before. Any other drive,
raw or GPT: 'USB: a drive that is not this system's stick -- not used'. On a pull the identity
partition is detached, then Artemis. sk_v4_identity_boot / sk_v4_run take the loader's BootDisk.
kernel_main.c: no virtio Artemis on the v4 path. kernel/Makefile: ARTDISK= attaches no virtio disk
(every QEMU target); ARTDISK stays for system=v3.
STICK.md 4 amended as ruled (R2): on v4 only this system's stick is used; users' drives minted by
this system's Zuse come with the users design after INSTALL. BRINGUP.md: Secure Boot off on every
PC (ruled 2026-10-28), RDRAND in QEMU, the SER5 smoke-test stick, A4 not on v4.
Tests: new stick_artemis (x3, no virtio disk), stick_zuse, two_sticks (replaces datastick),
artemis_unsigned, raw_ident; changed stick_boot, gpt_zuse, gpt_hotplug, midpull (replug) -- all
failed before, pass after (two test defects found and fixed: the signature's offset, the 63-char
log cut). Regression 20/20. Acceptance: amd64, aarch64, riscv64 to [zuse@Hera] ok>, POST 538/538,
dict_hash 0xfec8c47cfc6f437f; v4 suite and test-32 0 failures.
Plan: docs/v4.0.0/plans/2026-10-27-stick-4-one-disk.md, Task 4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The increment 2 final review's minors 1-4 (Captain Bob, 2026-10-27: "fix 1-4"):
1. A protective MBR with no "EFI PART" header is GPT_BAD, not GPT_NONE: the
MBR says it is a GPT disk, and a GPT disk is never attached whole.
2. A disk whose block 0 cannot be read is GPT_EIO: "GPT: a USB drive could
not be read -- not used", nothing attached (it was read raw, attached
whole, and a raw Zuse drive behind it logged her in).
3. The entry array's LBA is checked against the disk's size directly.
4. gpt_bad_entries: the entry array changed, the header valid -- refused.
Harness modes gpt_no_header, gpt_read_error (blkdebug EIO on sector 0),
gpt_bad_entries. The plan carries the amendment.
Red (amd64): gpt_no_header 070335, gpt_read_error 070717 FAIL;
gpt_bad_entries 070048 PASS (existing check, now pinned). Green: every GPT
mode on amd64 071113-072933; gpt_zuse, gpt_read_error, gpt_no_header on
aarch64 073222-074047 and riscv64 074526-075535; the identity gate's and
increment 1's sessions 080036-082345, emergency 083130 (082635 was run
without EMERGENCY_CONSOLE_ENABLED=1 -- a harness invocation error).
Accepted: suite EXIT 0; 093121, 093352, 093750, POST 538/538, dict_hash
0xfec8c47cfc6f437f; test-32, 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The stick, increment 2 (plan 2026-10-26-stick-2-gpt-reader; docs/v4.0.0/
STICK.md 3 and 4: GPT for boot media and installed disks). kernel/src/gpt.c:
a CRC-32; gpt_read(), which checks the protective MBR, the primary header and
the partition-entry array (both CRCs, the revision, the sizes, the geometry);
and a partition as a blkio device, refusing blocks past its end. The two
StarshipOS partition types: identity 23BEBFE1-4817-4CF3-A537-9069BF159788,
Artemis BCD7B586-1C8E-4EEF-86A3-273A8D8EE624.
The v4 drive path: a GPT disk is used only through its StarshipOS identity
partition, and that partition -- never the disk -- is given to the block
subsystem, in a raw identity drive's place; a GPT disk with no StarshipOS
partition, or a damaged table ("GPT: a USB drive has a damaged partition
table -- not used"), is not attached (Captain Bob's ruling on the plan); a
disk with no GPT is read raw, as always. Pulling the drive detaches the
partition and ends her session. Reads only.
Red (amd64): gpt_zuse 193158, gpt_hotplug 193804, gpt_foreign 194713,
gpt_bad_crc 194947. Green: 195409, 195627, 195854, 200136; gpt_zuse aarch64
200424, riscv64 200839 (the 64 MiB partition attached, 48724 user blocks,
not the 72 MiB disk); the identity gate's and increment 1's sessions,
201119-203403, all pass.
Accepted (with the KERNEL_ARGS fix that follows): suite EXIT 0; 214300,
214507, 214911, POST 538/538, dict_hash 0xfec8c47cfc6f437f; test-32, 39
modules, 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the fresh review of the stick's increment 1 (With fixes):
- Critical: after logout Zuse's last screen, her prompt included, could be
drawn back on a headless machine. The logout clear (ESC[2J) blanked the
pixels only: Alt+TAB, still read while headless, redrew the terminal
from its saved text; and if she had Alt+TABbed to the graphics view
before the pull, the clear did nothing and her session stayed on screen
(and the console did not draw at her next login). Now vt100_reset()
forgets everything the terminal has shown -- the on-screen text, the
scrollback and their boot snapshots -- makes it visible again and fills
the screen; and Alt+TAB does nothing while the screen is off (v3's never
is, so v3 is unchanged).
- Important: an EMERGENCY_CONSOLE_ENABLED build no longer clears the screen
when Zuse's drive is pulled: its console goes on running.
Red: alttab_after_logout 183308 (4515 lit), alttab_before_pull 183529
(4539 lit; the console not redrawn), emergency_pull 183759 (0 lit).
Green: 184115, 184328, 184552; regression 184805-185956 (screens,
keyboard, hotplug, emergency, v3 typed by keyboard) all pass.
Accepted: 190111, 190312, 190710, POST 538/538, dict_hash
0xfec8c47cfc6f437f; test-32, 39 modules, 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The stick, increment 1, task 2 (plan 2026-10-26-stick-1-screen-and-keyboard;
docs/v4.0.0/STICK.md 2, 5.3, 5.4). On the v4 path:
- the framebuffer is brought up after POST, cleared and dark: console text
goes to serial only ("splash only, log on serial") until Zuse logs in;
then the screen is the console (TTF, a blinking cursor, as v3's), and
when she logs out it is cleared and dark again (console_fb_mirror(),
console_fb_clear(); the mirror is on by default, so v3 is unchanged);
- the line editor and the node's KEY read every keyboard -- PS/2, USB, the
virtio keyboard -- through the shared bridge (sk_input_getc()); the
headless wait drops keyboard input as well as serial bytes;
- an EMERGENCY_CONSOLE_ENABLED build has the screen from the start.
The framebuffer setup is one helper both paths call (kernel_main.c).
Red (before): screen_dark_until_login 153051 (10835 lit), console/keys
153309, logout 153609, kbd_typeahead 153918, virtio_kbd aarch64 154245
and riscv64 154718; no_framebuffer 155546 passed before and after.
Green: 155926 (0 lit), 160156, 160409 (0 lit after logout), 160630,
virtio_kbd 161414 and 161820, no_framebuffer 161206; usb_kbd with the USB
keyboard the only one (-machine i8042=off) 164548, v3 control 164757;
the identity gate's sessions 162245-163802 all pass; v3 typed by keyboard
164009, 164115. (usb_kbd 160857 and 164313 failed in the harness:
input-send-event's device names a display, not a keyboard.)
Accepted: suite EXIT 0; three clean qemu boots, POST 538/538, dict_hash
0xfec8c47cfc6f437f unchanged (174704, 174911, 175315); test-32, 39
modules, 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the fresh whole-branch review of the identity gate (With fixes; no
Critical, four Important):
1. A drive pulled while a line ran: the drives are looked at before every
prompt (sk_v4_identity_check_now), so no prompt is shown once she is
gone and what was typed after the pull does not run as Zuse.
2. Pulled while a line WAITS: the wait is ended at logout, by an empty
line, and not carried into her next session (Captain Bob, 2026-10-26:
"end the waiting line at logout"); re-attaching brings the prompt back.
3. EMERGENCY_CONSOLE_ENABLED: v4 honours it as v3 does -- a bring-up build
has a prompt with no drive (boards/BRINGUP.md A2).
4. Keys typed at a headless machine are read and dropped, not run as
whoever logs in next. (v3 has the same gap, sk_repl_headless_wait;
reported, not changed.)
Each a QEMU session red then green (amd64): midpull 111650 -> 112953,
waitpull 111907 -> 113154, emergency 112524 -> 113609, typeahead
112248 -> 113359; headless 113800 and hotplug 114043 still pass.
Accepted: three clean qemu boots to [zuse@Hera] ok>, POST 538/538,
dict_hash 0xfec8c47cfc6f437f on all three; hotplug on aarch64 and
riscv64; make -C v4 test test-32 sanitize hosted-check EXIT 0;
make -C v4 test-32 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The identity gate on the v4 boot (plan 2026-10-26-engine-step4-identity-gate,
task G2; ENGINE.md 2, ruled 2026-10-20). As v3, by the kernel with no v3 VM:
after POST and the block chain, Zuse's public key from the genesis marker,
every xHCI controller brought up, the heartbeat started; then headless until
her drive is attached and checked (capsule_zuse_boot_check, G1). A drive that
arrives is given to the block subsystem by the kernel (Artemis is the
kernel's); her session is kept in the kernel for Hera, never in a node's
memory. The prompt is v3's, [zuse@Hera] ok>; pulling her drive ends the
session and the system goes headless; putting it back brings the prompt.
A blank pair stays headless: genesis writes Artemis's disk, Phase 5 (D2).
kernel/Makefile: ARTDISK and ZUSEDISK default to the minted pair,
disk/zuse-scratch/ (D4); .claude/CLAUDE.md's acceptance says so.
Fixed with it, at Captain Bob's word:
- kernel/Makefile: the ROM-and-map rule is a grouped target (&:). As two
targets of one plain rule its recipe ran twice at once under -j and the
copies met ("cp: cannot create regular file ... File exists"): 3 of 20
parallel builds failed, now 0 of 20.
- The v4 boot drops the logger to --log-level after POST, as v3 does
(sk_log_level_after_post, shared by both paths); INFO lines after POST,
3 before, 0 now.
Accepted: three clean qemu boots as built to [zuse@Hera] ok>, POST 538/538,
dict_hash 0xfec8c47cfc6f437f on all three (unchanged), 0 INFO after POST;
the hot-plug session (attach, line, detach, headless, re-attach) on amd64,
aarch64 and riscv64; headless with no drive; make -C v4 test test-32
sanitize hosted-check EXIT 0; make -C v4 test-32 0 failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The host node starts in its ROM (entry 12), and the boot feeds it the
system's source through the console: capsules/v4/sys/map-64.f18, the
fourteen .f18 blobs and end.f18, each found, hash-checked and
signature-checked as any capsule, then EOT. The ROM assembles the system
and starts it at (IDLE); the host reads the image record
(v4_image_from_node) for what only the assembled system knows. A ROM
error is reported with its blob, line and meaning, then PARITY:FAIL.
Deleted: v4/tools/mkimage.c, capsules/v4/nucleus-64.f18, v4/capsule/*.v4,
v4/tools/v4tof18.py. The .f18 files are the source now.
- text.c speaks the rest of the ROM's language (set, here, store, start;
v4_text_value, v4_text_start), so the oracle reads the same source:
test_host_rom_nucleus compares the ROM against text.c on the .f18 files,
word for word, image record included. Host tests load the .f18 files.
- mkmap image writes v4_map_image (the map's description of the system);
hosted and the kernel link the ROM and the map instead of v4_image_64.c.
- Hera (Captain Bob's ruling, option 1): births stream the system the ROM
assembled, as assembled code; hera.4th is unchanged.
test_host_unit's host builds the system with the ROM.
- New parity lines: PARITY:V4_F18 per blob, PARITY:V4_ROM words=2049
rom_hash=0xa6c792b5821010b7 instructions=60725216.
Acceptance: amd64, aarch64, riscv64 boot to ok> with PARITY:V4_POST
tests=538 pass=538 fail=0, POST: PASSED, and PARITY:V4_SYSTEM
word_count=317 dict_hash=0xe5a57f3eb72ff94a on all three -- the same hash
as the prebuilt nucleus gave. make -C v4 test test-32 hosted-check: all
pass. Logs: logs/20261015-141648, -141938, -142350.
Startup cost under QEMU TCG, the ROM's assembly (last V4_F18 line to
V4_ROM): amd64 26.7 s of 132 s to ok>; aarch64 41.0 s of 195 s; riscv64
20.5 s of 101 s. Hosted native: +0.85 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The final review of step 6e, by a fresh reviewer told to break it, and
what was ruled: fix them all, test first. Each has a test that failed
before the mend.
- Two texts from one node could be begun in the wrong order when room
for their answers came between the node's look at the first and at the
second. Texts whose answers go by the same port are begun in the order
they came: (D-MASK), quit.v4. 102 of 700 timings were wrong before.
- A word written with PORT! to a node that speaks whole messages, or to
a console's port, blocked its writer for ever and cut off every node
beyond it. It is error 18 now, whether that node is asleep, running
or stuck: fabric.c, node.c.
- A node asleep for room on one wire is woken when another of its wires
is cut.
- A line there was no room for on the lone node's wire was reported as a
fault and ended the hosted program. It is Message refused, and the
prompt comes back: V4_BOOT_LINE_REFUSED. hosted-check has the case;
no bare-metal boot has typed it.
- Ruled: text behind held output for the same port is refused though it
would fit. MESH.md 7d.4 said otherwise and is corrected.
- The removal test also cuts and re-makes a wire, and has an exchange in
which output is held at a full wire: 26,814 disturbances in all.
- MESH.md and v4/README.md no longer say that a stuck node holds up no
other node, that only printing waits, or that no node is left waiting
on one that is gone; 7d.6 lists the limits and what the products do
that they did not before.
make -C v4 test and sanitize at both widths, hosted-check, lint. Three
bare-metal boots: logs/20261013-194827 (amd64), -195131 (aarch64),
-195518 (riscv64). POST 538 of 538, word_count=317,
dict_hash=0xe5a57f3eb72ff94a on all six; the three sessions are the same
character for character. The boots were built with the working tree's
uncommitted change to kernel/Makefile that adds the ISA to the kernel's
version line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The bare-metal host hands a waiting node the next line typed, as the
hosted one does: 5 AWAIT, then 7 8 * . -- Interrupted, then 56.
amd64, aarch64 and riscv64 boot, POST 538 of 538, word_count=317,
dict_hash 0x54520ade672566ad, the same as the three hosted programs:
logs/20261007-202737, -203230, -203616.
MESH.md step 6c: what was built, how it was verified, and two things
found on the way that are not mended and are for ruling: Hera blocked
writing to a stuck node, and a message that can go round for ever between
two nodes whose ways disagree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A block request with fewer than two values on the stack is refused; it
had acted on whatever the stack ring held and stopped the node.
Bare metal: when the kernel's chain takes the place of POST's block RAM
the node's two buffers are emptied, so it no longer holds POST's copy of
a block; and the chain's fast RAM is cleared, so a node cannot read what
was in the kernel's heap.
blocks.c is built with each test under that test's own warnings and
sanitizers; it had been left out of both. The hosted link cleans its
object directory first: it had linked the withdrawn store_v3.o left there
from the day before.
node.h and DECOMPOSITION.md D-19 no longer describe the message device or
the four registers as current. MESH.md 8.5 records two findings for
ruling: a node's own copy of a block, and a block read over a node's code.
From a clean build: make -C v4 test, sanitize and hosted-check pass;
amd64, aarch64 and riscv64 boot, POST 538 of 538, same hashes, blocks 1
and 2047 clean at the prompt: logs/20261007-085017, -085254, -085636.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A block is a kernel request, as ENGINE.md 3.3 has it: the node puts the
block's number and the address of 256 cells on its stack and writes the
request to port 0, and the kernel leaves the status there. The requests
are -1, read, and -2, write, the same for every node. v4/system/blocks.c
serves them from the kernel's block subsystem, which is v3's. The four
storage registers are gone from the engine.
The device that spoke block messages (4a505a15) is withdrawn with its
test and its message types: Captain Bob ruled on 2026-10-07 that it, a
node's own drive, and nodes with no storage had left the OS as designed
(docs/v4.0.0/MESH.md 8.5).
Hera no longer sends POST to the nodes she births: POST is the kernel's,
once. Every node has its kernel on port 0; it serves a node's blocks and,
for Hera alone, her requests for nodes and capsules.
Bare metal: the node boots and is POSTed against POST's own block RAM,
and the kernel's chain -- fast RAM, the ramdrive, the virtio disk -- is
set up after POST and before the prompt, as on the v3 path. The disk is
read and not written: nothing in v4 yet gives the owner's word that it
may be formatted. A hosted program has the chain's fast RAM, as hosted
v3 has with no disk. Error 17 is Storage refused.
make -C v4 test and sanitize pass at both widths; hosted-check passes on
three ISAs; amd64, aarch64 and riscv64 boot, POST 538 of 538, with the
typed session: logs/20261007-081603, -081839, -082226. The hashes are
the same on all six.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MESH.md step 3. The ports are the transport; the message is what is
transported: to, from, type, heat and TTL, ACL tag, sequence, length, then
text four characters to a word.
- quit.v4: a node with nothing to do is blocked reading "any port"; text
for it is interpreted; (FINISH) sends what it printed and then how the
text ended, and it waits again
- core.v4: EMIT keeps what is printed, (FLUSH-OUT) and (HDR) send it to the
sender on the port the message came on. EMIT still needs one free data
cell and no more; it works on the return stack and in A and B
- message.h/.c: the same format for whatever is on a port and is not a node
- boot.c: the boot is the node's console on port 1 and its kernel on port 0
- the prompt tests are a console that speaks messages
- gone: v4_line_begin, v4_line_done, v4_line_status; writing a node's input
buffer and setting its P from outside; any use of CONSOLE-TX
Verified: make -C v4 test (test_host_quit.c 1283 checks, the full-stack
figures unchanged) and make -C v4 sanitize pass; hosted-check passes on
three ISAs with POST 550 of 550; clean qemu with STARFORTH_V4=1 passes POST
and answers lines typed at each prompt on amd64, aarch64 and riscv64
(logs/20261006-110551, -111621, -111341). -110837 is an aarch64 run ended
by the test wrapper's limit while still in UEFI firmware; it shows nothing
about v4.
Not done: KEY, EXPECT and QUERY still read the console's input registers;
a message not for this node is let go (step 4).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ENGINE.md step 2, the carrier. Ruled 2026-10-05 (V3-PARITY.md 1i), on
DECOMPOSITION.md section 6: a write to a port blocks until the neighbour
reads.
- node: v4_node_port_attach, v4_node_port_served; a store to the port keeps
the value as the request and blocks the node
- exec: a blocked node executes nothing; served, it goes on from the opcode
after the store, in the same instruction word; a fault meanwhile abandons
the rest of the word
- compile.v4: n KERNEL-WORD name makes a word whose body writes n to the
port; its arguments and results are on the data stack
- boot: the kernel's words are made by handing the node text, and requests
are served between the node's opcodes; one no one serves is error 12
- BYE, the first kernel word: hosted it leaves the program, as hosted v3;
on the lone node it is v3's cold restart
- ENGINE.md 3a: multiuser, multitasking, preemptive and cooperative, and
what that asks of the engine
Verified: make -C v4 test passes at both widths, with tests/test_port.c;
hosted-check passes on three ISAs; clean qemu with STARFORTH_V4=1 on amd64,
aarch64 and riscv64 passes POST with the same hashes as hosted, and a
kernel word no one serves and BYE typed at each prompt are answered
(logs/20261005-185506, -185734, -190101; -185234 is an amd64 run in which
those two lines were not typed).
Not done: v3's own C functions serving a node.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A v4 node no longer reads its own command line or prints a prompt. Its
host puts a line of text in the node's input buffer and starts it at
(LINE); the node interprets it and stops at (IDLE), leaving in
(LINE-STATUS) how it ended: completed, an error, or QUIT. The host says
" ok" or " ERROR" and prompts, as the kernel's REPL does for a v3 VM. A
line may be 1024 characters, a block, as v3's. Ruled 2026-10-05
(V3-PARITY.md 1b); design ENGINE.md 3.1.
- quit.v4: (REPL), the node's prompt loop, is gone; (LINE) (IDLE) (DONE)
- image.h/.c: v4_line_begin, v4_line_done, v4_line_status; the node is
idle at switch-on
- boot.c: v4_boot_line, the one loop the hosted binary, the kernel and the
capsule loader hand a line with; the code that took " ok" and the prompt
back out of the node's output is gone
- hosted.c, sk_v4.c: the prompt and the line editing are the host's
- test_host_quit.c: the tests are the node's host; two tests of the old
80-character prompt line now test a whole line, 1024 and 1025 characters
Verified: make -C v4 test passes at both widths; hosted-check passes on
three ISAs; clean qemu with STARFORTH_V4=1 on amd64, aarch64 and riscv64
passes POST (550 of 550) with the same hashes as hosted, and three lines
typed at each bare-metal prompt through the serial port are answered
correctly (logs/20261005-180922, -181152, -181541).
Still the lone node: kernel_main.c starts it before the fleet tables.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One boot, v4/system/boot.c, for both products: it starts the nucleus image,
finds each capsule in the baked capsule directory, recomputes its hash,
checks its signature, gives its blocks to the node a line at a time, and
prints PARITY:V4_NUCLEUS, PARITY:V4_CAPSULE and PARITY:OK before the prompt.
A line the node does not accept ends the boot with the capsule, block and
line named. docs/v4.0.0/NUCLEUS.md.
- hosted Linux product for amd64, aarch64 and riscv64 (make -C v4 hosted);
make -C v4 hosted-check boots all three and requires identical output
- the kernel's v4 entry (STARFORTH_V4=1) calls the same boot
- capsules/v4/forth79.4th, block 6000: no definitions yet
- mkimage builds the nucleus only; no FORTH source is compiled at build time
- capsule_blocks.c: the Block-header parse, free of any VM, for every loader
Verified: make -C v4 test passes; hosted-check passes on the three ISAs with
the same hashes; the kernel compiles with STARFORTH_V4=1 on the three.
Not verified: no bare-metal boot of v4 has been run.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
