bus-tester-scripts/ — CMD:OPERAND# scripts for the Bus Test Card (sent by embedded/command-sender/ or
tools/busdrv.py): ALU/*.new (2020 signal names), IO/, Index Register/, Memory Card Tests/,
Gen Test Vectors/ (C program that generates vector scripts), ramtest.logicsettings (Saleae Logic capture setup).
deprecated/gen1-2016/ = the 2016 gen-1 scripts in the OLD signal names (RESET, ALU-FUNC, RD-AC, WDATA…) plus fix,
the sed converter that produced the .new files; deprecated/address-register-2020-07/ = tests for the retired card.
assembler/ — yacc1test.asm, the CPU test program (2020-10-31), its 15 dated 2020 snapshots, a 2026 run log; ledcount/
(the 16-byte switch-ROM LED counter); brur/ (the BRUR $AD test, ABC0123 on the emulator, 2026-09-22).
memory/ — rom_verify.py (~30 s): every ROM byte against the image assembled from basic.img + monitor.img;
memory_status.py (~1 min): boot remap, every ROM byte against basic.img + monitor.img (what the emulator
loads; tools/romimage.py), low RAM spots, and a classification of every 4K block above $8000 (RAM / ROM / video /
undecoded). memory_full_test.py (~16 min with the blocks-1 bus-tester firmware): ROM, address lines, two full RAM
patterns over $0000-$CFFF written then verified in separate sweeps, the video RAM, ROM again. Logs beside them:
2026-09-21 all RAM cells good, ROM = sources, video RAM good.
sequencer/ — run-mode boot transcripts of the sequencer-memory ATmega (2026-09-21): the clobbered copy with -BUS-EN
asserted, the clean copy with the bus quiet, and the Sequencer4 boot (16 s copy, 29 s verify, RAM == EEPROM).
video/ — video_ram_test.py, the video card's display-RAM test over the bus tester (patterns, inverse, neighbour
isolation, the block-0/9 write-through check, read stability); 8/8 on 2026-09-21 after the +5V/VCC join.
basic/ — test, a small BASIC program (LET/FOR loops) used to exercise the interpreter.
compiler/ — the C compiler's test programs (*.c + expected .out/.in/.err) and run.py, which compiles each with
software/compiler/y1cc.py --boot, assembles it and runs it on emulator -x; --oracle regenerates the expectations with
the host C compiler through host_shim.h. 15/15 on 2026-09-22 (switch.c also compiled with --no-brur as switchnb.c); make cc-test.
22 programs since 2026-09-24 (the recursion tests rfact rmutual rlocals rcalc, errors recurse rmain adjstr). Beside it:
corpus.py (every C program the tree compiles, with its options), diffcheck.py (an old y1cc.py from git against the
working one over the corpus: identical assembly), twin.py (y1cc.py against its C twin software/compiler/c/y1cc, also
--16; --chain/--chain16 against the multi-pass compiler y1ccp, cc1..cc9), twinfuzz.py (random subset programs and
60 invalid ones through both compilers; --chain too), passes.py (each pass of the multi-pass compiler against the Y1/OS
program area: image, tables, measured stack; the corpus through the passes with their Y1/OS table sizes). Since
2026-09-24 run.py, twin.py, twinfuzz.py, passes.py and ../ucemu/run.py take --xisa (every compile with
y1cc's --xisa: the LDZ/STZ page, ADDIW, SHL16), xisa.c (// y1cc: --xisa) is the 23rd program, and diffcheck.py
proves the default output unchanged.
os/ — Y1/OS: run.py builds os/disk.img, boots it on both emulators with the console script basic.session
after the monitor's O command, and diffs the transcripts (basic.int.out, basic.uc.out; the latter shows the
monitor's input echo). make os-test. XISA=1 python3 tests/os/run.py (2026-09-24) runs the same sessions with
every /BIN program built by y1cc --xisa (make -C os XISA=1); the program sizes ls/load print are masked.
asm/ — the native assemblers /BIN/ASMC (asm.c, 2026-09-25) and /BIN/ASM (asm.asm, in YACC1 assembly, 2026-09-26;
run.py also runs it under Y1/OS on the instruction-level emulator on the whole corpus and compares its messages
and files with asm.c's, --uc 27 sources on the microcode emulator too): run.py builds os/commands/asm.c for the Mac against an
emulation of the Y1/OS syscalls (host_asm.c, host_sys.c: int = unsigned short, unsigned char) and compares it with
the host assembler RC/asm on every source in the tree (the y1cc corpus plain and --xisa, the firmware, the
hand-written tests, y1os.asm, and src/: quirks.asm with two INCLUDE levels, six sources both must refuse), as
Intel hex and as program files: 297 sources, 0 different. run.py --target (target.py, target.session, the
transcripts target.int.out/target.uc.out) runs asm under Y1/OS on both emulators - y1cc programs assembled
and run, the ROM monitor, isa.asm, compiler pass cc4 - and compares every file it wrote with RC/asm's output.
make asm-test; both in make check.
disasm/ — /BIN/DISASM (os/commands/disasm.c, 2026-09-29): run.py builds it for the Mac (host_disasm.c,
host_sys.c: int = unsigned short, OPEN/GETC/CLOSE/ENTRY/SEEK and peek emulated) and round-trips - disasm -s,
RC/asm, the same bytes, the listing true to them - over tests/asm's corpus, every program of the OS build, the ROM
from memory, random bytes and every opcode + operand byte (382, ~20 s), then START/COUNT ranges and the errors;
tools/gen_y1_distab.py --check first. The run on Y1/OS is tests/os/disasm.session. make disasm-test; in
make check.
ucemu/ — the same programs (and assembler/brur) on the MICROCODE-level emulator software/ucemu with the monitor ROM
loaded (run.py, 14/14 on 2026-09-22, chars.ucout = the expectation with the monitor's input echo), and isa.asm, a
differential test of every instruction whose port-2 byte stream must be identical on both emulators (it is, BRDEV aside;
since 2026-09-24 it covers LDZ/STZ/ADDIW/SHL16 too, 114 bytes, and tests/bench runs it on the machine).
kermit/ — /BIN/KERMIT (2026-09-26): run.py boots Y1/OS on both emulators with the console on a pseudo-terminal
and runs tools/y1kermit.py against kermit -r, -s and -x: text, all 256 byte values, a 70K file, long runs, an
empty file, a name to clean; block checks 1-2, 8th-bit prefixing, text mode, -n, -a; damaged, missing and
unanswered packets both ways, three Ctrl-Cs; the server's SEND/GET/FINISH. Files compared on the host and on the
disk image (p8xfs.py get, fsck); the throughput from the instruction-level emulator's PC histogram; --calib
checks os/kermit_io.asm's timing on the microcode emulator (a timeout second = 1,000,176 clocks, 217 clocks a
received character, with the 2026-09-29 microcode). In make check (~2.5 minutes).
Hardware findings of 2026-09 (memory-card block map, EPROM identity, video-card write-through) are in the card READMEs.