# Toshiba T100 exhibit: changelog Each version is the exhibit's RSL version (`simulators/t100/rsl.yaml`); a registered version is final, and a change to the package's content is a new version. ## 0.2.6 (15 Nov 2026) Calculator keys for a program that asks for them (#1301, #1302). No change to the machine, its saved state (format 2), its storage keys or the on-screen keyboard. Built against box-runtime 1.4.6, as 0.2.5 is. - **A program can ask for calculator keys** by carrying the eleven bytes "T100KEYS", the address of a byte and a mask anywhere from 8000h up (the HP-41CV on the T100 does from its pack 1.8.0). While they are there, the ROM is switched out and (byte AND mask) is 0, the page passes the visitor's keyboard as a calculator needs it (`t100/src/machine/hostKeys.ts`): SHIFT with a digit, the period or the comma key goes in as GRAPH and that key by its place (the program's second function: on a German keyboard SHIFT and 0 is not "="), "_" as GRAPH and the minus key, a letter has SHIFT only while SHIFT is held (Caps Lock is no SHIFT), and a typed + - * / is the calculator's operator key whatever shift state the layout needed. With the mask bit set (the HP-41's ALPHA mode) and for every other program the keys go by the character typed, as before. The key handler's chord logic moved out of the panel into that module, with its tests. ## 0.2.5 (15 Nov 2026) The on-screen keyboard's third view, the full keyboard under the screen (#1291). No change to the machine, its saved state (format 2) or its storage keys; the panel's kept settings (`t100.prefs.v1`) gain one field, `kbdPart`. Built against box-runtime 1.4.6, whose shell tells a host the height of a part that stands under the machine. - **Three views on the keyboard's switch**: "ABC", "PF · cursor · pad" and "Full keyboard". The first two are the two-part keyboard of before, beside the screen in the museum's stage and under it on a phone. The third is the machine's whole keyboard in one piece under the screen: the top row with GRAPH, ESC, PF1 to PF8, LABEL, COPY and BREAK, the typewriter block, the edit and cursor keys over the numeric pad, with the model's own legends (the T100's GRAPH legends, the Pasopia's kana, its カナ and KANJI keys). The layout is the one drawn from the Maintenance Manual PAD8120-1, Figure 1.1.4 "Layout of the Keyboard" (p. 1-6), which the automatic whole keyboard of the own page and of the full screen already used. - **The screen grows into the room, and the keyboard takes none of it.** In the museum's stage the full keyboard stands in height the box adds under the stage, so the page grows and the status line moves down with it. One column, centred: the screen as tall as the room the box gives, the switch in a slim row at the keyboard's top edge, the keyboard. The screen takes the column the two-part keyboard took beside it: 572 x 429 -> 644 x 483 px at 1440 x 900, 454 x 341 -> 468 x 351 at 1024 x 768, the same 511 x 383 at 1280 x 800 (there the room's height already limited it); its top stays where it was. The keyboard's top edge stands where the status line stood, in the first screen of the page; picked on the switch, the page scrolls just far enough to show the screen with the PF keys and the number row. On a phone the screen does not move (326 x 244.5 px in all three views). In a room the machine now starts right under the toolbar: centred, the screen moved 3 px between "ABC" and "PF · cursor · pad". - **No key under 44 px.** The full board is at least 864 px wide (19.625 key units of 44 px). In a narrower place (a phone) it scrolls sideways in its own area with the screen fixed above it; a sideways move of the finger is the scroll there, a tap is the key. - **SHIFT, CTRL and GRAPH latch** on it as on the parts (once for the next key, twice to lock), the Pasopia's カナ key shows its lock, a held key repeats, and a key held on the computer's keyboard lights the same key on it. - **The choice is kept** per visitor and machine with the panel's other settings. - **The numeric pad stays true to the machine**: digits, the point, minus and RETURN. The same figure of the Maintenance Manual shows no plus, times or divide key on the pad, so none was added to "PF · cursor · pad" (#1290 asks for them: on the full keyboard the keys that carry + * / are all there). - In a box that does not add the height (an older museum web, another host) the full keyboard still stands under the screen at the screen's unchanged size, and the machine's area scrolls to it. ## 0.2.4 (15 Nov 2026) The drives' access lights (#1286). No change to the machine, its saved state (format 2) or its storage keys. Built against box-runtime 1.4.5, which carries a slot's activity in the status line. - **A lamp per drive in the status line.** The exhibit tells the box what each drive does (the host contract's `"status"` event, a slot's `activity`: idle, selected, read, write), and the box draws a small light before the drive's name. Drive A: and B: report from the drive unit's own signals, watched on the board's side of the µPD765 and nothing inside the chip: the motor bit of the control latch (E6 bit 6), the unit select of the command on the data port, and the main status register (a command in progress, a drive seeking, a data byte of a read or of a write). The flux images go the same way, through the same chip. Between two commands the µPD765 selects no drive, so a drive that waits at a prompt is dark, with the motor bit still on or not. Measured in Chromium: CP/M's boot shows `idle > selected > read > … > idle`, dark at `A>`. - **The cassette's light** is lit while the motor relay is on and the deck plays or records, reads while the ROM samples the read line (`CLOAD`) and writes while the write line changes (`CSAVE`). - **What the real unit has** (the exhibit's accuracy notes name the pages): one red "Select lamp" per drive on the PA7200U's front (its Owner's Guide PAD8142-1, p. 1; the T100 Owner's Manual PAD8110-1, p. 2-53: "the red busy indicator light"). The lamp's circuit was not found in the papers read. The cartridge slots have no lamp and no light. - The lamps' marks are no part of a saved state (`state/t100State.ts` skips them): a machine that was watched and one that was not save the same bytes (`driveLamps.test.ts`). ## 0.2.3 (15 Nov 2026) On top of 0.2.2. Built against box-runtime 1.4.4 (0.2.2 was built against a 1.4.2 that the acceptance site refused: a second, different layer of that version, #1283). No change to the machine's saved state (format 2) or to its storage keys. - **The background colour of a line is the one that stood when the beam left its displayed part** (`t100/src/video/video.ts`): a change of port 08h bits 2-0 made later in the line, in the horizontal blanking, shows from the next line on, as on the tube. Until now the whole line was drawn in the colour that stood at its end. Needed by the opening demo's colour bars (`t100.disk.opening-demo`); a program that sets the colour once sees no difference (`video.test.ts`). - Two new public disks start from the Software applet on both models: "The Machine and Its Story" (`t100.disk.opening-demo` 1.4.0) and "HP-41CV on the T100" (`t100.disk.hp41cv-on-t100` 1.7.0). They are library packs, not part of the package. ## 0.2.2 (15 Nov 2026) The go-live gate's open points of 0.2.0, on top of 0.2.1 (built against box-runtime 1.4.2, as 0.2.1 is). - **The HD46505S is a shared component** (#1199): `chip.mc6845` 1.1.0 with its variant `hd46505s` (R8's skew bits, R12 to R17 readable, the programmable VSYNC width, the cursor's display modes). The package's own register file is gone, and no shared part is left inside the package. The beam stays in the display interface. The pictures, the frames and the machine time of the ROM's and the disks' runs are those of 0.2.0 (`t100/tools/read-path-bench.ts`, `t100/tools/crtc-bench.ts`). - **The saved state is format 2** (the CRTC's fields are the shared chip's). A state of format 1, from 0.2.0 or from the earlier site, resumes: the page accepts it, the machine migrates it on load, and it runs on field for field as a machine of this version (`stateFormats.test.ts`, with a Pasopia's and a T100's state written by 0.2.0). A state saved by this version is refused by 0.2.0 with its format's number. - **One profile.** The package reads the exhibit's own profile only; the Pasopia's is that profile as the model `pasopia` sees it (its id and storage scope, title and library as before), held equal to the earlier site's by a test. The package imports nothing of `simulators/pasopia/` any more. - **The own page's two older browser tests** follow the shell as it is (#1219): the keyboard is "Automatic" on arrival, its row in Machine Settings is a choice, the Library opens on its first tab, a save says "Saved in this browser". 79 and 144 checks pass in Chromium. ## 0.2.1 (15 Nov 2026) Start from the museum's Software applet loads a cassette (#1258). No change to the machine's saved state (format 1) or to its storage keys. - **A hand on the deck's PLAY key.** A start recipe has nobody at the recorder. While its steps play, the deck's PLAY key is pressed for the visitor once the ROM has had the recorder's motor on for a tenth of a second with a tape in a stopped deck (`CLOAD`, `LOAD`). Measured before: 60 s after the steps typed the CLOAD the deck stood in STOP with the motor relay on and the counter at 0.00 s. The tenth of a second keeps the hand off the 13 µs for which both ROMs hold the relay's line low 3.6 ms after power-on. The deck's own keys work as before: STOP pressed while the tape runs holds until the ROM asks again, and outside a start recipe nothing is pressed. - **`wind` in a start recipe** (docs/RML-FORMAT.md §5): the tape goes to three seconds before the file a step names (`wind: { file: "BJ" }`), or to a position in seconds, at once. Pasopia Game Highlight's BJ lies 20 minutes into its tape; its recipe winds there before the `CLOAD`. - Built against box-runtime 1.4.2, which plays the `wind` step and tells the machine while a start recipe plays (`core/src/state/hosted.ts`; additive, the contract stays rsl-host/1). ## 0.2.0 (14 Nov 2026) The T100 as an exhibit package of museum v2, the third after the IME 84 and the NeXT. It holds everything of 0.1.1 below; neither 0.1.0 nor 0.1.1 was signed or registered. - Built against box-runtime 1.4.1 (the one-row bar with its "more" menu in a box's room, the box's status line, the sheets over the device; #1220, #1215). - **Components.** The Z80 with the interrupt daisy chain, the Z80 CTC, the Z80 PIO, the 8255A, the µPD765 and the floppy drive are shared component packages now (`chip.z80`, `chip.z80-ctc`, `chip.z80-pio`, `chip.i8255`, `chip.upd765`, `periph.floppy-drive`, 1.0.0 each), each with its clean-room record, an article and tests that need no ROM. The controller reaches a disk only through the drive's medium interface, so another kind of medium needs no change to it. The machine behaves as before and its saved state is unchanged (format 1). The drives also take flux-level images (MFI, KryoFlux streams, SCP) through `chip.data-separator` and `periph.flux-medium` (#1223). - **Its own entry** on the host contract rsl-host/1 (`t100/src/entry.ts`): the handle has `snapshot()` and `restore()` (the whole machine with the diskettes in its drives), and the disks a box offers for the drives are the machine's library. - **One bar.** `toolbar: [model, fullscreen]`: a box draws no menu line over the machine's own bar. The bar offers only the manifest's models (no Pasopia in the T100's exhibit). - **The wallet.** In a box the wallet of packs is the museum's My Wallet. A pack's disk started from the museum is the visitor's own copy of it when this browser keeps one, so their changes go on. The disks kept in the older Disk wallet are named in My Wallet once per browser; their records are not touched (MIGRATION.md, Storage). - **The RSL** declares what the exhibit offers: `features`, the "?" text as `help:` (the help pack `t100.docs.how-to-use` 0.2.0 is written from it), the component graph and the gates, the clean-room record of the machine's own code, and the two public disk packs offered for the drives. - **Gates:** the components' tests, the pack-disk and migration tests, and the package e2e in the reference box at desk, tablet and phone width in Chromium and WebKit (T-BASIC from the ROM, CP/M from drive A:, DIR on the keys, snapshot and restore, the visitor's copy kept). - **Documents:** this changelog, `HISTORY.md`, the inventory `MIGRATION.md`, the saved state's schema `STATE.md`, the clean-room record `CLEAN-ROOM.md` and the build notes `BUILD.md`. ## 0.1.1 (14 Nov 2026) - Built against box-runtime 1.4.1, the one version that holds both changes to core's shell: in a box's room the toolbar is one row at every width with a "more" menu, the sheets lie over the device and the box draws the status line (#1220); and a machine's tool marked `keep` stays in that row (#1215). Additive, the contract stays rsl-host/1. - The on-screen keyboard's control stays in the room's toolbar row, right after Pause, down to a phone's width and in the box's full screen: it was folded away with the other tools, and a phone or a tablet had no control in sight to bring the keyboard up (#1215). - In a room too short for the screen over the whole keyboard (the museum's stage on a desktop or a tablet) the two-part keyboard stands beside the screen, both whole. Before, the screen was left 160 x 120 px and on a tablet the lower rows of keys lay under the room's edge (#1215). - Gates: Reset and Power Cycle on the real ROM, without a disk and with CP/M in drive A: (`t100/src/machine/ui/restart.test.ts`); the keyboard's place (`keyboardPlace.test.ts`); the on-screen keyboard and Restart in Chromium and WebKit at 1440 x 900, 1024 x 768 and 390 x 844, on the own page, in a host's room and in its full screen (`t100/e2e/osk-restart-e2e.mjs`). ## 0.1.0 (to 13 Nov 2026) - The T100 as an exhibit package: the system ROM as a required pack of its own tier, the shell fitted to a host's room (#1089, #1090).