[ww8l]

Closes #47.

Programs the AnyTone AT-D578UV over its USB cable. It supports:

• channels, zones, scan lists and talkgroup contacts;
• 264 menu settings, plus the radio's own DMR ID and name;
• the 500,000-entry call-sign (caller-ID) database;
• the air-band memory, which only the D578UV PLUS has.

How it fits

The D578 speaks the same clone protocol as the D890UV, but over the D868/D878-family memory map with D578 overrides (firmware 1.21).

• Shared with the D890: the transport, the window write engine and the codeplug planner. They are now parameterized (PlanLimits, IdentExpect).
• The D578's own: its map, record codecs, program and settings sessions, and call-sign DB.
• Firmware gate: sessions refuse any radio whose ident isn't revision V111, the revision the 1.21 map was built for. No readable region carries the firmware version.
• Program dialog: the D578 uses the AnyTone dialog. Its Verify, Restore and latest-backup commands now take a driver_key, and refuse a backup written for the other model.

Measured on a real D578UV (fw 1.21)

• Writes rebuild a whole flash half. A write rebuilds the entire 128 KB half it lands in, from only that session's data. A 16 KB settings write erased the GPS-roaming table next to it. Writes now carry every window of the half that holds data.
• The radio owns the end-of-half marker. Sending it, even unchanged, left new data in the inactive half, and the radio kept showing old channels. Writes now never send that frame.
• CTCSS: no 150.0 Hz entry. Borrowing the D890's table would shift every tone from 151.4 Hz up.
• Call-sign DB layout: the end-of-database address is a real address, gaps included (qdmr writes it differently). Names stop at 15 bytes.
• Simplex offsets: simplex channels store their RX frequency in the offset field.
• Settings labels: "Boot Zone A/B" is really the current zone. Backlight is stored 0–4 for menu steps 1–5.

Verification

Re-encode gates. Each takes the radio's own data and must reproduce it byte for byte:

• all 392 channels, 21 zones, 21 scan lists and 38 contacts, plus the contact ID table;
• the radio's existing 281,304-entry call-sign database (an env-gated test against a local dump, kept out of the repo because it is radioid.net data).

Hardware ladder on the radio, each step read back and restored:

Step	Result
Identity write	✅
One-byte name change	✅
Full 208-channel codeplug	✅ zones checked on screen
Restore	✅
Band probe, 136–520 MHz edges	✅ 6/6
Settings (brightness, name colour)	✅ checked on screen
Call-sign DB, 310,769 entries	✅ 18 MB read back exactly

Full regression through the shipping program and settings paths: 120 windows swept before and after. Nothing changed outside the plan.

Dialog: Verify and Restore ran on the radio. The frontend was driven with stubbed IPC and real fixtures.

Not verified:

• Air band: the test radio isn't a PLUS. The README says so.
• Transmit, on any band.
• Inverted DCS (stored as code + 512, per qdmr and dmr-tools): the test radio has none.
• Settings beyond the two checked on screen: they rest on qdmr and dmr-tools agreeing (189 of 191 offsets). 59 fields are withheld.

Also in this PR

• Bluetooth tab: a new settings tab for every radio that has Bluetooth (D578, D890, BT-9000, FTM-300, ID-5100, ID-52, TH-D75). Connections keeps the COM-port, CI-V and data settings.
• Text encoding: settings text and radio names are folded to ASCII (Zürich → Zurich). Before, they were cut off at the first accented character or filled with ?. This also changes the D890.
• Channel-fit check: ExpandedChannel is now Clone. The D578 seed lists the air band as receive-only, so the codeplug screens stop calling air channels out of range.

Follow-up for the D890 (not in this PR): #143.

🤖 Generated with Claude Code