UCI adapter: CIA1 TOD is never started, so the TOD-bounded waits (uci_wait_idle / uci_wait_not_busy / resp reads) are unbounded on real hardware #145

Found while chasing c64-wireguard#58 on a U64E (fw 3.15), using the uci_wait_idle design your adapter originated (c64-https#37) — this applies to the original as much as to our port.

Measurement

With the C64 hung inside net_udp_send, I hooked $0314 with a routine copying $DC08–$DC0B into RAM on every IRQ: 207 samples over 3 s, TOD = 00:00:00.0 every time, hour byte $91 (the CIA's untouched reset value). The CIA TOD does not run until its TENTHS register is written, and neither adapter ever writes it. So the cmp last_tenths / beq loop bound never advances: every "wall-clock bounded" wait is unbounded on hardware. (VICE's CIA starts the TOD running, which is why the emulator never shows this.)

Proof of mechanism: a one-shot sta $DC08 from the hook started the clock, and the hung wait expired within 2 s and returned C=1 with the adapter's error.

Fix (c64-wireguard src/net/uci/uci_cmd.s, uci_tod_start, called first thing in net_init)

lda $DC0F ; and #$7F ; sta $DC0F     ; CRB bit 7 = 0: writes set the clock, not the alarm
lda #0
sta $DC0B ; sta $DC0A ; sta $DC09    ; hours (halts), min, sec
sta $DC08                            ; tenths — starts the clock

Run it once before any bounded wait. Note every run_prg/reset stops the TOD again, so it belongs in init, not in a one-off.

Contract

Reported to the c64-lib-contract lane for a normative line in §13.4 ("adapter MUST start the TOD before the first bounded wait; a stopped TOD makes the bound infinite").