ATtiny1634 GPIO smoke — first attempt failed (SBI on PINB)
First run of the new simavr-attiny1634 GPIO smoke, after the CPU model gained
ATtiny1634 port mirroring and PINx write-one-to-toggle (data 0x27/0x2B/0x2F),
but before SBI/CBI learned the ATtiny1634 PINx addresses.
- Run:
scripts/run_renode_onboarding_inventory.py --only simavr-attiny1634, 100,000 instructions, stop reasonmax_steps. - Result: FAIL 2/3.
memory_value @0x130expected0x7F, got0x6F. The PB0 edge and stop-reason assertions passed. - Missing bit
0x10is the "SBI on PINB toggles PB0 only" check. SBI was executed as a read-modify-write of PINB: with PB3 driven high it wrote0x09back to PINB, which toggled PB3 as well as PB0. - The new unit test
attiny1634_ports_mirror_and_pin_togglefailed at the same point: PORTB was0x01afterSBI PINB,0, expected0x09.
Cause: write_one_register in crates/core/src/cpu/avr/exec/bitops.rs
fell back to data_addr == 0x36 for every toggle-capable model not named in
its match, so the ATtiny1634's PINB at 0x2B was not treated as write-one.
The fix gives the ATtiny1634 its own arm and makes the fallback false.
junit.xml and run_smoke.log here are copied from that failed run.