mirror of
https://github.com/MCKero6423/uv-k5-v3-emulator.git
synced 2026-10-07 05:27:21 +00:00
mckero
56d87cfaef
Round 102's client, unchanged, against the current binary: zero low-row reads and no probe output at all. Round 104's own rule covers it, so the first job was to make the probe prove it can see anything -- and it can. Counting IDR reads per port, with the keypad port's first few logged unconditionally: the firmware reads port B's input register about three million times, and in every one of those reads the four row bits 12..15 come back high, with F held for eight seconds (v = 0x0000fc87). That is narrower than anything before it. The model pulls a row low only when a held key sits on a currently-low column; the firmware reads the rows three million times and never sees one low; so the model's notion of which cell is held does not line up with the cell the firmware's own 5x4 keyboard table looks in. Two notes: the sanity line was added only after the replay came back empty, which is the right order, since an empty probe is a claim rather than a result; and its first version logged the first five IDR reads of any port, which were all port F, so the keypad was never in view while the instrument looked like it was working. Counting per port is what turned that into the number above. Next: put the machine's keypad cell mapping beside App/driver/keyboard.c's keyboard[5][4] table and find the disagreement.