Files
CoCoCapitalQuiz/docs/ANSWER_SLOTS.md
T

46 lines
2.7 KiB
Markdown

# Nonrepeating correct-answer slots — November 11, 2026
The first multiple-choice question randomly places its correct answer in any
of the four slots. Each following question chooses equally among the other
three slots, excluding the previous question's correct slot. This follows the
correct answer's location even when the player chooses a wrong answer.
Distractor capitals remain unique and shuffled. `quiz_start` resets the slot
history for each new game; typed answers retain their existing behavior.
## Build and verification
This working change is based on revision `46f4b9c` plus the November 11 progress
map changes. The root `CAPITAL.DSK` is the Release build produced with installed
**VCC Plus 3.6.1 (54)** and its bundled tools. Its SHA-256 is:
`5dbc30b7c69f95f325cf3457ba88effb79837501f089f7615d22ae6a6b7ee6d6`
- Release, Debug, and Test disk builds succeeded within the existing memory limit.
- `make test` checked 100,000 option sets across 2,000 seeded games with no
consecutive repeated correct slots. All four starting slots and all twelve
permitted transitions occurred; choices stayed distinct and included exactly
one correct answer. Repeating each of 32 seeds after a complete game reproduced
its first options, verifying that the previous slot resets on replay.
- Existing capitalization, known-wrong capitals, aliases, distance, bounds,
scoring/reset, and all map-geometry checks passed. Python test syntax passed.
- The actual Release disk completed a 50-question multiple-choice quiz on each
emulated MC6809 and HD6309 using the installed app's isolated `vcc-playtest`
helper, 512 KB RAM, and private CoCo 3 / Disk BASIC ROMs. Correct positions were
decoded from the displayed choices and checked against the previous displayed
question. Neither run repeated a correct slot consecutively; all four slots
occurred. Deliberate correct and incorrect key selections tested scoring and
ensured the constraint did not follow the player's previous key.
- Those frame checks also verified the answer-history map colors, all three map
views, the final score, neutral replay title, and the first two replay
questions. Both runs reported zero disk writes and preserved the disk hash.
The public MCP regression driver `tests/playtest_mc.py` now checks adjacent
correct slots as it reads each question. The grant-based suite was not rerun;
this change was checked directly through the installed helper. Raw captures,
slot-sequence receipts, and logs remain in ignored local build storage. The
normal emulator and personal media were not used. No physical CoCo or host
Intel qualification is claimed by these checks.
See [the preceding progress-map record](MAP_PROGRESS.md) for the map design
and the unchanged 512 KB graphics-memory requirement.