Code review P2: small verified bug fixes (zero seed impact)

[HansGR]

Second of five stacked code-review PRs. Based on the P1 infrastructure PR — merge that first; until then this diff includes P1's commit. This tier's own commits are 7573784..8fbdb37.

What's included (5 commits, one per fix)

• 7573784 — data/character.py: inverted init_run_success setter. It stored value - MAX_RUN_SUCCESS (negative) where the getter expects MAX_RUN_SUCCESS - value; a set/get round-trip returned the wrong value and the negative raw value would corrupt the bit-packed byte at init_data[21]. Latent: nothing calls the setter today, so no seeds are affected — but the first feature to use it would silently corrupt character init data. Regression tests added.
• 2fdc731 — objectives/objective.py: NameError in a failure path. The assert message referenced bare suplex_train_quest_name instead of cls.suplex_train_quest_name, so a failed quest lookup crashed with NameError instead of the intended diagnostic. Now an explicit RuntimeError (also survives python -O).
• 8359042 — args/challenges.py: dead code. An unreachable return opts (opts was never defined) and a duplicate definition of _format_spells_log_entries whose first copy was silently shadowed. The surviving definition is the one that already ran, so no behavior change.
• 59d8092 — event/whelk.py: dead check. self.reward.type == RewardType.NONE could never be true: Reward.type is initialized to Python None and RewardType.NONE is never assigned anywhere (verified by grep).
• 8fbdb37 — three silent-failure fixes: utils/compression.py printed an error and returned data with a truncated size header when output exceeded 65535 bytes (silent corruption for the consumer) — now raises ValueError, with a test; metadata/flag_metadata_writer.py had if type(group_title): which is always true (a type object is truthy) — replaced with explicit str/None/callable dispatch preserving the manifest output exactly; battle/scaling.py labeled the xp/gp routine "scale hp/mp" in its space description (debug output only).

Seed compatibility

Verified byte-identical ROMs (SHA-256) against the pre-change baseline for: the default flagset, a 20-flag heavy flagset, and a -rls ultima,quake build specifically exercising the modified challenges.py log path. Flag-metadata JSON output is also byte-identical. Spoiler logs identical.

https://claude.ai/code/session_01QnF26DB2TYi1hpJE8L6EJr

[gemini-code-assist]

Code Review

This pull request introduces a comprehensive suite of unit tests under the tests/ directory, updates documentation, and refactors memory handling by introducing a specific RomSpaceError to prevent out-of-bounds ROM corruption. It also addresses several bugs, such as correcting the character run success setter logic, removing duplicate functions, and refining shop item swapping. Feedback on the changes highlights a critical issue in data/shops.py where removing a safety check on no_dried_meat_shops could result in an IndexError if the list becomes empty.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Removing the safety check if not no_dried_meat_shops: break can lead to an IndexError when self.args.shop_dried_meat is configured to be larger than the number of available shops (e.g., due to exclusions or empty shops). It is highly recommended to restore this guard to prevent potential crashes during seed generation.

            for index in range(number_shops_with_dried_meat, self.args.shop_dried_meat):
                if not no_dried_meat_shops:
                    break
                random_shop = random.choice(no_dried_meat_shops)

[HansGR]

This branch was not correctly rebased to dev prior to the pull request, which caused this issue. This and all other pull requests in this series should now be correctly rebased to dev.