🩹Neo Angband: Bug Fixes v0.16.0

Added

• Text corrections, a new section, on by default. Angband 4.2 has one weapon slot and no two-handed rule, so nothing stops a character wielding a Pike and a shield, and four gamedata descriptions still say otherwise: the Two-Handed Great Flail, the Pike, the Trident of Wrath and Mundwine. The core game keeps upstream’s text exactly, so the correction lives here. Text only, and no weapon’s damage, weight or slot changes.
• The section owns the whole class of defect, “the description says something the game no longer does”, so a future correction joins it rather than adding another row to the options menu.
• The test derives each record reference from the real content pack rather than mirroring the naming rules, and asserts both directions: that the core game still makes the claim, and that the replacement drops it. Without the first half this would keep passing after core dropped the wording on its own, leaving a section advertising a fix it no longer performs.

Fixed

• The install section claimed the game checks every file against a SHA-256 that ships inside it, and that a replaced tag or an intercepted download therefore fails rather than runs. The game does not do that. It records a digest of the bytes that arrived, which answers whether the copy on your machine has changed since it was installed, and cannot answer whether what arrived is what was published here. What the install does give is a pinned tag rather than a branch, so what arrived cannot change under you afterwards.
• The build section said the game’s content pack and plugin builder come from a checkout of the game rather than from npm. Both have been published packages since 0.12.0, so npm ci is the whole setup. A sibling checkout is an override for developing against an engine change that has not reached the registry yet, and the builder’s own documentation had the same order backwards.

Changed

• Tested against engine and content 0.24.0 rather than 0.19.0. A mod tested against an engine five release lines older than the one it installs onto has been tested against the wrong thing.
• The four pinned stranded-floor seeds were re-measured and replaced. Two came back healthy under 0.24.0’s level generation and a third changed direction, which is the staleness this table has always been expected to accumulate. Only the control assertion can detect it, because a healthy floor passes the repair assertion trivially, so three of the four rows had quietly stopped testing anything. Each replacement is measured in both directions, and spares are recorded next to the table so the next reseal needs no scan.