- Rust 79.3%
- Shell 12%
- Nix 5.9%
- Python 2.8%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
Ivan
2553f89600
Same shape as cli-utils' yt-audio release: CI on the tag (version matches Cargo.toml, clippy, nix build with its tests, a publish dry run), then cargo publish and a Forgejo release from the annotated tag's message, each skipped if already done. The package gets a LICENSE, repository metadata and an include list, so crates.io gets the crate and its data, not the game harness. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> |
||
| .forgejo/workflows | ||
| data | ||
| game | ||
| src | ||
| tests | ||
| .gitignore | ||
| build.rs | ||
| Cargo.lock | ||
| Cargo.toml | ||
| flake.lock | ||
| flake.nix | ||
| LICENSE | ||
| README.md | ||
mcopts
A library for managing your Minecraft options. It sounds ridiculous until you try to.
Read, check, merge and sync Minecraft's options.txt and mod configs, for
Minecraft 1.21 to 26.3.
With the tested modern keybinds, every supported client logs Failed to load options and falls back to defaults if version: is missing. This does not
mean every versionless file fails: legacy files can be migrated. Unknown
options are dropped when the game saves. mcopts knows each version's options
from what the game writes (data/defaults/<version>.txt); check validates
keys and value kinds, not every range or enum. It warns about the graphics
preset interactions observed below; warnings do not change the exit status.
Mod keybinds are allowed for modded instances, but vanilla drops them on save.
mcopts versions # name, data version, option count
mcopts check -m 1.21.1 options.txt # check version, keys and value kinds
mcopts apply -m 1.21.1 out/ dir1 dir2 # lay option dirs over out/, in order
mcopts sync instance/ dir # take an instance's preferences
mcopts diff a.txt b.txt # options in a that b lacks or changes
An option directory mirrors the game folder: its options.txt merges key by
key and gets the version's version:, and every other file (mod configs)
replaces the target's. sync leaves out configs that are not a player's
preferences (data/not-preferences.txt).
Proving it against the game
game/ launches real clients offline on Xvfb, never on the desktop, one at a
time, silently, each held to 2 GiB of heap. It inspects screenshots, logs and
the options.txt the game writes:
nix develop .#game
game/collect.sh # data/defaults/<v>.txt from a fresh launch of each
game/verify.sh # apply + check hutao's options, launch, compare
Collection clicks Continue on the first-launch accessibility screen; quitting alone does not save a fresh 1.21.x instance's options. Verification opens Options and leaves with Escape, and checks that a save occurred: F11 twice did not save in 1.21. Only keybinds present in that version's vanilla defaults are compared; mod binds remain in the applied file but require their mods to survive a save.
In 1.21.11, 26.1, 26.1.1, 26.1.2, 26.2 and 26.3, graphicsPreset:"fancy"
overwrote entityDistanceScaling:0.75 and simulationDistance:6 with 1.0
and 12. Use graphicsPreset:"custom" to preserve individual settings.
check and apply warn when a non-custom preset accompanies either of those
keys in those versions. Verification still fails on changed vanilla values.
Game files are cached in ~/.cache/mcopts-game.
Forgejo CI runs nix flake check and cargo test on every push, including
tags. A separate game workflow verifies the versions whose defaults changed;
for anything else (game/ or the options fixture changed, a manual run) it
verifies the oldest and newest, which cover both launch paths. All 17 is a
local game/verify.sh. It uses tests/fixtures/hutao, a copy of hutao/vps's options (vps
checks its live ones with mcopts check in its own build), and caches game
downloads when the runner supports caching.
The game workflow does not gate releases.
Development
nix develop # cargo, clippy, rustfmt, pre-commit hooks
cargo test
nix build