Supported versions¶
This page says exactly what has and has not been tested. Read "untested" literally: nobody has run it there yet.
How to read the matrix¶
| Word | Meaning |
|---|---|
| Verified | The acceptance checklist passed on real hardware. Nothing is verified yet. |
| Preview | Works in limited testing. The manifest calls this state "experimental". |
| Available | Installs. Xol makes no integration claims about it. |
| Not offered | Hidden from the launcher's picker |
| Smoke test | A headless run on Linux x64 with software rendering and a diagnostics session. It proves install, launch and join. It proves nothing about performance, audio, input or graphics drivers. |
| Untested | Not run at all |
Profiles¶
| Profile | State | Linux x64 (development) | Windows x64 | macOS Intel | macOS Apple Silicon |
|---|---|---|---|---|---|
| Xol 1.8.9 (Forge 11.15.1.2318 and the Xol client) | Preview | Smoke test: title screen, joined a private Paper 1.8.8 server, HUD drawn, brand Xol |
untested | untested | untested; x64 Java through Rosetta |
| Xol 1.7.10 (Forge 10.13.4.1614, UniMixins 0.3.1, GTNHLib 0.11.49, Hodgepodge 2.7.208, ArchaicFix 0.8.0 and the Xol client) | Preview | Smoke test: joined a private vanilla 1.7.10 server, HUD drawn, brand Xol, Forge's registration intact |
untested | untested | untested |
| Forge 1.8.9 | Available | Installs and boots, as the base of Xol 1.8.9 | untested | untested | untested |
| Forge 1.7.10 | Available | Installs and boots, as the base of Xol 1.7.10 | untested | untested | untested |
| Forge + Optimization 1.7.10 | Preview | The same stack as Xol 1.7.10; not launched separately | untested | untested | untested |
| Forge + Optimization 1.8.9 | Not offered | No redistributable stack whose files can be pinned by checksum exists | |||
Xol 1.8.9 with OptiFine (optional import into mods/) |
Preview | Import and resolution unit-tested with a synthetic jar. Never launched with real OptiFine. | untested | untested | untested |
| OptiFine 1.8.9 (your file, through launchwrapper) | Preview | Import flow and its error path tested in the launcher with a synthetic jar. The launch path reached the title screen with a synthetic tweaker that does nothing. Never launched with real OptiFine. | untested | untested | untested |
| OptiFine 1.7.10 (your file, through launchwrapper) | Preview | The same code path, unit-tested only. Never launched. | untested | untested | untested |
| Xol 1.7.10 with OptiFine | Not offered | Hodgepodge, in the pinned stack, conflicts with OptiFine. The launcher says so instead of offering the import. | |||
| Modern versions (Fabric and later) | Not started |
Not tested anywhere: Minecraft itself on Windows or macOS of any architecture, macOS on Intel hardware, a real graphics card, audio, input devices, sustained play, performance.
Linux appears in the matrix because development testing happens there. No Linux build is published.
The launcher app¶
| What | Where | Result |
|---|---|---|
| The main screen, settings, repair, clone, the redacted support log, and a real 1.8.9 launch from the interface | Linux x64, headless | Run and checked, with unit tests |
Xol-<version>-macos-arm64.dmg |
An automated macOS 14 arm64 build machine | The disk image mounts, Xol.app has the right architecture, codesign --verify passes (ad-hoc signature), the app stays up for 12 seconds |
Xol-<version>-macos-x64.dmg |
The same machine, through Rosetta | The same checks |
Xol-<version>-windows-x64-setup.exe |
An automated Windows Server 2022 build machine | Silent install, Xol.exe present, silent uninstall keeps %APPDATA%\Xol\instances |
| Self-update (0.2.2) | The same two machines | macOS: staging, swap, signature check, start, clean-up, for both architectures. Windows: a silent update over an install, after which Xol starts again. |
No person has walked through the interface on Windows or macOS yet, and no Minecraft launch was done on those machines. The builds are unsigned development builds; on macOS they carry an ad-hoc signature only.
Other parts¶
| Part | State |
|---|---|
| Microsoft sign-in | Implemented. Never exercised against a real account, because it needs an approved application ID. |
| Signed update metadata: signature, pinned keys, replay protection, authorised rollback | Unit-tested, and checked end to end against the production service with a development key: the offer was accepted, the downloaded bytes matched the signed hashes, a replay was refused, a revoked key was refused |
| Access key, announcements, remote manifest and configuration, diagnostics | Unit-tested and run in the interface against a local stand-in service; the update, manifest and configuration calls were also checked against the production service |
| The in-game client 0.2.0 | Run on 1.8.9 (GUI scales 1 to 3) and 1.7.10 on Linux with real input. Some parts are compiled and unit-tested only; they are listed on The client. |
Minimum operating systems¶
Windows 10 or 11 x64, and macOS 12 or later. These are planned minimums and have not been verified on hardware. macOS 12 is what the application framework Xol is built on requires; builds up to 0.2.2 were packaged with macOS 11 written into them as the minimum, which was a mistake in the packaging settings and is corrected from the next build.
Performance¶
There are no benchmark results yet, and nothing on this site may be read as a performance claim. The development machine has no graphics card, so frame rates from it would be meaningless for players.
The planned method: the same machine, world, resolution (1920 × 1080 windowed), options, Java runtime and heap for every configuration; a 60-second warm-up, then three runs of 120 seconds on a fixed camera path; average FPS, 1% low, frame-time percentiles, heap after garbage collection and start-up time. Changes in visual quality (presets) will be reported separately from gains in the code. The configurations to compare are clean vanilla, Xol with modules on and off, OptiFine, and for 1.7.10 Forge alone against Forge with the pinned stack.
Because optimization on 1.8.9 has not been measured, Xol 1.8.9 cannot become "verified" yet.
One fact is true by construction rather than by measurement: the FPS module allocates nothing per frame while its value is unchanged.