Conversation
Minecraft 26.1 ships fully deobfuscated, which the existing modstitch 0.7.0-unstable/Loom 1.14.6 toolchain has no concept of. Bump modstitch to 0.8.5 (the version that added unobfuscated-MC support, splitting the `loom` platform into `fabric-loom-remap` for existing obfuscated versions and `fabric-loom` for 26.1+), Loom to 1.17.20, Gradle to 9.5.1, Fabric Loader to 0.19.5, and Kotlin to 2.4.20 (needed for the new Java 25 target on 26.1+). Adds versions/26.1.2-fabric (26.1's second patch, since Architectury 20.1.14 requires >=26.1.2) and versions/26.2-fabric with researched dependency coordinates. Existing 1.20.1-1.21.11 fabric versions are updated for the platform rename only and remain otherwise untouched; verified they still compile after the shared toolchain bump.
Minecraft 26.1 renamed GuiGraphics to GuiGraphicsExtractor and moved from immediate-mode rendering to a CPU-side render-state extraction model: Screen/Widget `render()` became `extractRenderState()`, AbstractButton's override point became `extractContents()`, AbstractSelectionList.Entry's became `extractContent()`, and `hLine`/`vLine`/`drawString` were renamed to `horizontalLine`/ `verticalLine`/`text`. EditBox also dropped `setFilter` entirely. Adds a `GuiGraphics` typealias (-> GuiGraphicsExtractor) and a `Renderable.render()` extension shim so the ~20 files that just reference the type or call `.render(...)` on a widget keep compiling unchanged; adds `hLine`/`vLine`/`drawString` extension shims backed by the new `fill`/`text` methods. The handful of files that directly override a renamed Mojang method (FABaseScreen, TextOnlyButton, CycleTextOnlyButton, EnrouteWaypointsList, SystemManagementList) get a new `>=26.1` stonecutter branch alongside the existing ones. TypeStrictEditBox replaces `setFilter` with an `insertText` override that reverts invalid input, replicating the old behavior. This is compatibility work only - no HUD behavior or appearance is intended to change on existing versions.
Three separate issues in the world-space -> screen-space pipeline that fed the attitude ladder, heading tape, and flight path vector, found by comparing against Octol1ttle#92: - LevelRenderer.renderLevel's three matrices (frustum/projection/ culling-projection) collapsed into a single Matrix4fc parameter on 26.1. It's the frustum/view matrix, not the projection matrix as first assumed - the real projection matrix now lives on CameraRenderState.projectionMatrix. Using the wrong matrix in the wrong role corrupted every downstream world-space HUD calculation. - ScreenSpace read the GL viewport via glGetIntegerv, which is stale during 26.1's new CPU-side render-state extraction phase. Now reads mc.window's dimensions directly instead. - 26.1 uses a reversed-Z depth range, which broke the existing on-screen check: points behind the camera were mirrored back into the visible depth window instead of being rejected, so e.g. both ends of the pitch-limit indicator would render regardless of actual camera orientation. fromWorldSpace now explicitly rejects points at or behind the camera plane (clip.w <= 0) before projecting. Also stopped reading RenderSystem.getModelViewMatrix() on 26.1+, which is equally stale during extraction - it's now taken as identity, matching the reference fix.
flightPitch/flightYaw derive the direction of travel via velocity.normalize(), which amplifies floating-point noise into an essentially random angle once velocity gets close to zero (e.g. standing on the ground, or just after landing). StallComputer feeds that angle straight into its max-pitch protection target, so it would intermittently clamp down to near-zero for no real reason while barely moving. Fall back to the player's actual look-based pitch/yaw below a small velocity threshold, where the travel-direction angle isn't meaningful anyway.
- toggle_auto_thrust toggles Auto Thrust on its own, addressing half of the long-standing TODO on the old autopilot_disconnect binding - toggle_all_autopilot flips Flight Directors, Auto Thrust, and Autopilot together, mirrored off the current autopilot state - toggle_holding_pattern enters/exits a new HoldingPatternLateralMode: it holds at the flight plan's current target waypoint if one exists (via a newly-public FlightPlanComputer.getEnrouteTarget), otherwise at present position and heading. The mode alternates between two points offset from the fix to trace a racetrack, delegating the actual steering to DirectCoordinatesLateralMode. Also added as a HOLD option in the Autoflight setup screen for manual fix/course entry All three keybinds are unbound by default, same as the existing toggle_enabled binding.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
versions/26.2-fabric, though only 26.1 has been built and run-tested hereWhat changed
Toolchain
0.7.0-unstable->0.8.5(the version that added unobfuscated-MC support, splitting theloomplatform intofabric-loom-remapfor existing obfuscated versions andfabric-loomfor 26.1+)1.14.6->1.17.20, Gradle9.2.1->9.5.1, Fabric Loader0.16.10->0.19.5, Kotlin2.2.0->2.4.20(needed for the new Java 25 target on 26.1+)versions/26.1.2-fabricandversions/26.2-fabricHUD rendering port (Minecraft 26.1's
GuiGraphicsrewrite)GuiGraphics->GuiGraphicsExtractortypealias, plusRenderable.render()/hLine/vLine/drawStringextension shims so most call sites don't need to changeextractRenderState/extractContents/extractContentstonecutter branches added where a Screen/Widget/list-entry directly overrides the renamed Mojang methodTypeStrictEditBox: replaced the removedEditBox.setFilterwith aninsertTextoverride that reverts invalid inputLevelRenderer.renderLevel's matrix parameters restructured forCameraRenderState/singleMatrix4fcProjection fixes
(Found by comparing against #92 partway through debugging - see note below)
Matrix4fcparameter onrenderLevelis the frustum/view matrix, not projection - the real projection matrix now lives onCameraRenderState.projectionMatrixScreenSpace's viewport now reads frommc.windowinstead of the stale GL viewport during 26.1's CPU-side render-state extraction phasefromWorldSpacenow explicitly rejectsclip.w <= 0Bug fix found during testing
AirDataComputer.flightPitch/flightYawwere numerically unstable at near-zero velocity (normalizing a near-zero vector amplifies floating-point noise into an arbitrary angle), which intermittently madeStallComputer's pitch protection clamp far more aggressively than intended while barely moving. Falls back to look-based pitch/yaw below a small velocity threshold.New features
toggle_auto_thrust,toggle_all_autopilot,toggle_holding_patternkeybinds (unbound by default, same astoggle_enabled)HoldingPatternLateralMode: holds at the flight plan's current target waypoint if one exists, otherwise at present position/heading; built on top ofDirectCoordinatesLateralModeHOLDoption added to the Autoflight setup screen for manual fix/course entryRelationship to #92
This was developed independently and only found #92 partway through debugging the reversed-Z projection issue - comparing against it confirmed the root cause and shape of the fix (credited in the relevant commit). There's likely overlap in
ScreenSpace.kt,GuiGraphics.kt, and theGuiGraphicsextension shims between the two PRs since they solve the same underlying rendering-model change; probably worth reconciling before merging both rather than merging separately.Validation
:26.1.2-fabric:buildsucceeds; mod boots and runs in-game (HUD, mixins, and mod init all verified live):1.20.1-fabric,:1.20.1-forge,:1.21.1-neoforge, and:1.21.11-fabricverified to still compile after the shared toolchain bump:26.2-fabrichas researched dependency coordinates but has not been built or run