Conversation
Signed-off-by: Loqor <isaiahcrace@gmail.com>
github-actions
Bot
added
size/L
github-actions
Bot
added
S: Untriaged
fix: issues with disconnecting = crash fix: log spam fix: heads of entities would spin like crazy Signed-off-by: Loqor <isaiahcrace@gmail.com>
fix: booth had incorrect transparent door boolean fix: player animations now render fix: entity heads no longer fritz out bug: animated textures dont.. animate? bug: particles dont render Signed-off-by: Loqor <isaiahcrace@gmail.com>
…amera Signed-off-by: Loqor <isaiahcrace@gmail.com>
…red() for no reason Signed-off-by: Loqor <isaiahcrace@gmail.com>
…lock positions is apparently not accurate enough! Signed-off-by: Loqor <isaiahcrace@gmail.com>
fix: block entity rendering fix: block entities from ait are now animated! fix: vortex and other skyboxes now render properly through portals, along with (fast) clouds Signed-off-by: Loqor <isaiahcrace@gmail.com>
fix: isOpen is now the inverse of isClosed fix: vortex and other skyboxes properly render fix: entities render properly across both sides of the portal Signed-off-by: Loqor <isaiahcrace@gmail.com>
drtheodor
added
the
S: DO NOT MERGE
Signed-off-by: Loqor <isaiahcrace@gmail.com>
The clear-by-draw fallback drew its full-screen quad at NDC z=1.0, exactly on the far clip plane. The GL spec treats z==w as inside the frustum, so NVIDIA keeps those fragments, but Apple's GL driver clips them away. With zero fragments the stencil clear never ran (stale mask -> interior smear) and the depth clear never ran (interior had no depth to test against and drew over the whole screen -> world went white). Draw the quad at z=0 instead, and clamp glDepthRange to [1,1] while writing depth so every fragment still stores the far value without riding the clip boundary. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Both glClear and clear-by-draw fail identically on the affected Mac with the interior smearing full-screen, which is the signature of the afbo having no usable stencil storage there (glStencilFunc(EQUAL,1) then always passes). Dump the FBO status, stencil attachment type/bits and GL vendor/renderer once so we can confirm before touching the clear path again. Revert after diagnosis. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apple's GL-over-Metal rejects framebuffer attachment introspection (GL_INVALID_ENUM), so the previous size/type queries were inconclusive. Replace them with a functional probe: set stencil=1 in a tiny centre region, paint red where stencil==1 over black, then read back a centre and a corner pixel. corner red => stencil never gates (the smear); centre black => writes dropped. Revert after diagnosis. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…temporary) The functional probe proved stencil gating works on the affected M4, so the clear was never the problem. Instrument the real pass instead: log the live viewport/scissor/bound-fb at stencil-clear time, and read back the afbo's actual stencil values right after the mask draw (centre vs corner). corner==1 means the mask spread stencil across the whole buffer -> the real smear cause. Revert after diagnosis. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ary) Two-point stencil sampling read 0/0 after the mask, but that is ambiguous: the doorway may not sit under the sample pixels, and GL_STENCIL_INDEX readback may be silently unreliable on Apple's GL-over-Metal. Add a self-test that reads stencil back where the functional probe KNOWS it is 1, and replace the 2-point sample with a full-buffer histogram (max/ones/nonzero) so we can tell whether the mask wrote any stencil at all. Revert after diagnosis. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Stencil, mask and gating are all correct on the M4 (23% doorway, clean), so the smear is downstream in the blits. copyFramebuffer blits COLOR|DEPTH in one call, but the main FB is DEPTH32F_STENCIL8 and the afbo is DEPTH24_STENCIL8 - a depth blit with mismatched formats is GL_INVALID_OPERATION per spec and copies nothing, which on Apple's strict driver would leave the never-cleared afbo stale (the smear). Log glGetError after each blit to confirm. Revert after. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…Apple Root cause of the Mac "smear", confirmed by instrumenting the real pass on an M4 (stencil, mask and gating all measured correct; the depth-inclusive blits returned GL_INVALID_OPERATION): copyFramebuffer blitted GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT in one call, but the main framebuffer and the afbo have different depth formats. A glBlitFramebuffer that includes depth requires identical depth formats or it is GL_INVALID_OPERATION and copies nothing - all or nothing. NVIDIA tolerates the mismatch; Apple's GL-over-Metal enforces it, so the whole blit (including the colour copy) failed. Because the afbo is never cleared, its colour never refreshed and last frame's pixels accumulated -> the interior smeared across the screen / white. Blit colour and depth as separate calls so the colour refresh can never be blocked by the depth copy failing, and swallow the depth blit's error on drivers that reject the mismatched format (depth occlusion no-ops there; the visible smear is fixed). NVIDIA behaviour is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Framebuffer-agnostic composite refactor across the four BOTI variants, gated on an in-game instrumentation step that determines the correct hook. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…t (Iris-safe) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…nder-scale) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… (Phase B M1) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The Phase B dep commit hardcoded a local nix-store JDK path, which breaks other contributors and CI. Pass the JDK at invocation instead: nix develop --command bash -lc './gradlew <task> -Dorg.gradle.java.home="$JAVA_HOME"' The daemon-picks-JDK25 issue is a local ~/.gradle/gradle.properties override. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
M1 in-game gate: a nested WorldRenderer.render() double-composites the main world (Iris's final pass writes to the main framebuffer) and avoiding it needs deep Iris-internal render-target redirection. Pivoting to gbuffer-injection. Iris modCompileOnly + isShaderPackInUse guard are kept (injection uses them). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…tion design/plan In-game gate: injecting the interior into the main gbuffer at AFTER_ENTITIES with setPhase(TERRAIN_SOLID) comes out shaded by the pack, no doubling. Adds the throwaway probe (IrisPhase, GbufferInjectionProbe, WorldGeometryRenderer.debugInjectTerrainIntoGbuffer), the gbuffer-injection design spec, and the M1 (doorway clip) plan. M1 reworks the probe into a real clipped injector. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Extract TardisDoorBOTI.drawDoorApertureMask() as a reusable static helper that stamps stencil=1 into the doorway aperture using the exterior variant's portal transform. GbufferInjectionProbe now uses it to clip the terrain injection: logs GL_STENCIL_BITS once, draws the aperture into stencil with the door's camera-relative transform, restricts the terrain draw to stencil==1, then fully restores all GL stencil state. Falls back to unclipped draw if GL_STENCIL_BITS==0 or no door is in the render queue. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Register WorldRenderEvents.START callback (Iris-only) that ensures client.getFramebuffer() has a stencil attachment before any draw this frame; setIsStencilEnabled is idempotent and handles shaderpack reloads. - Remove GL11.glGetInteger(GL11.GL_STENCIL_BITS) from GbufferInjectionProbe (emitted GL_INVALID_ENUM in core GL profile); replace with AITRenderHelper.getIsStencilEnabled(client.getFramebuffer()). - Add IrisPhase.setTerrainCutoutMipped() and setTerrainCutout(); refactor the set path to a private helper. - Extend WorldGeometryRenderer.debugInjectTerrainIntoGbuffer() to draw RenderLayer.getCutoutMipped() and RenderLayer.getCutout() after solid, each wrapped in the matching IrisPhase + reset(). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Cache lastPortalCamera + lastPortalWorld in WorldGeometryRenderer.render() for reuse by the gbuffer injector (one frame stale, acceptable for probe) - IrisPhase: add setBlockEntities() and setEntities() wrapping the BLOCK_ENTITIES and ENTITIES WorldRenderingPhase values - WorldGeometryRenderer.injectBlockEntitiesAndEntities(tickDelta): new public method that mirrors render()'s model-view+projection save/restore exactly, draws BEs wrapped in BLOCK_ENTITIES phase and entities in ENTITIES phase, with RenderSystem.depthMask(true) so injected geometry writes depth - debugInjectTerrainIntoGbuffer: add enableDepthTest + depthMask(true) guard - GbufferInjectionProbe: call injectBlockEntitiesAndEntities after terrain in both the stencil-clipped path and the unclipped fallback Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…tity injection Addresses task review: injectBlockEntitiesAndEntities left the BE/entity dispatchers pointed at the portal world between AFTER_ENTITIES and END, and both injection methods forced depthMask(true) without restoring it. Both now restore in their finally blocks (dispatchers mirror render()'s restore). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ns the interior)
The inside-out door view was rendering twice under a shaderpack: Phase A's
unshaded afbo composite AND the shaded gbuffer injection ("two rendered
parts"). Keep geometry.render() (it caches the portal matrices/VBOs the
injection reuses) but skip only the afbo->screen blit when a shaderpack is
active, so the injection is the sole visible interior. The door frame is still
drawn (shaded) by the normal DoorRenderer pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ar (experimental) Evidence-driven progress (all behind isShaderPackInUse): - Confirmed Iris's bound gbuffer has 8 stencil bits (valid glGetFramebufferAttachmentParameteri query). - Root-caused the clip bleed: DOOR_RENDER_QUEUE is empty at AFTER_ENTITIES under Sodium; now cache the interior door in BOTI.LAST_RENDERED_DOOR from TardisDoorBOTI (END) and reuse it in the injector. - Added BOTI.clearDepthInStencilRegion() and punch a depth hole in the aperture so the portal draws over blocks behind the door. KNOWN-OPEN (resume here): the aperture depth-clear corrupts the door frame's depth -> frame flicker; smear persists (likely shaderpack TAA). Needs a frame-safe depth approach (tight aperture to the opening, or re-render the door frame after injection) + a TAA decision. See ledger. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…gression) Keeps this turn's real win - the injection now correctly stencil-clips to the doorway (via the BOTI.LAST_RENDERED_DOOR cache; Iris's gbuffer confirmed to have 8 stencil bits). Removes the depth-clear that wiped the open door panels' depth and made them flicker. Depth-occlusion of the portal by blocks behind the door, and the frame-safe depth approach, are the next designed step (see ledger). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The door now correctly occludes the injected portal instead of flickering: - clear depth in the aperture so the portal draws over the room/blocks behind the door; - inject the portal world (shaded by Iris); - re-render the door block entity on top, still stencil-clipped to the aperture, with normal depth test/write, so the open panels/frame occlude the portal and their pixels+depth (wiped by the clear/injection) are restored. Uses the block-entity dispatcher so the door's exact transform/animation are reused. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…iew + translucent occlusion) - Clear the stencil buffer each frame before stamping the aperture. We added the stencil attachment ourselves and MC/Iris don't clear it, so it was accumulating every past door position -> the injection smeared across the screen. This fixes the smear. - Recompute the portal view from the CURRENT camera at AFTER_ENTITIES (updatePortalView) instead of reusing the 1-frame-stale END-render view. - Flatten the aperture depth to NEAR after injecting so translucent geometry behind the door (glass) is occluded by the portal instead of showing through. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…r, sky backdrop - Only inject when the door is actually open (isRemoved + leftRot/transparent check), dropping the cache otherwise: no portal over closed/destroyed doors. - Re-render the door with the CONTEXT matrix stack (camera view) instead of a fresh identity stack, which was rendering it billboarded to the camera; also re-clear aperture depth before it so it reliably occludes the portal. - Paint the exterior fog colour as the sky backdrop in Iris's SKY phase so it's unlit (was blowing out to white as lit gbuffer geometry). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…dering Signed-off-by: Loqor <isaiahcrace@gmail.com>
Signed-off-by: Loqor <isaiahcrace@gmail.com>
github-actions
Bot
added
the
S: Merge Conflict
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
# Conflicts: # src/main/java/dev/amble/ait/client/AITModClient.java # src/main/java/dev/amble/ait/client/boti/BOTI.java # src/main/java/dev/amble/ait/client/boti/TardisDoorBOTI.java # src/main/java/dev/amble/ait/client/boti/TardisExteriorBOTI.java # src/main/resources/ait.mixins.json
Signed-off-by: Loqor <isaiahcrace@gmail.com>
github-actions
Bot
removed
the
S: Merge Conflict
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
2 participants
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.
About the PR
Rewrite the entire system from the ground up to ensure functionality and compatibility. Will be tested by beta testers and on multiplayer for discrepancies.
Why / Balance
DOCTOR WHO TARDIS BAHAHAHAHA
Technical details
Media
Requirements
Breaking changes
Changelog
🆑