issues · pear-git-v3

#129 Post-mortem 2026-10-19: L3 cable extras, #127 perf, L4 assets (deployed 0.1596.0 + 0.1618.0) open

by uds-local · 10/19/2026, 10:54:12 PM

#128 Editor fold latency: next optimization candidates (after #127) open

by uds-local · 10/19/2026, 4:02:59 PM

#127 Performance review: cable strands (tabled) and high-poly faceted shapes — per-edit full refold cost open

by uds-local · 10/19/2026, 12:21:17 PM

#126 Post-mortem 2026-10-18 (keetd): M9a–M9c-4c committed, M9c-5a hyperconf gated but UNCOMMITTED (ctest stopped at 140/145, 0 failed); bundles: none to port open

by uds-local · 10/18/2026, 10:23:56 PM

#125 Post-mortem 2026-10-18: #124 editor areas, #120 L1 brush modifiers, #120 L2 BoxCutter (deployed 0.1584.0) open

by uds-local · 10/18/2026, 10:16:16 PM

#124 Editor UI: panels block hover, modal ops own the mouse, one-tree organizer, and Blender-style areas (3D view as a panel type, resizable) open

by uds-local · 10/18/2026, 12:51:46 PM

#123 dc-terrain: merge_brushes / move_brush_points hull in float32 far from the origin (brush.merge / brush.reshape replay) open

by uds-local · 10/18/2026, 12:02:01 PM

#122 Post-mortem 2026-10-17 (hyper-tps): F5 curves, follow-curve, Shift+D/Alt+D shipped (0.1527.0); L3 cables mid-flight — hull precision blocker open

by uds-local · 10/17/2026, 9:56:54 PM

#121 Post-mortem 2026-10-17 (keetd): M8c complete (all 74 contract routes); M9 started — room engine gated single-writer open

by uds-local · 10/17/2026, 9:54:46 PM

#120 Editor plan 2026-10-17: node tree + modifier stacks, BoxCutter/HardOps, Cablerator + Random Flow as convex brushes — layered from foundations open

by uds-local · 10/17/2026, 12:03:25 PM

#119 Post-mortem 2026-10-16 (keetd): L8, M8a, M8b done; M8c started (hyperdispatch) — unshipped specs recovered and proven open

by uds-local · 10/16/2026, 11:24:24 PM

#118 Post-mortem 2026-10-16 (keetd): M6c, M6d, M6e, M7 done — the autobee port is complete open

by uds-local · 10/16/2026, 4:49:12 PM

#117 Post-mortem 2026-10-15 (keetd): Phase B Windows, 4e-4, M6a, R4g, R4h, M6b done — and a review of the line-by-line method open

by uds-local · 10/15/2026, 10:45:51 PM

#116 Post-mortem 2026-10-14 (keetd): R4f groups 1–4 done, group 5 ported (drill controls red, unresolved), next steps open

by uds-local · 10/14/2026, 9:32:00 PM

#115 Post-mortem 2026-10-13: creatures→bots (0.1473.0), shape parts, proportional editing, Mirror — 8 commits unpushed open

by uds-local · 10/13/2026, 9:37:05 PM

#114 Audit 2026-10-13: --dedicated mode — what it runs, what it should be, creature/bot options A–E, benchmark plan, security findings open

by uds-local · 10/13/2026, 11:10:00 AM

#113 Post-mortem 2026-10-12: editor keymap → Blender 2.7x steps 1–6.5 (live cuts, Q menu, F1 by context, layouts, overlays); 3 commits unpushed pending suite open

by uds-local · 10/12/2026, 10:43:01 PM

#112 Post-mortem: keetd R4c-4d E3 survivors → E4 caller-first → E4a(1a…2a-ii) + upstream parity maps (2026-10-12) open

by uds-local · 10/12/2026, 10:40:26 PM

#111 Post-mortem 2026-10-11: invite bug → Keet 4.23 lockstep (0.1432.0), stortron sidecar wedge, keep-2 retention, headless --dedicated at 765% CPU open

by uds-local · 10/11/2026, 11:37:40 PM

#110 Post-mortem: keetd R4c-4d E0→E3 + Keet 4.23 storage re-pin (2026-10-07 → 10-11) open

by uds-local · 10/11/2026, 11:35:52 PM

#109 Things to Work on open

by fleeky-tabertron · 10/9/2026, 9:36:12 AM

#108 Post-mortem 2026-10-06: Geiss rebuild → terraced world, collision/LOD/arrival fixes, atmosphere + audio overhaul, map-scope race (perf next) open

by uds-local · 10/6/2026, 10:10:06 PM

#107 Post-mortem 2026-10-06: keetd R3c landed + R4 async storage R4a → R4c-3b-3 (scaffold flaw, unfaithful deferrals, harness bugs, upstream findings) open

by uds-local · 10/6/2026, 10:06:42 PM

#106 Post-mortem 2026-10-05: hyper-tps editor + movement session (handspring built wrong, live dev tree broken, keys, faces/CSG/duplicate shipped) open

by uds-local · 10/5/2026, 10:14:03 PM

#105 Post-mortem 2026-10-05: keetd testing redesign (P0/P1, 8 bugs found), R3b complete, R3c in its gate; port review open

by uds-local · 10/5/2026, 9:46:04 PM

#104 Post-mortem 2026-10-04: undersea/P5 shipped, mountain speed measured, bot nav step 1 done / step 2 WIP open

by uds-local · 10/4/2026, 10:23:04 PM

#103 Post-mortem 2026-10-04: keetd M5g-4e + runtime R1→R3a; testing is now the bottleneck open

by uds-local · 10/4/2026, 10:20:57 PM

#102 JS removal for hyper-tps 1.0: inventory, replacements, owners open

by uds-local · 10/4/2026, 2:32:39 PM

#101 Bug (intermittent): carver ADDED instead of removing (Shift+carve too) — gone after restart, 0.1332.0 open

by uds-local · 10/3/2026, 9:30:18 PM

#100 Post-mortem 2026-10-03: keetd port M5g-3c → M5g-4d (state, bugs, upstream facts, next steps) open

by uds-local · 10/3/2026, 9:15:55 PM

#99 Post-mortem 2026-10-03: demo build tools — settings, UV/hotspots, rotate, editor E4/E5 (0.1332.0) open

by uds-local · 10/3/2026, 9:12:52 PM

#98 Post-mortem 2026-10-02: keetd port M5g-1 → M5g-3b (state, bugs, harness, next steps) open

by uds-local · 10/2/2026, 9:33:32 PM

#97 Post-mortem 2026-10-02: texturing → parallax tier, generated normals, editor E3, wheel open

by uds-local · 10/2/2026, 9:07:58 PM

#96 Post-mortem 2026-10-01: keetd port M5b→M5g-0 (state, bugs, backlog plan) open

by uds-local · 10/1/2026, 10:32:19 PM

#95 Post-mortem 2026-10-01: 1060 hitching, terrain GPU submit sizing, pgh never swaps a running app — and what's left for tomorrow open

by uds-local · 10/1/2026, 10:28:53 PM

#94 F4 editor: Scythe + TrenchBroom texturing parity — research, capability matrix, design and plan (hotspots, texture lock) open

by uds-local · 10/1/2026, 5:08:15 PM

#93 Demo / 1.0 roadmap: creative build mode on Mountain, Erosion, Flat — status review of all issues and the plan open

by uds-local · 10/1/2026, 3:16:28 PM

#92 keetd (C/C++ keet-daemon port) — status 2026-10-16: M0–M7 done (autobee port complete); next M9 keet-core or the vertical slice open

by uds-local · 9/30/2026, 10:02:57 PM

#91 Post-mortem 2026-09-30: mountain material perf -- mesh-time bake (#90), results, next steps open

by uds-local · 9/30/2026, 9:53:53 PM

#90 Mesh-time material bake: mountain material noise once per vertex, not per pixel per frame (close-up 22 ms -> ~1 ms target) open

by uds-local · 9/30/2026, 6:55:36 PM

#88 Post-mortem 2026-09-29: carve latency, LOD livelock, v0.1207.0 Windows release, deep perf audit + parked plan open

by uds-local · 9/29/2026, 10:01:04 PM

#87 Mountain clouds/spawn/CSG-detail shipped (uncommitted); moiré traced to Bump × slope thresholds on the noise lattice, root cause open — post-mortem + paths open

by uds-local · 9/28/2026, 10:45:08 PM

#86 Mountain load 228 s → ~3 s terrain / ~7 s with material; moiré root-caused → analytic mipmaps designed — post-mortem + next steps open

by uds-local · 9/28/2026, 3:06:47 PM

#85 voronoi5 end-to-end: rim fixed, native material groups, density cause found, 30.1 s -> 8.6 s — post-mortem + next steps open

by uds-local · 9/27/2026, 10:45:58 PM

#84 Bridged mountain v3+: geometry garbled — post-mortem, 6 fixes landed, 2 causes still open open

by uds-local · 9/22/2026, 10:49:20 PM

#83 Post-mortem 2026-09-21b: the bridged material reaches dc-terrain and still looks wrong — two unfixed defects, one emitter win, five broken rulers open

by uds-local · 9/21/2026, 10:45:21 PM

#82 Post-mortem: Godot 4.8 RTX migration, three DLSS-RR bugs, and the limb/frame-rate bug open

by uds-local · 9/21/2026, 9:33:55 PM

#81 Post-mortem 2026-09-20: DCTerrain honours its transform — and six claims I published as findings before measuring them open

by uds-local · 9/20/2026, 7:26:15 PM

#80 Post-mortem: bots reported "no nav source" — four stacked causes between a working nav graph and a bot you can see open

by uds-local · 9/15/2026, 10:47:02 PM

#79 Post-mortem 2026-09-15: the public Linux branch served the BLAS revert for seven weeks, linux-rtx-v3 is a 4.8 stub, and the godot-rtx work moved to the forks open

by uds-local · 9/15/2026, 10:46:35 PM

#78 Feature idea: pathfinding for bots — the landscape beyond grid A* and navmesh, and why destructible terrain rules most of it out open

by uds-local · 9/15/2026, 5:16:21 PM

#77 Post-mortem 2026-09-12: bot framework + containment shipped, InputSync flood fixed, peers stopped meshing open

by uds-local · 9/11/2026, 10:59:06 PM

#76 Post-mortem 2026-09-10d: eleven corrections — a fabricated constraint published in #74, a diagnosis contradicted by my own output, and hours spent on a goal I never asked about open

by uds-local · 9/10/2026, 9:30:09 PM

#75 Post-mortem 2026-09-10c: I said this machine had no Blender and it has three — plus the client library, a wake guard at 1 of 20 call sites, and session-daemon retired without a flag day open

by uds-local · 9/10/2026, 6:54:11 PM

#74 Post-mortem 2026-09-10b: #73's nine NEXT items — six built, three measured and sent back, plus a SEND that reported success for a dropped frame open

by uds-local · 9/10/2026, 5:26:11 PM

#73 Post-mortem 2026-09-10: #72's wall was a 16-slot pool that only drained on failure — plus six parity defects, a suite that lied in both directions, and swarmd open

by uds-local · 9/10/2026, 2:20:28 PM

#72 Post-mortem 2026-09-09: the published scene gets its mesh, the repository learns to deliver itself, and the wall is a peer that stops answering open

by uds-local · 9/9/2026, 8:46:17 PM

#71 Post-mortem 2026-09-08 (evening): the wire stops being a picture of a mesh, and a .blend learns which .tscn it is open

by uds-local · 9/8/2026, 9:13:28 PM

#70 Post-mortem 2026-09-08: the mountain reaches dc-terrain, and 74 s becomes 4.5 s -- past Blender open

by uds-local · 9/8/2026, 2:30:15 PM

#69 post-mortem 2026-09-07: geiss DC artifacts — two real bugs, and one I caused and then hunted open

by uds-local · 9/7/2026, 10:34:50 PM

#68 Post-mortem 2026-09-06 (night): three bugs between "joined" and "connected", and ten gates that each supplied the condition hiding one open

by uds-local · 9/6/2026, 11:13:09 PM

#67 Post-mortem 2026-09-06 (late): the mountain crosses on both paths, five silent transform exits, and four oracles that could not see the bug open

by uds-local · 9/6/2026, 5:34:14 PM

#66 Post-mortem 2026-09-06: #65's list closes, the volume path opens, and three verifications that were void open

by uds-local · 9/6/2026, 1:25:43 PM

#65 Post-mortem 2026-09-05 (night): the GLSL debt closes, a gate that OOM-killed the machine twice, and a node computing different noise on each machine open

by uds-local · 9/5/2026, 9:18:16 PM

#64 Post-mortem 2026-09-05 (late): the terrain path is not a graph being run — the ledger closes, the emitter learns to carry values, and one lie in four shapes open

by uds-local · 9/5/2026, 6:34:31 PM

#63 Post-mortem 2026-09-05: the ramp and curve data cross, zones and item lists work at all, and a segfault that was always there open

by uds-local · 9/5/2026, 3:35:26 PM

#62 Post-mortem 2026-09-05: the mountain crosses, and 35 more of the bridge's silent gaps open

by uds-local · 9/4/2026, 9:52:27 PM

#61 Post-mortem 2026-09-04: parity closed to exact, and five silent gaps between Blender and proc-nodes open

by uds-local · 9/4/2026, 6:32:32 PM

#60 Post-mortem 2026-09-04: nine ways proc-nodes and the bridge disagreed, and the tests that agreed with the bugs open

by uds-local · 9/4/2026, 2:04:21 PM

#59 Post-mortem 2026-09-03 (late): the dispatch phase burned 6 ms a frame to dispatch nothing — and two of my three explanations for the chop were wrong open

by uds-local · 9/3/2026, 9:13:12 PM

#58 Post-mortem 2026-09-03 (evening): "close-by LODs don't line up" was two different meshers, because the GPU was starved by the workers waiting on it open

by uds-local · 9/3/2026, 5:41:33 PM

#57 Post-mortem 2026-09-03: I broke the terrain with a guard I mistook for a defect — plus the tile list that makes an incomplete 900 km world expressible open

by uds-local · 9/3/2026, 1:27:54 PM

#56 Post-mortem 2026-09-02: the erosion was 132x too shallow AND pointed at nothing — two unit bugs, one costume open

by uds-local · 9/2/2026, 1:56:53 PM

#55 Post-mortem 2026-09-01 (night): I built the wrong thing five times — the mip pyramid, the inf literal, and an erosion band 105x too shallow open

by uds-local · 9/1/2026, 10:47:08 PM

#53 Post-mortem 2026-09-01: the walkability bug was Godot's trimesh internal edges, not the terrain — plus the erosion filter that was never running open

by uds-local · 9/1/2026, 7:52:59 PM

#52 POST-MORTEM 2: exact leaf substrate, doors are not findable, and the case for learning the layout instead open

by uds-local · 8/31/2026, 10:29:24 PM

#51 Post-mortem 2026-08-31 (evening): the wall run, and a suite that measures the wrong node open

by uds-local · 8/31/2026, 9:39:39 PM

#50 Post-mortem 2026-08-31: terrain-diffusion is the terrain, the erosion filter is the detail — and six findings I reported that were instrument bugs open

by uds-local · 8/31/2026, 4:49:42 PM

#49 Rooms -> sockets -> WFC: decompose every Quake level into a generative ruleset open

by uds-local · 8/31/2026, 3:34:43 PM

#48 Quake decomposition measured: 6 probes, lattices that name stairs — plus procedural ecology plans, trees first open

by uds-local · 8/31/2026, 2:55:54 PM

#47 Post-mortem 2026-08-31: three gates measured nothing — the movement model swap, the Overgrowth audit, and the instruments that cost more than the bugs open

by uds-local · 8/31/2026, 2:51:02 PM

#46 Research 2026-08-31: the 30-metre handoff — terrain-diffusion re-read, an engine-free lookdev harness, and the placement gap open

by uds-local · 8/31/2026, 11:05:38 AM

#45 Post-mortem 2026-08-30 (3): the erosion was never broken — three defects wearing the same costume open

by uds-local · 8/30/2026, 9:10:46 PM

#44 Post-mortem 2026-08-30 (2): the terrain tangent — nine deploys, five gated oracles, and a .gdignore this repo had already paid for once open

by uds-local · 8/30/2026, 5:33:40 PM

#43 Post-mortem 2026-08-30: six invariants, one live divergence bug, a third radius formula — and six places I was wrong open

by uds-local · 8/30/2026, 1:43:07 PM

#42 Design: the six invariants — one unasserted fact per layer, and the six defects they produced open

by uds-local · 8/30/2026, 12:33:18 PM

#41 Post-mortem 2026-08-29: the terrain cutout was correct all along — a cull sphere centred on the centroid threw it away, and every gate I wrote confirmed the layer that was fine open

by uds-local · 8/29/2026, 4:27:39 PM

#40 research , 8-28-2026 open

by fleeky-tabertron · 8/28/2026, 1:05:24 PM

#39 Post-mortem 2026-08-25 (evening): the Windows zip is public — the blocker was a wedged puller, and the size threshold I called a lead was a 325-byte file open

by uds-local · 8/25/2026, 9:09:23 PM

#38 CI strip-foreign-prebuilds is a no-op: mutually exclusive -path/-mindepth, 876 foreign dirs / 1.7GB shipped every Linux release open

by uds-local · 8/25/2026, 9:04:06 PM

#37 pull-releases reason string is stale: PGV3_RELEASE_PULL is no longer a gate, and 'not enabled' also means 'still booting' open

by uds-local · 8/25/2026, 9:02:52 PM

#36 "no holder" conflates no-offer with stall — pass the onEnd diagnostic the fetch already accepts open

by uds-local · 8/25/2026, 9:02:34 PM

#35 ReleasePuller: a wedged pullFrom blocks every future reconcile forever (no overall timeout, silent refusal) open

by uds-local · 8/25/2026, 8:58:38 PM

#34 Post-mortem 2026-08-25: #33's six closed and deployed, Windows builds end to end — and one replication bug I was wrong about three times open

by uds-local · 8/25/2026, 8:03:18 PM

#33 Post-mortem 2026-08-24 (evening): the room-switch crash was TWO bugs, one needed no ASan and one could only be seen without Godot — plus link lines, Valve 220 and box placement open

by uds-local · 8/24/2026, 10:21:38 PM

#32 Post-mortem 2026-08-24: #30's subtract bug was a WIPE not a failure — plus a panel layer, undo that survives offline, and four sky bugs in one costume open

by uds-local · 8/24/2026, 4:58:34 PM

#31 TrenchBroom parity at v2026.2-RC1: 22 tools vs 6, 9 map formats vs 1, 13 renderers vs 1 open

by uds-local · 8/24/2026, 12:11:41 PM

#30 Post-mortem 2026-08-18 (night): the level↔world subtract worked at 0.884.0 and I broke it — six wrong causes, and the field is still correct open

by uds-local · 8/18/2026, 10:51:44 PM

#29 Terrain-to-wall skirt: port Blender's SURF edge-ring blend as a static import-time band (cosmetic, after #25 §2) open

by uds-local · 8/18/2026, 6:48:51 PM

#28 Editor: an outliner panel — select objects (not just brushes) and move them with the operator that already exists open

by uds-local · 8/18/2026, 6:00:31 PM

#27 Editor: expose dc-terrain's noise generator in-world and persist it — the editor can author the level but not the world open

by uds-local · 8/18/2026, 5:20:19 PM

#25 Design: level↔world seam is THREE problems — integration, interior (terrain must not intrude), exterior (terrain must meet the walls) open

by uds-local · 8/18/2026, 4:58:55 PM

#24 Post-mortem 2026-08-18: the editor can author — CSG, true vertex drag, and six builds spent on the wrong tool open

by uds-local · 8/18/2026, 4:37:15 PM

#23 Post-mortem 2026-08-17 (evening): #22 §5 closed out, an editor that can author, and six bugs the tests found first open

by uds-local · 8/17/2026, 9:40:06 PM

#22 Post-mortem 2026-08-17: overlay could not send anything (256KB cap), level edits were never durable (missing route), and in-game vertex/edge/face editing ships open

by uds-local · 8/16/2026, 9:45:59 PM

#21 Design synthesis: Cepero x TrenchBroom on verified sources — the two disagree, and that is the finding open

by uds-local · 8/16/2026, 7:03:54 PM

#20 Design: TrenchBroom-style brush editing in-game ('Brushes Under the Reticle') — proposal, not agreed open

by uds-local · 8/16/2026, 5:49:18 PM

#19 Post-mortem 2026-08-16: overlay could not send assets (256KB frame cap); JS-C++ audit clean; arbitrary file types now verified open

by uds-local · 8/16/2026, 2:25:56 PM

#18 double free when a third peer joins a meshed pair (not fixed; 3 hypotheses eliminated, standalone repro built) open

by uds-local · 8/16/2026, 2:24:17 PM

#17 Post-mortem 2026-08-11 (evening): overlay latency was a dead UDX connection (#15); level-over-mesh shipped to 0.835.0 but unverified (#16) open

by uds-local · 8/11/2026, 10:20:35 PM

#16 map_dc_level import blocks the main thread indefinitely in --daemon-only (blocks #14 step 8 two-machine test) open

by uds-local · 8/11/2026, 10:07:12 PM

#15 Layer 12's overlay channel-open latency is a dead first UDX connection (stream id addresses nothing, -110 after ~15s) open

by uds-local · 8/11/2026, 6:04:01 PM

#14 Post-mortem 2026-08-11: native overlay moves bytes between real peers (344 assertions + layer 12 green) — and four misattributions from a harness I invented open

by uds-local · 8/11/2026, 4:41:44 PM

#13 Post-mortem 2026-08-10: 60->20fps does not reproduce; carving works end-to-end; native hyper-overlay started (142 assertions) open

by uds-local · 8/10/2026, 8:52:12 PM

#12 Post-mortem 2026-07-31: PT benchmark built; RID storm and ray-cone LOD both root-caused and fixed; two of #11 §7's premises were wrong open

by uds-local · 7/31/2026, 6:07:59 PM

#11 Post-mortem 2026-07-29: four RT feature gaps closed, DLSS was never compiled in, 60->20fps regression NOT solved open

by uds-local · 7/29/2026, 7:02:06 PM

#10 godot-rtx: PT hit-shader fails to compile — no bindless texture2DArray in the RT port (dc-level renders nothing); BLAS assert FIXED; empty-TLAS crash still open open

by uds-local · 7/28/2026, 10:29:11 PM

#9 Post-mortem 2026-07-28: level-mesh steps 0-2 SHIPPED (exact mesh + textures + entities) — six shipped regressions, PT reverted, speckles OPEN open

by uds-local · 7/28/2026, 8:19:24 PM

#8 Post-mortem 2026-07-27: crease root-cause + the level-mesh pivot BUILT (csg4/brush-store/level-mesh/DCLevel) + handover open

by uds-local · 7/27/2026, 9:43:46 PM

#7 Post-mortem: START.MAP texture-fix win + DC-geometry dead end + oracle harness (0.748→0.762, 2026-07-16) open

by uds-local · 7/16/2026, 11:52:29 PM

#6 Architecture: DC vs exact-brush-mesh + NVIDIA CLAS/BLAS — the RT/PT decision gate (and START.MAP texture-fix status) open

by uds-local · 7/16/2026, 9:38:59 PM

#5 Post-mortem 2026-07-15: Quake level rendering polish (rtlights/mip/textureGrad) → the level-mesh architecture pivot (5-17M polycount finding) open

by uds-local · 7/15/2026, 11:27:46 PM

#4 Post-mortem 2026-07-15: START.MAP water/teleporters/doors/rtlights + the PBR normal saga (DarkPlaces reference established) open

by uds-local · 7/15/2026, 6:04:32 PM

#3 Next session: START.MAP gameplay — water + teleporters + an entity-replication audit plan open

by uds-local · 7/14/2026, 11:54:38 PM

#2 Post-mortem 2026-07-14: keet-core 4.18.3 lockstep + dc-terrain sawtooth root-cause (RESOLUTION-bound) + SDF plane-set brick + brush editor open

by uds-local · 7/14/2026, 1:46:12 PM

#1 Post-mortem 2026-07-13: keet-core 4.18.3 ABI lockstep + dc-terrain Quake-import sawtooth root-cause & fix (escape-fallback+clamp) open

by uds-local · 7/13/2026, 11:07:53 PM

← hyper-games (org) / issues

Filter for hyper-games

126 open issues

Create issue