manifest(windows-x64): 73d20a40b verification closed — oracle 10/10 zero VUIDs, DIC 22/0, cull_pool 4524/4524, headless smoke clean

manifest.json

@@ -26,10 +26,10 @@
       "sha256_console": "265f8133999528aa06f919870f58f7c04e55a448da0c6221e093ce76710e5bc2",
       "target": "editor",
       "built_on": "2026-10-19",
-      "built_by": "Ozan/Forge via ZCode (fast publish on owner request)",
+      "built_by": "Ozan/Forge via ZCode (fast publish, oracle + smoke verified post-push)",
       "toolchain": "MSVC 14.3 (VS2022 Build Tools), Windows SDK 10.0.26100, SCons 4.10.1, .NET SDK 10.0.103",
       "scons_flags": "platform=windows target=editor module_mono_enabled=yes -j32",
-      "note": "Windows carries 73d20a40b \u2014 SAME hash as macos-arm64 (built at 73d20a40b8): the d49400463 pair (parallel replay+recording with the VUID-00050 pool reset; see its note in git history) PLUS 219b914c18 fix(rd) \u2014 parallel draw recording sub-recordings nest: a recording worker that waits inside its slice (get_pipeline waiting on an ubershader compile) could run another slice's task on the same thread and its end cleared the thread-local instruction target, so the outer slice appended into the MAIN draw list concurrently with the render thread (realloc under a live instruction) \u2014 Windows FATAL p_idx >= _len in add_draw_list_bind_vertex_buffers on the Ultra bench (#250); begin now pushes the current target and end restores it; oracle Vulkan/MoltenVK 10/10, Metal 5/5. PLUS 73d20a40b8 fix(texture_rd) \u2014 the RS texture is allocated before the render-thread bind is queued, so main-thread get_rid() no longer races the render thread's write of texture_rid (a material could keep the magenta placeholder for the session; Ariki's FFT ocean drew milky/opaque on Mac and Windows); Texture2DRD, TextureLayeredRD, Texture3DRD. GodotSharp regenerated from this binary \u2014 Api AND Tools. Fast publish per owner request: oracle + headless smoke on this exact hash still owed (the recording-nest fix is in the engine's own oracle coverage; its run.sh now drops Metal shader validation per the commit). Owed on Windows (carried): tests=yes unit suites, crash checklist (8 steps), template_release still 1e7be28b. D3D12 Agility DLLs not shipped (Vulkan game)."
+      "note": "Windows carries 73d20a40b \u2014 SAME hash as macos-arm64 (built at 73d20a40b8): the d49400463 pair (parallel replay+recording with the VUID-00050 pool reset; see its note in git history) PLUS 219b914c18 fix(rd) \u2014 parallel draw recording sub-recordings nest: a recording worker that waits inside its slice (get_pipeline waiting on an ubershader compile) could run another slice's task on the same thread and its end cleared the thread-local instruction target, so the outer slice appended into the MAIN draw list concurrently with the render thread (realloc under a live instruction) \u2014 Windows FATAL p_idx >= _len in add_draw_list_bind_vertex_buffers on the Ultra bench (#250); begin now pushes the current target and end restores it; oracle Vulkan/MoltenVK 10/10, Metal 5/5. PLUS 73d20a40b8 fix(texture_rd) \u2014 the RS texture is allocated before the render-thread bind is queued, so main-thread get_rid() no longer races the render thread's write of texture_rid (a material could keep the magenta placeholder for the session; Ariki's FFT ocean drew milky/opaque on Mac and Windows); Texture2DRD, TextureLayeredRD, Texture3DRD. GodotSharp regenerated from this binary \u2014 Api AND Tools. VERIFIED POST-PUSH on the published binary (RTX 5080): parallel_draw_replay oracle vulkan --gpu-validation 10/10 exit 0 ZERO VUID lines; draw_indirect_count d3d12 + debug layer 22/0 + expected SKIP base-instance line; multimesh_cull_pool d3d12 + debug layer 4524/4524, 0 D3D12 errors; headless game boot clean (0 binding errors). Owed on Windows (carried): tests=yes unit suites, crash checklist (8 steps), template_release still 1e7be28b. D3D12 Agility DLLs not shipped (Vulkan game)."
     }
   }
 }