The oracle showed Cycles treats a huge exponent as integral for a negative base, so the speculative
b >= 2^31 rule is gone. CultMath.Cycles.glsl is stale at this commit (the HLSL changed), so
CommittedGlslEqualsLowering fails here; the regenerated file follows. Builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The film zeroes an infinite result, which had hidden both the power quirk and Compare's epsilon floor
from the fixtures (two hand mutants of the HLSL survived). The fixtures are regenerated in a later commit.
Builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A mutant of CultMath.Cycles.hlsl that changes a Cycles result the fixtures reach now fails a test on its own;
before, only the random-input mirror comparison reached the HLSL. Builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Written by tools/cycles-oracle/oracle.py and GlslLowering on Yggdrasil; the tool's second run was
byte-identical. CultMath.Tests passes 379 of 379 at this commit's tree. Builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The oracle's first fixtures showed Cycles' AVX2 kernel fuses multiply-adds and its film zeroes
non-finite results; the tool now exits 4 off the SSE4.2 kernel (CYCLES_CPU_NO_AVX2=1) and the
test skips cases Cycles cannot report. The Unity copy of CultMath.hlsl carries the new include.
Builds. Fixtures and CultMath.Cycles.glsl are still to come.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Builds. The oracle fixtures and CultMath.Cycles.glsl are still to come, so the oracle and golden tests fail here.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
oracle.py bakes each Cycles node under Blender 5.2.2 LTS (build d13f752e3b9c, refusing any other)
into float images; CyclesOracleTests holds math.cycles_* to the fixtures it writes. No fixtures
exist at this commit, so the tests fail here; builds.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Ports of Blender v5.2.2's intern/cycles kernel functions, one function per node
operation, with the HLSL mirror, the Unity copy, the lowering table entry for a
separate Apache-2.0 GLSL file, the golden-fixture families, the package licence
expressions and the notices section.
Builds. Does not pass yet: CultMath.Cycles.glsl and the regenerated golden
fixture arrive in a later commit, so CommittedGlslEqualsLowering and
GoldenFixtureMatchesCSharp fail at this commit.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
leaf-vectors-cs Soul-held at pass 3 (cultlib-gaps:verdict:cut-leaf-vectors-cs.s3); py-typed-leaves Soul-held at pass 2 (cultlib-gaps:verdict:cut-py-typed-leaves.s2).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Soul-held at pass 3 (cultlib-gaps:verdict:cut-union-bounds-ts.s3). Lands follow-up ts-codec-pin-gaps.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The wide-count test announced map32 as 32770 pairs, which fits in 16 bits, so a
count read as its low 16 bits survived. map32 now announces 65537 pairs and
array32 65540 children; both low halves close the container early.
Builds: yes.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Fixes the previous commit (05f9937b, mislabelled wip): its test file did not compile (TS2322) and the pack probe threw on a lazy codec's getters. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
leaves.py (string, boolean, bytes, int32, int64, float32, float64, any, list, array, nullable, lazy) held to the
numeric-leaves vectors. Removed: DatabaseEntryField.many, callable codecs and _resolve, the System.Object literal,
the present-nil default, the union's nil pass-through, interop.py's hand-written catalog. A float32 keeps its width
through _pack_tree. Nil is refused unless wrapped in nullable; a default fills only a slot past the payload's end.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Finding pack-denial-by-name: the package-internal readers (readWire, resolveDocumentFormatter, @msgpack decode) are loaded from the packed dist by path and searched by identity through every packed entry, including default and namespace members; every other exported function is shown a MessagePack map and must not return it decoded.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Findings custom-formatter-bytes-walked (ruling depth-walk-covers-library-msgpack: the owner skips the walk when the definition carries a formatter; the grey 0x91 image round-trips, plain and codec stay bounded; README says so) and wide-length-unpinned (bin32/str32/ext32 of 65539 bytes and array32/map32 of more than 65535 children before a chain at and past the limit). The same tests pin the inspector refusing to recover a stamp from a past-limit payload. Not yet built.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Finding inspector-recover-unwalked: one decodeBoundedPayload (walk, then decode) serves both the stamp recovery and the preview, so a payload past the limit is never decoded by either.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
leaf-record.msgpack is unchanged. The any leaf now has a ext refusal (d40100) and the record leaves read a complete second slot.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Kills the survivors of the first mutation sweep (fixext2 skip, 0xc1, the fixarray/fixmap/fixstr upper bounds, map16 and map32 child counts).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The C# cases and the TS assertions are here; the vector file is written from the reference in the next commit, so the
slot-nil tests of both runtimes fail at this commit.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
checkPayloadDepth is an iterative header walk run by decodeDocumentPayload before the formatter and by
encodeDocumentPayload on the formatter's bytes; the two plain-decode inspection doors call it first. The plain
formatter's encoder depth is set above the limit so the walk decides. Tests: the C# bounds vectors, the walk's
header and scalar rules, the owner, the inspector, the legacy version inference, and the mesh network doors;
test:pack:ts denies decoder-shaped exports by enumerating the packed entries.
Builds and passes untested at this commit (verification runs on Yggdrasil next).
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
