Repository navigation
Description
Summary
2026-11-13: PRs #1961, #1962, #1965, #1967, #1968, #1969, #1970, #1971 and #1972 were closed and combined into #1973, which is built and tested with the combined EbsdLib branch BlueQuartzSoftware/EbsdLib#86. Items that one of these pull requests fixes now name it and #1973. No item is ticked until #1973 is merged.
Some filters do not handle degenerate input: an empty geometry or DataArray, a selection that matches nothing, a zero-length vector, a Feature with no cells, a cosine that rounds just past ±1, or an index computed from bad data. The filter then crashes, or it writes NaN, inf or a plausible-looking wrong value and reports success. As of develop 419f9db (2026-10-21), this class has 44 open findings in 39 filters: 11 CRASH, 23 WRONG_DATA, 10 MINOR.
Mechanism
The algorithms assume valid, non-empty input. Preflight checks that the arrays exist, but not that they have elements. Shared helpers divide or index with no guard. This class merges five sub-patterns:
- Empty input. The geometry or DataArray has 0 elements, or nothing is selected. The code dereferences
std::max_elementof an empty range (that isend()), reads element 0 of an empty vector, divides by a count of 0, or keeps FLT_MAX/lowest extents. Helpers such asFindEulerCharacteristicValues(GeometryHelpers.cpp) andQuickSortVertices(VertexUtilities.cpp) have no empty check. - Zero-length vector. A direction, rotation axis, Rodrigues axis or face normal is divided by its length with no zero check. Several parameter defaults are {0,0,0}, so the default arguments give NaN.
- Empty or reserved Feature. A Feature with no cells or faces, or Feature 0, is not handled. Per-Feature outputs get 0/0 NaN, an unscaled count, or a phantom last row.
- acos and pole coordinates.
acosgets a cosine that rounds to 1+ulp and returns NaN. The square-grid coordinate x/fabs(x) is NaN at x = y = 0. The NaN then goes to a wrong bin, rejects a valid match, or goes into the output. - Data-derived index. An index computed from data (an unclamped hex column) is used for a read with no range check.
The shared Vec3::normalize divides by the magnitude with no check (Array.hpp#L795-L799):
Vec3 normalize() const
{
T mag = magnitude();
return Vec3((*this)[0] / mag, (*this)[1] / mag, (*this)[2] / mag);
}ComputeBoundaryStrengthsFilter calls it on the Loading Direction (XYZ) parameter, whose default is {0.0, 0.0, 0.0}. All three components become NaN, and the filter reports success. Eigen normalize() behaves differently: it leaves a zero vector unchanged, which is also wrong for an axis (ComputeMisorientationsFilter#2).
No filter of this class is fixed on develop or in an open PR yet.
Rule
These parts follow the project rules for all classes:
- A check that needs only the parameters or the geometry size goes in preflight. Examples: a zero-length axis or direction parameter, a geometry with 0 vertices, faces or edges, a DataArray with 0 tuples.
- A check that needs array values goes at the start of execute (for example a mask that is all false, or no face in the selected phase).
- Remove execute-time checks that a new preflight check makes unreachable.
Decision needed:
- Empty input. Option A: return a preflight error for a geometry or DataArray with 0 elements. Option B: succeed with a defined empty output and a warning. Option B keeps pipelines that sometimes produce an empty mesh running. Option A is simpler. A mix is also possible (error for filters whose output has no meaning when empty).
- Zero-length parameter vectors. Proposed: preflight error for a zero-length or non-finite axis or direction, and change each {0,0,0} default to a valid axis (for example {0,0,1}). Check that a new default does not change existing tests.
- Zero-length data-derived vectors (a Rodrigues vector (0,0,0), a zero-area face normal). Proposed: write one defined value. For an identity Rodrigues vector, write (0,0,1,0). For a degenerate face normal, write (0,0,0).
- Empty Features. Choose one sentinel (0 or NaN) for per-Feature outputs of a Feature with no cells, and reset Feature 0 outputs.
- acos. Proposed: add one shared clamped acos helper (for example in
GeometryMath.hpp) and one shared square-grid coordinate function that maps (0,0) explicitly. No such helper exists at 419f9db.
How to fix
- Empty input: add the check in preflight when the size is known there (
getNumberOfVertices(),getNumberOfFaces(),getNumberOfEdges(),getNumberOfTuples()). Guard everystd::max_element,front(),back(),size() - 1and division by a total count in the algorithm and in the shared helpers (FindEulerCharacteristicValues,QuickSortVertices,PrintDataSetsToSingleFile). - Zero-length parameter vectors: return a preflight error when the length is 0 or not finite. Change the {0,0,0} defaults (
ComputeBoundaryStrengthsFilter,WriteGBCDGMTFileFilter).ComputeGBCDMetricBasedFilteralready has an axis check (-7231); use it as a model. - Zero-length data-derived vectors: check the length before the division and write the defined value. Change both the in-core and the chunked paths.
- Empty Features: count the elements of each Feature and write the sentinel before the computation. Do not compute 0/0. Move existing zero guards before the first division (
ComputeShapes.cppline 594). - acos and pole coordinates: clamp each cosine to [-1, 1] with
std::clampbeforeacos. Handle x = y = 0 in the square-grid coordinate function. - Data-derived index: clamp or range-check each computed index against the size of the target container before the read.
- Filters with more than one algorithm (Direct and Scanline): fix every algorithm.
ComputeGBCDPoleFigureScanline.cppandComputeSurfaceAreaToVolumeScanline.cpphave the same code as their Direct versions. - Error codes: each filter uses one range more negative than -10000 (for example -47500..-47599). Check that a new code is unique with
git grep -nE -- "-NNNN([^0-9]|$)", and also in open PR refs.
How to test
- Write the test first. It must fail on develop.
- Empty input: build a geometry with 0 vertices, faces or edges, or a DataArray with 0 tuples. Assert the preflight error code, or assert the defined empty output and the warning code. On develop, these tests crash or return success.
- Zero-length vector parameter: set the axis to {0,0,0} and assert the preflight error code.
- Zero-length data-derived vector, empty Feature, acos: build a small synthetic input (one identity Rodrigues tuple, one Feature id with no cells, an exact bicrystal match, a face normal of (0,0,1)). Assert the exact output values. Get the expected values from an independent calculation (by hand, or from a closed-form result), not from the output of the code.
- Assert that no output value is NaN or inf, unless NaN is the chosen sentinel.
- If the filter has more than one algorithm, run each test for each algorithm with the existing algorithm-selection test scope (
ForceInCoreAlgorithmGuard/ForceOocAlgorithmGuardinAlgorithmDispatch.hpp).
Conventions
- One PR for the whole class. One commit per filter. Commit prefix
BUG:. Add aSigned-off-byline. Do not add AI attribution. - Every fix starts with a unit test that fails on develop. If you cannot reproduce a finding, write that and the reason in the PR and on this issue. Do not force a fix.
- Value checks need an independent expected result (oracle), not the output of the code under test.
- Run ctest serially. Never use
-j, because the tests depend on each other. - Format with clang-format 19.1.1 and the repository
.clang-format. - Do not add per-filter documentation sentences about input requirements. The generated parameter table documents them.
- If a fix touches a filter that has a V&V report (
src/Plugins/<Plugin>/vv/<Filter>.md), update every file of that filter's V&V package (report, deviations, provenance, test tables). V&V documents reference DREAM3D 6.5.171 only. - Rebase on develop. Never merge develop into the branch.
Checklist
-
CRASH · ApproximatePointCloudHullFilter · ApproximatePointCloudHull.cpp:96 — Preflight accepts a Vertex Geometry with 0 vertices; the extents stay at FLT_MAX/lowest,
static_cast<int64>(floor(inf))is undefined, andvertsInVoxelsis allocated with a garbage size.ApproximatePointCloudHullFilter#2 -
CRASH · ComputeFeatureClusteringFilter · ComputeFeatureClustering.cpp:271 — With exactly two Features in the phase, max == min so
stepSizeis 0, the bin index is NaN cast to int, andrdfCacheis written out of bounds; 0 or 1 Feature writes FLT_MAX and 0. The derived floating-point bin counts atComputeFeatureClustering.cpp:46and:329are converted to integers unchecked, so a very small but nonzero step can exceed the integer range.ComputeFeatureClusteringFilter#2 -
CRASH · ComputeShapesTriangleGeomFilter · GeometryHelpers.cpp:40 —
FindEulerCharacteristicValuesin GeometryHelpers.cpp dereferencesstd::max_elementof an empty Face Labels DataArray (0 faces), which readsend().ComputeShapesTriangleGeomFilter#7 -
CRASH · ComputeTriangleGeomVolumesFilter · ComputeTriangleGeomVolumes.cpp:41 —
*std::max_elementis read on an empty Face Labels DataArray (0 faces); all labels -1 resize the Feature Attribute Matrix to 0 tuples, and labels <= -2 give a size of SIZE_MAX.ComputeTriangleGeomVolumesFilter#3 -
CRASH · ConvertHexGridToSquareGridFilter · ConvertHexGridToSquareGrid.cpp:187 — With an X Spacing finer than the hex XSTEP, the offset-row column index is not clamped, so edge points read the next row and the last row reads past the arrays. (related: BUG: Hex-grid conversion selects the wrong coincident source node #1816, partial overlap)
ConvertHexGridToSquareGridFilter#1 -
CRASH · CreatePythonSkeletonFilter · PythonPluginTemplateFile.hpp:98 — With Use Existing Plugin and a 0-byte
__init__.pyorPlugin.py,splitreturns an empty vector andlines.back()is called on it (lines 98 and 137).CreatePythonSkeletonFilter#1 -
CRASH · HierarchicalSmoothFilter · HierarchicalSmooth.cpp:612 — A Triangle Geometry with 0 faces reaches
vs.mesh(0, 0),(0, 1)and(0, 2)on an empty Eigen matrix, which reads invalid memory in Release builds.HierarchicalSmoothFilter#4 -
CRASH · IdentifyDuplicateVerticesFilter · VertexUtilities.cpp:76 — With 0 vertices,
QuickSortVertices(sorted, 0, sorted.size() - 1, ...)gets end = SIZE_MAX and readssorted[0]of an empty vector (VertexUtilities.cpp).IdentifyDuplicateVerticesFilter#2 -
CRASH · PointSampleEdgeGeometryFilter · PointSampleEdgeGeometry.cpp:217 — With an Edge Geometry that has 0 edges,
*std::max_elementis read on an empty store, which dereferencesend().PointSampleEdgeGeometryFilter#5 -
CRASH · SplitDataArrayByTupleFilter · SplitDataArrayByTupleFilter.cpp:460 — An input DataArray with tuple shape {0} leaves
tupleShapesempty, and preflight readstupleShapes[0](with Output Container = New Attribute Matrix).SplitDataArrayByTupleFilter#2 -
CRASH · WriteFeatureDataCSVFilter · OStreamUtilities.cpp:649 — A Feature Attribute Matrix that holds only NeighborLists (or nothing) gives an empty path list, and
PrintDataSetsToSingleFilereadsobjectPaths[0](OStreamUtilities.cpp).WriteFeatureDataCSVFilter#1 -
WRONG_DATA · AlignGeometriesFilter · AlignGeometries.cpp:209 — A node geometry with 0 vertices gives a 0/0 NaN centroid (or an origin of FLT_MAX), and the moving geometry is translated by it with no error.
AlignGeometriesFilter#4 -
WRONG_DATA · AlignSectionsFeatureCentroidFilter · AlignSectionsFeatureCentroid.cpp:111 — A Z-slice with no mask cells divides by a count of 0; the NaN centroid is cast to int64 (undefined) and enters the cumulative shifts, and execute succeeds.
AlignSectionsFeatureCentroidFilter#3 -
WRONG_DATA · ApplyTransformationToGeometryFilter · ImageRotationUtilities.cpp:184 — Rotation Axis-Angle = (0,0,0,90) is normalized by 0 in ImageRotationUtilities.cpp, so all node vertices become NaN; for an Image Geometry, NaN is cast to— moved to BUG: C14: Parameter values not validated in preflight in 65 filters #1873.usizedimensions in preflight.ApplyTransformationToGeometryFilter#8 -
WRONG_DATA · ComputeBoundaryStrengthsFilter · ComputeBoundaryStrengths.cpp:52 — The default Loading Direction (XYZ) is {0,0,0}, and— moved to BUG: C14: Parameter values not validated in preflight in 65 filters #1873.LD.normalize()divides by 0, so the m-3m outputs are NaN-based and the filter reports success.ComputeBoundaryStrengthsFilter#2 -
WRONG_DATA · ComputeGBCDFilter · ComputeGBCD.cpp:253 —
getSquareCoordcomputes x/fabs(x) at x = y = 0 (normal (0,0,±1), common for voxel meshes); the NaN passes the limit checks and the area goes into a corner bin instead of the center bin.ComputeGBCDFilter#2 -
WRONG_DATA · ComputeGBCDMetricBasedFilter · ComputeGBCDMetricBased.cpp:187 —
acosarguments at lines 187 and 286-289 are not clamped; a boundary that matches the Fixed Misorientation exactly can give 1+eps, so acos is NaN and the match is dropped.ComputeGBCDMetricBasedFilter#6 -
WRONG_DATA · ComputeMisorientationsFilter · ComputeMisorientations.cpp:246 — With Reference Axis-Angle = (0,0,0,30), Eigen— moved to BUG: C14: Parameter values not validated in preflight in 65 filters #1873.normalize()leaves (0,0,0), the reference quaternion is not unit, and the user's angle is lost with no error.ComputeMisorientationsFilter#2 -
WRONG_DATA · ComputeMomentInvariants2DFilter · ComputeMomentInvariants2D.cpp:72 — A Feature that covers one cell gives dim = 1, so
xx / maxCoeffandfNorm = (dim - 1) / 2divide by 0, and Omega 1 and Omega 2 are NaN.ComputeMomentInvariants2DFilter#4 -
WRONG_DATA · ComputeShapesFilter · ComputeShapes.cpp:583 — Axis lengths are computed as (A^4)/(B*C) and written before the A/B/C == 0 guard at line 594, so a Feature with no cells gets NaN Axis Lengths.
ComputeShapesFilter#7 -
WRONG_DATA · ComputeShapesFilter · ComputeShapes.cpp:321 — The copy loop writes the raw cell count for every Feature, but the scaling loop starts at Feature 1, so Volumes[0] keeps an unscaled count.
ComputeShapesFilter#8 -
WRONG_DATA · ComputeTwinBoundariesFilter · ComputeTwinBoundaries.cpp:131 —
std::acos(GeometryMath::CosThetaBetweenVectors(...))is not clamped; a Face Normal exactly parallel to <111> can give 1+ulp, so a coherent twin is rejected.ComputeTwinBoundariesFilter#4— PR BUG: Clamp cosines before acos in Compute Twin Boundaries #1972 (EbsdLib side: BUG: Clamp the quaternion scalar part before acos() in conversions EbsdLib#80), combined into BUG: Combine the EBSD fixes that need the next EbsdLib release #1973 and REL: EbsdLib v3.1.4 Official Release EbsdLib#86 -
WRONG_DATA · ComputeVertexToTriangleDistancesFilter · ComputeVertexToTriangleDistances.cpp:442 — With a Target Triangle Geometry that has 0 faces, distances start at FLT_MAX and
sqrt(FLT_MAX)= 1.84e19 is written as if it were a real distance.ComputeVertexToTriangleDistancesFilter#5 -
WRONG_DATA · FeatureFaceCurvatureFilter · CalculateTriangleGroupCurvatures.cpp:278 — A seed triangle with a (0,0,0) Face Normal or a duplicate centroid makes the rotation NaN; the loop breaks with the patch in the global frame, and the fit still writes finite curvatures. (related: BUG: Curvature fitting returns NaN directions from a singular frame #1819)
FeatureFaceCurvatureFilter#8 -
WRONG_DATA · LabelTriangleGeometryFilter · LabelTriangleGeometry.cpp:95 —
triangleCountsstarts as {0,0} and grows by one per region, so the Feature Attribute Matrix is resized to R+2 tuples and gets a phantom last Feature with NumTriangles 0.LabelTriangleGeometryFilter#1 -
WRONG_DATA · LaplacianSmoothingFilter · LaplacianSmoothing.cpp:112 — A vertex that no face or edge uses has
numConnections= 0, sodeltaArray[in0] / numConnections[i]is NaN and it is written even when lambda is 0 (the Taubin pass at lines 157-160 does the same).LaplacianSmoothingFilter#2 -
WRONG_DATA · MapPointCloudToRegularGridFilter · MapPointCloudToRegularGridFilter.cpp:125 — With Sampling Grid Type = Manual and a Mask that is all false (or 0 vertices), the extents are not updated, so spacing and origin become inf with no error.
MapPointCloudToRegularGridFilter#2 -
WRONG_DATA · ReadGrainMapper3DFilter · ReadGrainMapper3D.cpp:362 — With Create Compatible Orientation Data on, a (0,0,0) Rodrigues tuple (identity orientation) is divided by its length 0, and NaN axis components are written.
ReadGrainMapper3DFilter#4 -
WRONG_DATA · RodriguesConvertorFilter · RodriguesConvertor.cpp:52 — The identity Rodrigues vector (0,0,0) is divided by its length 0, so NaN axis components are written with no error or warning.
RodriguesConvertorFilter#1 -
WRONG_DATA · TriangleDihedralAngleFilter · TriangleDihedralAngle.cpp:66 —
acosarguments are not clamped; for a nearly collinear face the ratio can round to 1+ulp, and the face gets NaN instead of about 0.TriangleDihedralAngleFilter#2 -
WRONG_DATA · WriteFeatureDataCSVFilter · OStreamUtilities.cpp:683 — With Write Number of Features Line on and a Feature Attribute Matrix with 0 tuples,
featureCount--on ausizewrites 18446744073709551615 as the first line.WriteFeatureDataCSVFilter#2 -
WRONG_DATA · WriteGBCDGMTFileFilter · WriteGBCDGMTFile.cpp:131 — The default Misorientation Axis-Angle is {0,0,0,0}; the zero axis is normalized to NaN, no bin matches, and every point is written as 0/0 = NaN with no error.
WriteGBCDGMTFileFilter#1— Axis validation is tracked in BUG: C14: Parameter values not validated in preflight in 65 filters #1873. The zero-default replacement remains here pending its scope decision. -
WRONG_DATA · WriteGBCDGMTFileFilter · WriteGBCDGMTFile.cpp:294 —
GetSquareCoordgives NaN for (0,0,1); with Triclinic symmetry no bin matches on the phi = 0 row, sosum / countwrites 0/0 = NaN.WriteGBCDGMTFileFilter#6 -
WRONG_DATA · WriteStlFileFilter · WriteStlFile.cpp:246 —
vecA.cross(vecB).normalize()divides by 0 for a face with collinear or repeated vertices, so NaN normal components are written to the STL file (lines 246 and 557).WriteStlFileFilter#5 -
MINOR · ComputeGBCDMetricBasedFilter · ComputeGBCDMetricBased.cpp:298 — When no face has both Face Labels in Features of the Phase of Interest,
m_TotalFaceAreaandm_NumDistinctGBsare 0, and NaN is written to the .dat files with no warning.ComputeGBCDMetricBasedFilter#5 -
MINOR · ComputeGBCDPoleFigureFilter · ComputeGBCDPoleFigureDirect.cpp:74 — Misorientation Angle-Axis = {60,0,0,0} normalizes a zero axis, so the matrix is NaN and the MRD output is all zero with success (the Scanline path has the same code at line 75).— moved to BUG: C14: Parameter values not validated in preflight in 65 filters #1873.ComputeGBCDPoleFigureFilter#5 -
MINOR · ComputeGBPDMetricBasedFilter · ComputeGBPDMetricBased.cpp:254 — When no face has both Face Labels in Features of the Phase of Interest, the code divides by a total area and boundary count of 0, and NaN files are written with no warning.
ComputeGBPDMetricBasedFilter#4 -
MINOR · ComputeShapesTriangleGeomFilter · ComputeShapesTriangleGeom.cpp:365 — A Feature with no faces has Vol = 0, so
Cacc / Volis NaN and the Axis Euler Angles are NaN, while Axis Lengths and Aspect Ratios use the -1 sentinel.ComputeShapesTriangleGeomFilter#6 -
MINOR · ComputeSurfaceAreaToVolumeFilter · ComputeSurfaceAreaToVolumeDirect.cpp:144 — A Feature with 0 cells gives a 0/0 NaN ratio, and a Feature with surface area 0 gives Sphericity = inf (the Scanline path has the same divisions).
ComputeSurfaceAreaToVolumeFilter#3 -
MINOR · ComputeVolumeFractionsFilter · ComputeVolumeFractionsFilter.cpp:146 — A Cell Phases DataArray with 0 tuples gives
count / 0= NaN volume fractions.ComputeVolumeFractionsFilter#2 -
MINOR · PointSampleTriangleGeometryFilter · PointSampleTriangleGeometry.cpp:84 — 0 faces, or Face Areas that are all 0, negative or NaN, reach
std::discrete_distributionwith no check; 0 faces ends instd::out_of_range, which the execute backstop reports as error -2.PointSampleTriangleGeometryFilter#6 -
MINOR · RandomizeFeatureIdsFilter · RandomizeFeatureIdsFilter.cpp:132 — A Feature Ids DataArray with 0 tuples leaves the maximum at INT32_MIN, and error -82640 reports a maximum Feature Id of -2147483648.
RandomizeFeatureIdsFilter#3 -
MINOR · RobustAutomaticThresholdFilter · RobustAutomaticThreshold.cpp:75 — A Gradient Magnitude Data DataArray that is all 0 gives a 0/0 NaN threshold, so the output mask is all false with no message.
RobustAutomaticThresholdFilter#2 -
MINOR · SharedFeatureFaceFilter · SharedFeatureFace.cpp:80 — With Randomize Face IDs on and 0 faces,
uniform_int_distribution(1, 0)is created, which violates its precondition; the shuffle loop does not run.SharedFeatureFaceFilter#4 -
WRONG_DATA · RectGridGeom · RectGridGeom.cpp:127 — NaN bounds pass the increasing-order check and reach nonfinite cell-size calculations. (origin: sweep)
-
WRONG_DATA · GradientMagnitudeRecursiveGaussianImageFilter · GradientMagnitudeRecursiveGaussianImageFilter.cpp:98 — Infinite existing image spacing passes the shared guard and produces invalid coefficients. (origin: sweep)
-
WRONG_DATA · LaplacianRecursiveGaussianImageFilter · LaplacianRecursiveGaussianImageFilter.cpp:98 — Infinite existing image spacing passes the shared guard and produces invalid coefficients. (origin: sweep)
-
WRONG_DATA · SmoothingRecursiveGaussianImageFilter · SmoothingRecursiveGaussianImageFilter.cpp:100 — Infinite existing image spacing passes the shared guard. User Sigma validation is separate. (origin: sweep)
-
WRONG_DATA · GradientMagnitudeImageFilter · GradientMagnitudeImageFilter.cpp:95 — Infinite active image spacing passes validation and suppresses nonzero derivatives. (origin: sweep)
-
CRASH · MinMaxCurvatureFlowImageFilter · MinMaxCurvatureFlowImageFilter.cpp:76 — Invalid existing image spacing can produce nonfinite sample offsets before integer conversion. (origin: sweep)
-
WRONG_DATA · GradientAnisotropicDiffusionImageFilter · GradientAnisotropicDiffusionImageFilter.cpp:113 — Invalid active image spacing contaminates gradients and updates. (origin: sweep)
-
WRONG_DATA · CurvatureAnisotropicDiffusionImageFilter · CurvatureAnisotropicDiffusionImageFilter.cpp:113 — Invalid active image spacing contaminates gradients and updates. (origin: sweep)
-
WRONG_DATA · DiscreteGaussianImageFilter · DiscreteGaussianImageFilter.cpp:95 — Invalid active image spacing produces invalid effective variance. Large valid variance remains a separate numerical-stability defect. (origin: sweep)
-
WRONG_DATA · ApproximateSignedDistanceMapImageFilter · ImageProcessingFilterUtilities.hpp:3755 — Existing image spacing reaches distance interpolation without its required domain check. (origin: sweep)
-
WRONG_DATA · DanielssonDistanceMapImageFilter · DanielssonDistanceMapImageFilter.cpp:82 — Nonfinite active image spacing contaminates distance calculations. (origin: sweep)
-
WRONG_DATA · SignedDanielssonDistanceMapImageFilter · SignedDanielssonDistanceMapImageFilter.cpp:79 — Nonfinite active image spacing contaminates the shared distance calculation. (origin: sweep)
-
WRONG_DATA · SignedMaurerDistanceMapImageFilter · SignedMaurerDistanceMapImageFilter.cpp:145 — Nonfinite active image spacing contaminates physical-coordinate distance calculations. (origin: sweep)
-
WRONG_DATA · ComputeBoundingBoxStatsFilter · ComputeBoundingBoxStatsFilter.cpp:262 — Existing negative Image Geometry spacing passes the zero-only check and invalidates bounding-box calculations. Origin: BUG: C14: Parameter values not validated in preflight in 65 filters #1873.
-
WRONG_DATA · IsoContourDistanceImageFilter · IsoContourDistanceEngine.hpp:715 — Existing invalid Image Geometry spacing reaches distance divisions without validation. Origin: BUG: C14: Parameter values not validated in preflight in 65 filters #1873.
Related issues
- BUG: Hex-grid conversion selects the wrong coincident source node #1816 BUG: Hex-grid conversion selects the wrong coincident source node. It overlaps
ConvertHexGridToSquareGridFilter#1in part. Coordinate the two fixes. - BUG: Curvature fitting returns NaN directions from a singular frame #1819 BUG: Curvature fitting returns NaN directions from a singular frame. It reports the same defect as
FeatureFaceCurvatureFilter#8.
Out of scope
- Findings in filters that have a V&V report are not in this issue. There are 9: ComputeFeatureReferenceCAxisMisorientationsFilter, ComputeFeatureReferenceMisorientationsFilter, ComputeIPFColorsFilter, ComputeNeighborhoodsFilter, ComputeTriangleGeomCentroidsFilter, DBSCANFilter, GroupMicroTextureRegionsFilter, RotateEulerRefFrameFilter, WritePoleFigureFilter. A separate V&V branch handles them.
Vec3::normalizeitself is not changed in this issue. Many callers depend on it. Fix the callers that can get a zero vector.- Every finding is a candidate. It was found and re-checked by reading the code at develop 419f9db, not by running it.
Activity
[-]BUG: Degenerate input not handled in 39 filters[/-][+]BUG: C16: Degenerate input not handled in 39 filters[/+]BUG: TriangleDihedralAngle - Use signed dot products for the corner a…
BUG: TriangleDihedralAngle - Use signed dot products for the corner a…
BUG: Fix 18 filter defects introduced during the 6.5.171 to NX port (#…