Archived issue #0032744

Modeling Algorithms - Endless loop in GCPnts_UniformDeflection

Open CASCADEOCCT:Modeling Algorithmsclosed16 public notes

Description

To reproduce memory leak and endless loop inside GCPnts_UniformDeflection constructor it’s enough to read Attachment 1 (BREP) by BRepTools::Read() function.
To find out why such “valid“ edges are produced by JTCAFControl_Reader converter, you can read JT file and transfer it to TDocStd_Documen document and then iterate through gotten shapes checking each edge by GCPnts_UniformDeflection algorithm.

Steps to reproduce

TopoDS_Compound comp;
BRep_Builder builder;
BRepTools::Read(comp, "/path/to/Attachment 1 (BREP)", builder);
BRepCheck_Analyzer analyzer(comp);
if (analyzer.IsValid()) {
  TopTools_IndexedMapOfShape edges;
  TopExp::MapShapes(comp, TopAbs_EDGE, edges);
  for (auto edgeNum = 1; edgeNum <= edges.Extent(); ++edgeNum)
  {
    auto edge = TopoDS::Edge(edges(edgeNum));
    Standard_Real firstParam = 0., lastParam = 0.;
    auto pCurve = BRep_Tool::Curve(edge, firstParam, lastParam);
    GeomAdaptor_Curve curveAdaptor(pCurve, firstParam, lastParam);
    GCPnts_UniformDeflection uniformAbs(curveAdaptor, 0.001, firstParam, lastParam); // <= endless loop and memory leak inside
  }
}

Additional information

From TSYS-16-006

Public activity

16 archived notes

Participants are labeled by their role within this record.

01Commenter 12021-12-20 19:27

Branch [archived branch] has been created by Author.

[revision removed]

Detailed log of new commits:

Author: Author
Date: Mon Dec 20 19:27:09 2021 +0300

    0032744: Modeling Algorithms - Memory leak in GCPnts_UniformDeflection

    fixed CPnts_UniformDeflection::More()

02Commenter 12021-12-21 13:33

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

03Commenter 32021-12-21 16:20

In this extremal case the parameter increment in CPnts_UniformDeflection::Perform() is less than the parameter's value precision. For this reason, the parameter value has not been changed, the loop becomes endless.
Tests results are here: http://jenkins-test-occt.nnov.opencascade.com/view/CR32744-master-Author/view/COMPARE/
Unstable cases are different for win & Linux platforms and seems like not related to this issue

04Commenter 42021-12-21 18:38

               myFirstParam == myFirstParam + myDu;

It is better using myDu < Epsilon (myFirstParam)

05Commenter 52021-12-21 18:43

About test case, it is better to read the shape using standard restore command and pass to the test command the shape instead of file name.

06Commenter 12021-12-22 12:52

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

07Commenter 12021-12-22 14:43

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

08Commenter 82021-12-22 20:18

remarks fixed.
tests results: http://jenkins-test-occt.nnov.opencascade.com/view/CR32744-master-Author/view/COMPARE/

09Commenter 12021-12-23 10:31

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

10Commenter 102021-12-27 15:17

    myFirstParam += myDu;
    myFinish = myLastParam - myFirstParam < myTolCur ||
               Abs(myDu) < myTolCur ||
               // to avoid less than double precision endless increment
               myDu < Epsilon (myFirstParam);

Please compute the value of Epsilon (myFirstParam) before entering the loop. It must be invariant.

  else if (aShape.ShapeType() != TopAbs_EDGE)
  {
    theDi << " Shape type must be EDGE\n";
    return 1;
  }

  const TopoDS_Edge& anEdge = *(TopoDS_Edge*) &aShape

Use the method TopoDS:Edge() for safe downcasting of a shape to edge. After that just check the cast result for null shape.

11Commenter 12021-12-27 16:10

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

12Commenter 122021-12-27 16:11

fixed

13Commenter 12021-12-27 17:17

Branch [archived branch] has been updated forcibly by Author.

[revision removed]

14Commenter 142021-12-27 17:20

For integration:
occt - CR32744
products - none

15Commenter 152021-12-31 12:49

Combination -
OCCT branch : [archived branch]
master SHA - [revision removed]
[revision removed]
Products branch : [archived branch] SHA - [revision removed]
was compiled on Linux, MacOS and Windows platforms and tested in optimize mode.

Number of compiler warnings:
No new/fixed warnings

Regressions/Differences/Improvements:
No regressions/differences

CPU differences:
Debian80-64:
OCCT
Total CPU difference: 18220.12000000046 / 18175.980000000385 [+0.24%]
Products
Total CPU difference: 11659.240000000125 / 11591.590000000113 [+0.58%]
Windows-64-VC14:
OCCT
Total CPU difference: 19997.15625 / 19974.59375 [+0.11%]
Products
Total CPU difference: 13120.265625 / 13088.65625 [+0.24%]

Image differences :
No differences that require special attention

Memory differences :
No differences that require special attention

16Commenter 12022-01-02 13:36

Branch [archived branch] has been deleted by Participant.

[revision removed]