Linux buildserver failures

[Russell Taylor]

I've been taking a bit of a look at the failures on the Linux buildserver.

First the LoadInstrumentTest one. The uncaught exception comes out of OpenCASCADE. Here's a backtrace:

#0  0x00000036b90c3d70 in __cxa_throw () from /usr/lib64/libstdc++.so.6
#1  0x0000000001ea4a4e in Standard_ConstructionError::Throw (this=0x20008a50)
    at ../../../drv/Standard/Standard_ConstructionError_0.cxx:52
#2  0x0000000001ea4b3c in Standard_ConstructionError::Raise (
    AString=0x7fe3cce88228 "Bnd_Box is void")
    at ../../../drv/Standard/Standard_ConstructionError_0.cxx:42
#3  0x00007fe3cce09460 in Bnd_Box::Get (this=0x7fffd68592e0, 
    x=@0x7fffd6859120, y=@0x7fffd6859118, z=@0x7fffd6859110, 
    X=@0x7fffd6859108, Y=@0x7fffd6859100, Z=@0x7fffd68590f8)
    at ../../../src/Bnd/Bnd_Box.cxx:201
#4  0x00007fe3ce0044da in BRepMesh_IncrementalMesh::Update (
    this=0x7fffd6859230, S=@0x7fffd6859248)
    at ../../../src/BRepMesh/BRepMesh_IncrementalMesh.cxx:160
#5  0x00007fe3ce004e88 in BRepMesh_IncrementalMesh::Perform (
    this=0x7fffd6859230)
    at ../../../src/BRepMesh/BRepMesh_IncrementalMesh.cxx:145
#6  0x00007fe3ce004fae in BRepMesh_IncrementalMesh (this=0x7fffd6859230, 
    S=@0x7fffd6859350, D=0.001, Rel=<value optimized out>, Ang=0.5)
    at ../../../src/BRepMesh/BRepMesh_IncrementalMesh.cxx:75
#7  0x00007fe3cdfe8a9d in BRepMesh::Mesh (S=@0x222bdf0, d=0)
    at ../../../src/BRepMesh/BRepMesh.cxx:15
#8  0x0000000000a1c973 in Mantid::Geometry::OCGeometryGenerator::AnalyzeObject
    (this=0x2c4bb90) at Geometry/src/OCGeometryGenerator.cpp:99
#9  0x0000000000aa89e5 in Mantid::Geometry::OCGeometryHandler::Triangulate (
    this=0x7fffd68593d0) at Geometry/src/OCGeometryHandler.cpp:50
#10 0x0000000000aa88a6 in Mantid::Geometry::OCGeometryHandler::NumberOfPoints (
    this=0x7fe3c9026f00) at Geometry/src/OCGeometryHandler.cpp:92
#11 0x0000000000aa2543 in Mantid::Geometry::CacheGeometryGenerator::Generate (
    this=0x2cd3bc0) at Geometry/src/CacheGeometryGenerator.cpp:37
#12 0x0000000000aa5da5 in Mantid::Geometry::CacheGeometryHandler::Triangulate (
    this=0x2cd4d40) at Geometry/src/CacheGeometryHandler.cpp:52
#13 0x0000000000aa5b49 in Mantid::Geometry::CacheGeometryHandler::NumberOfPoints (this=0x2cd4d40) at Geometry/src/CacheGeometryHandler.cpp:97
#14 0x0000000000a72a74 in Mantid::Geometry::vtkGeometryCacheWriter::addObject (
    this=0x2b970f0, obj=0x2cd3b00)
    at Geometry/src/vtkGeometryCacheWriter.cpp:85
#15 0x000000000071fd02 in Mantid::DataHandling::LoadInstrument::exec (
    this=0x7fffd685b030) at DataHandling/src/LoadInstrument.cpp:291
#16 0x0000000000d8a14b in Mantid::API::Algorithm::execute (this=0x7fe3c9026f00)
    at API/src/Algorithm.cpp:180
#17 0x0000000000421709 in LoadInstrumentTest::testExecIDF_for_unit_testing (
    this=0x654680) at LoadInstrumentTest.h:357
#18 0x000000000043b1fc in TestDescription_LoadInstrumentTest_testExecIDF_for_unit_testing::runTest (this=0x6549e0) at runner.cpp:58
#19 0x0000000000408df7 in CxxTest::RealTestDescription::run (this=0x6549e0)
    at ../../../Third_Party/include/cxxtest/RealDescriptions.cpp:96
#20 0x00000000004411da in CxxTest::TestRunner::runTest (this=0x7fffd685ec1f, 
    td=@0x6549e0) at ../../../Third_Party/include/cxxtest/TestRunner.h:74
#21 0x00000000004412be in CxxTest::TestRunner::runSuite (this=0x7fffd685ec1f, 
    sd=@0x654600) at ../../../Third_Party/include/cxxtest/TestRunner.h:61
#22 0x00000000004413b1 in CxxTest::TestRunner::runWorld (this=0x7fffd685ec1f)
    at ../../../Third_Party/include/cxxtest/TestRunner.h:46
#23 0x000000000044147c in CxxTest::TestRunner::runAllTests (
    listener=@0x7fffd685ec70)
    at ../../../Third_Party/include/cxxtest/TestRunner.h:23
#24 0x000000000044150f in CxxTest::ErrorFormatter::run (this=0x7fffd685ec70)
    at ../../../Third_Party/include/cxxtest/ErrorFormatter.h:47
#25 0x0000000000408f9b in main () at runner.cpp:16

Secondly, the crash of BackToBackExponentialTest, which doesn't appear to happen anywhere else (not on linuxs1 or Martyn's machine). This has been happening since check-in #2570.

It turns out that the crash is at line 186 of Algorithm.cpp when it tries to access the variable Algorithm::g_execCount. So, nothing directly related to the check-in. I suspect that the new code strays past the end of one of the C-style arrays that are being used to talk to gsl and writes into the memory where g_execCount lives - and that we just get away with it on other machines.

[Russell Taylor]

N.B. To set a breakpoint on the previously unknown exception thrown here, do "bcxa_throw" in gdb.

[Russell Taylor]

(In [2658]) Fix Linux buildserver test failure. It turned out to be the torus shape causing the problem, so we have removed support for toruses. Re #683.

[Nick Draper]

Completed in iteration 18

[Stuart Campbell]

This ticket has been transferred to github issue ​1531