[Debian] cppcheck_TensorApproximationAlgorithm_std on armel architecture

[pgrt]

Hello,

While packaging version 1.18 for Debian, I stumbled upon the following issue on armel architecture, for cppcheck_TensorApproximationAlgorithm_std:

446/483 Test #450: cppcheck_TensorApproximationAlgorithm_std ........................***Failed 0.56 sec
[32mINF - Build the iso-probabilistic transformation [0m
[32mINF - Same distribution for input vector=ComposedDistribution(Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), IndependentCopula(dimension = 3)) and basis=ComposedDistribution(Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), IndependentCopula(dimension = 3)) [0m
[32mINF - Build the iso-probabilistic transformation [0m
[32mINF - Same distribution for input vector=ComposedDistribution(Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), IndependentCopula(dimension = 3)) and basis=ComposedDistribution(Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), Uniform(a = -1, b = 1), IndependentCopula(dimension = 3)) [0m
[32mINF - Transform the input sample in the measure space if needed [0m
[32mINF - Working on rank=1 over 5 [0m
[32mINF - j=0(j varies from 0 to 2) [0m
[32mINF - j=1(j varies from 0 to 2) [0m
[32mINF - j=2(j varies from 0 to 2) [0m
Segmentation fault

Curiously, the error is showing up only when the test is build with the -O1, -O2 or -O3 flags. A gdb session lead to

Program received signal SIGSEGV, Segmentation fault.
dswap (n=<error reading variable: Cannot access memory at address 0xf1173849>, dx=..., incx=10, dy=..., incy=10) at dswap.f:103
103 dswap.f: No such file or directory.
(gdb) bt
#0 dswap (n=<error reading variable: Cannot access memory at address 0xf1173849>, dx=..., incx=10, dy=..., incy=10) at dswap.f:103
#1 0xb517b3c4 in dbdsdc (uplo=..., compq=..., n=10, d=..., e=..., u=..., ldu=1000, vt=..., ldvt=10, q=..., iq=..., work=..., iwork=..., info=0, _uplo=_uplo@entry=1, _compq=_compq@entry=1) at dbdsdc.f:494
#2 0xb51afde0 in dgesdd (jobz=..., m=1000, n=10, a=..., lda=1000, s=..., u=..., ldu=1000, vt=..., ldvt=10, work=..., lwork=770, iwork=..., info=0, _jobz=-1090528180) at dgesdd.f:803
#3 0xb67a775c in OT::MatrixImplementation::computeSVD (this=, u=..., vT=..., fullSVD=false, fullSVD@entry=248, keepIntact=false, keepIntact@entry=236) at /usr/include/c++/11/bits/stl_vector.h:1043
#4 0xb67a7c08 in OT::Matrix::computeSVD (this=this@entry=0xbeffdd58, u=..., vT=..., fullSVD=fullSVD@entry=false, keepIntact=keepIntact@entry=false) at ./lib/src/Base/Type/Matrix.cxx:254
#5 0xb64079ec in OT::SVDMethod::update (this=this@entry=0x537cf8, addedIndices=..., conservedIndices=..., removedIndices=..., row=row@entry=false) at ./lib/src/Base/Algo/SVDMethod.cxx:114
#6 0xb63fedd4 in OT::SVDMethod::solve (this=0x537cf8, rhs=...) at ./lib/src/Base/Algo/SVDMethod.cxx:129
#7 0xb63f64a0 in OT::LeastSquaresMethod::solve (this=this@entry=0xbeffdf50, rhs=...) at ./lib/src/Base/Algo/LeastSquaresMethod.cxx:92
#8 0xb6bb1a88 in OT::TensorApproximationAlgorithm::rankOne (this=0x4f23b0, this@entry=0x3, x=..., y=warning: can't find linker symbol for virtual table for `OT::Sample' value
..., tensor=..., i=0, i@entry=3204440576, marginalResidual=@0x4d5f40: 0, marginalRelativeError=@0x4ebe90: 0)
at ./lib/src/Uncertainty/Algorithm/MetaModel/TensorApproximation/TensorApproximationAlgorithm.cxx:299
#9 0xb6bdb1ac in OT::TensorApproximationAlgorithm::greedyRankOne (this=0x3, this@entry=0xbefff038, x=..., y=..., tensor=..., marginalResidual=@0x4d5f40: 0, marginalRelativeError=@0x4ebe90: 0)
at ./lib/src/Uncertainty/Algorithm/MetaModel/TensorApproximation/TensorApproximationAlgorithm.cxx:199
#10 0xb6bdd74c in OT::TensorApproximationAlgorithm::runMarginal (this=this@entry=0xbefff038, marginalIndex=3204442156, marginalIndex@entry=0, marginalResidual=@0xbeffe830: 1.516643423598251e-314, marginalRelativeError=<error reading
at ./lib/src/Uncertainty/Algorithm/MetaModel/TensorApproximation/TensorApproximationAlgorithm.cxx:160
#11 0xb6bddf14 in OT::TensorApproximationAlgorithm::run (this=0xbefff038, this@entry=0xbefff030) at ./lib/src/Uncertainty/Algorithm/MetaModel/TensorApproximation/TensorApproximationAlgorithm.cxx:135
#12 0x00418e20 in main () at t_TensorApproximationAlgorithm_std.cxx:54

Hoping this helps...

Best regards,
Pierre

I investigated in a gdb session, and

[jschueller]

we should remove TensorApproximationAlgorithm & co

[pgrt]

Hi,

I followed your advice and removed this problematic test, it's fine now.

Best,
Pierre