Glycol47 segfaults in LAPACK routine

[Martin Sjölund]

In MSL 3.2.1 Modelica.Media.Examples.Tests.MediaTestModels.Incompressible.Glycol47:

#0  0x00007ffff7447299 in dgelsx_ () from /usr/lib/liblapack.so.3gf
#1  0x0000000000408b54 in omc_Modelica_Math_Matrices_LAPACK_dgelsx(void,...)(char) ()
#2  0x0000000000408314 in omc_Modelica_Math_Matrices_leastSquares ()
#3  0x0000000000409915 in omc_Modelica_Media_Examples_Tests_Components_FixedMassFlowRate_Medium_Polynomials__Temp_fitting ()
#4  0x0000000000409a47 in omc_Modelica_Media_Examples_Tests_Components_FixedMassFlowRate_Medium_h__T ()
#5  0x000000000040c6c4 in eqFunction_32 ()
#6  0x000000000040f413 in updateBoundParameters ()
#7  0x0000000000466e40 in state_initialization (data=0x7fffffffdcb0, 
    optiMethod=<optimised out>, updateStartValues=1)
    at simulation/solver/initialization.c:437

I recently fixed a code-generation issue for leastSquares, so that it does compile. I even added a testcase for it and I get some result, so the code shouldn't be totally off (you could verify if the testcase gives correct output I guess).

Can someone check if this is an issue with code-generation or a bug in LAPACK where valid input data causes a segfault (unlikely)? I tried to debug it, but all I could see was valid input...