Issues with LAPACKE_xxx_work functions and workspace query

Hi,

My question follows up this old thread:

https://software.intel.com/fr-fr/forums/intel-math-kernel-library/topic/...

I have a similar problem with LAPACKE_dsytrf_work and other LAPACKE_xxx_work functions. Using Gennady's code:

    int work_sz = -1; // workspace query; optimal size is returned in tmp
    info = LAPACKE_dsytrf_work(LAPACK_ROW_MAJOR, 'L', m, NULL, m, NULL, (double *)&tmp, work_sz);

I get LAPACKE_dsytrf_work fails with "Parameter 3 was incorrect":

Major version:           11
Minor version:           3
Update version:          3
Product status:          Product
Build:                   20160413
Processor optimization:  Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) enabled processors
================================================================

Intel MKL ERROR: Parameter 3 was incorrect on entry to DSYTRF.
info=-4 tmp=-92559631349317830736831783200707727132248687965119994463780864.0000

When doing a workspace query, NULL for a and ipiv should be valid parameters, right? It seems that passing a dummy pointer workaround the issue only on the surface. Visual Studio reports a stack corruption around ipiv:

Run-Time Check Failure #2 - Stack around the variable 'ipiv' was corrupted.

As the OP mentioned, passing some allocated space (e.g. int ipiv[10]) kinda works but I would prefer a better option.

Thanks, Samuel