Compilation error with openblas

[mariavd]

I noticed a compilation problem when using openblas-0.3.20-gcc-11.2.1-a3n26so (on Alma Linux) but I don't have the time to investigate further. I am copying the output in case somebody else gets stuck. My workaround was to continue without blas because the problematic functions are defined differently with/without it.

[ 59%] Building Fortran object src/libgimic/CMakeFiles/gimic2.dir/jtensor.F90.o
/opt/gimic/20230210/src/libgimic/jtensor.F90:197:55:

  165 |         diapam = ddot(vec_size, this%denbf, 1, this%bfvec, 1)
      |                                               2
......
  197 |                 prsp1 = -ddot(vec_size, this%dendb, 1, this%drvec(1, m), 1)
      |                                                       1
Error: Element of assumed-shape or pointer array as actual argument at (1) cannot correspond to actual argument at (2)
/opt/gimic/20230210/src/libgimic/jtensor.F90:198:54:

  165 |         diapam = ddot(vec_size, this%denbf, 1, this%bfvec, 1)
      |                                               2
......
  198 |                 prsp2 = ddot(vec_size, this%denbf, 1, this%d2fvec(1, k), 1)
      |                                                      1
Error: Element of assumed-shape or pointer array as actual argument at (1) cannot correspond to actual argument at (2)
/opt/gimic/20230210/src/libgimic/jtensor.F90:205:49:

  165 |         diapam = ddot(vec_size, this%denbf, 1, this%bfvec, 1)
      |                                               2
......
  205 |               ppd = ddot(vec_size, this%pdbf, 1, this%drvec(1, m), 1)
      |                                                 1
Error: Element of assumed-shape or pointer array as actual argument at (1) cannot correspond to actual argument at (2)
make[2]: *** [src/libgimic/CMakeFiles/gimic2.dir/build.make:194: src/libgimic/CMakeFiles/gimic2.dir/jtensor.F90.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:241: src/libgimic/CMakeFiles/gimic2.dir/all] Error 2
make: *** [Makefile:146: all] Error 2

The first of the problematic functions where the tensor is contracted looks like this

#ifdef HAVE_BLAS
        vec_size = size(this%bfvec)
        call dgemv('n', vec_size, vec_size, 1.0d0, this%aodens, vec_size, this%bfvec, 1, 0.0d0, this%denbf, 1)
        diapam = ddot(vec_size, this%denbf, 1, this%bfvec, 1)
#else
        this%denbf=matmul(this%bfvec, this%aodens)
        diapam=dot_product(this%denbf, this%bfvec)
#endif

Googling hints that an array element might be passed but then inside the subroutine, there is an attempt to access the whole array. Sadly, I don't have time to take care of it.

[heikef]

Assigning @bast hope this is OK for you. Perhaps this is related to the other problems we face....