Segmentation fault in reshape (ifx 2025.2.0)

[V-T]

Consider the following minimal example:

integer,allocatable::a(:),b(:,:)
allocate(a(20000000))
allocate(b(2,size(a)/2))
b=reshape(a,shape(b))
end

The executable compiled with ifx 2025.2.0, ends with a segmentation fault in the line with `reshape`. It can be avoided by either:

• compiling with `-heap-arrays`,
• reducing the size of `a`, or
• explicitly specifying the shape in `reshape`, i.e.  `[2,size(a)/2]` instead of `shape(b)`.

Considering that `-heap-arrays` fixes this problem, I assume, it's stack overflow. However, I don't understand, why it's happening in `reshape`, and why explicit shape fixes this problem. Is it a bug?

[MarcGrodent]

I've tested the code on Windows 11 with 3 compilers:

• ifort (2021.10.0): stack overflow at line #4
• ifx (2025.3.0): same stack overflow at line #4
• gfortran (14.2): no problem

To me this is a bug. The call to the 'shape' function returns a huge temporary array that is stored on the stack triggering the overflow. As you pointed out, there is no problem if an array constructor is used for the shape argument (perhaps in this case the array is stored directly on the heap ...?).