Original answer deleted.
(The meaning of the term “native dll” is not obvious to me. Cf this page.)
You shouldn’t need the ByRef-clause as calling by reference is default.
For LibO BASIC the calling convention cdecl (from the C++ world) seems to be mandatory (or at least default; don’t know how to change).
As I tested with Declare you don’t even need to associate a type (As-clause) with a parameter to pass by reference. Technically a pointer into memory will be passed which does not contain information about the type - neither about the type of elements nor about their number nor about the number and the sizes of the dimensions. If you know better, please tell me.
The code of the dll, on the other hand will contain info about the ElementType of an array to be passed as parameter. Otherwise it cannot cut it it into the pieces needed for indexed access. As long as there isn’t implemented a convention (cf safeArray structure) about including this metainformation about dimensioning with the array itself (say, in the first few bytes), you will have to pass it explicitly in addition to the array reference. The byte-size of the single element, consistently specified for the array to pass/accept in BASIC and in the dll may be implicit.
A completely different thing may be the creation of com objects from registered dll. I never did it.
Now comes a rectified example:
First some BASIC code now reproducibly working
REM ***** BASIC *****
Declare Sub doubleArray Lib "...\ArrayToDll\ArrayToDllVar.dll" Alias "doubleArray" (p1(), ByVal h0 As Long, ByVal h1 As Long)
REM ... must be replaced with the actually correct folder path.
Dim rows As Long, columns As Long
rows = Ubound(pArray, 1) - Lbound(pArray, 1) + 1
columns = Ubound(pArray, 2) - Lbound(pArray, 2) + 1
doubleArray(pArray, rows, columns)
arrayTimes2 = pArray
Then the source of the used library (in Delphi style Pascal)
procedure doubleArray(var pA: Array of Double; h0, h1: Longint); cdecl;
var y, x, ind : Longint;
for y := 0 To h0-1 Do
for x := 0 To h1-1 Do
ind := y * h1 + x;
pA[ind] := pA[ind] * 2.0;
Please let me know if you also want the compiled library and a working .ods example.