MATLAB: Correct use of pointer arithmetic with lib.pointer

MATLABpointer arithmetic libpointer lib.pointer

Having read https://fr.mathworks.com/help/matlab/ref/lib.pointer.plus.html, I am doing some tests with pointer arithmetic.
First script:
clear
clc
p = libpointer;
setdatatype(p, 'int32Ptr');
p.Value = [0 0 0 0 0 0];
p2 = p + 3;
p2.Value(1) = 2;
disp(['p = ', mat2str(p.Value)]);
disp(['p2 = ', mat2str(p2.Value)]);
Second script:
clear
clc
p = libpointer;
setdatatype(p, 'int32Ptr');
p.Value = [0 0 0 0 0 0];
p2 = p + 3;
p.Value(1) = 1;
disp(['p = ', mat2str(p.Value)]);
disp(['p2 = ', mat2str(p2.Value)]);
The first script seems to always produce the (expected) output:
p = [0 0 0 2 0 0]
p2 = [2 0 0]
The second one, however, doesn't always produce the (expected) output which is
p = [1 0 0 0 0 0]
p2 = [0 0 0]
Sometimes, p2.Value will have what seems garbage values in it, as shown here:
p = [1 0 0 0 0 0]
p2 = [0 256035536 0]
The random results of script 2 make me suspect script 1 might not be reliable as well…
Is there something wrong in my scripts?

Best Answer

Assigning a new value to the base (original) pointer of a user created libpointer invalidates all pointers created from it. If the libpointer was returned from an external function (MATLAB does not own the memory pointed to) then modifications of the value will be inplace and derived pointers are not invalidated. Try the following code (I prefer creating the pointer in one step.)
p = libpointer('int32Ptr',[0 0 0 0 0 0]);
pa=p+0;  %pa=p does not do the job both pointers reference the exact same object
p2 = p + 3; 
pa.Value(4) = 3;
disp(['p = ', mat2str(p.Value)]);
disp(['p2 = ', mat2str(p2.Value)]);
I hope this is an experiment for use with loadlibrary/calllib. Otherwise I am curious what your use case is. There should be no need for libponters when not interfacing to a shared library.
Related Question