Another alternative which is:

        if capacity == <SIZE_t> SIZE_MAX:

but I think what you chose brings a correct conversion: capacity is SIZE_t AKA cnp.npy_intp AKA Py_intptr_t AKA "an integer that is the size of a pointer on the platform" whilst, SIZE_MAX is the maximum value for uint64 stored as a uint64 I suppose.

I find using SIZE_MAX for this encoding weird. Would there be configuration for which this would break? Previously, was capacity cast to a size_t (likely a uint64 on most architecture nowadays) or was SIZE_MAX cast to a Py_intptr_t?

@OmarManzoor: can you provide the warning you get on your platform?

This is the precise warning:
sklearn/neighbors/_quad_tree.c:8389:34: warning: comparison of integers of different signs: '__pyx_t_7sklearn_9neighbors_10_quad_tree_SIZE_t' (aka 'long') and 'unsigned long' [-Wsign-compare] __pyx_t_1 = ((__pyx_v_capacity == SIZE_MAX) != 0);