Looks like the GPIO interrupt is indeed broken. For sure in the AM33x and most probably also in AM387x and AM389x.
I corrected the obvious problem in the intr.c file in \COMMON_ti_v1\AM33X\OAL\INTR\intr.c where the "else if" tests were short circuiting because the software interrupts were moved to higher irq numbers than the GPIO.
The setup calls now at least do not fail but I am still not getting the interrupts to fire.
Is this a known problem? does anyone see a fix in our future?