You are right, the manual does say that. However, it doesn't complete that thought elsewhere, so we're left to assume that something like setting the NMI IWDT enable causes the IWDT interrupt to be non-masking, but not setting the enable causes it to be masking. I experimented and found that setting ICU.NMIER.BIT.IWDTEN = 1 causes the IWDT underflow to branch to the NMI interrupt handler (called "__NMI_Handler", although I can't find where that name is defined as a vector), and setting ICU.NMIER.BIT.IWDTEN = 0 causes the IWDT underflow to not cause branch to anything. At least if I can branch to the NMI handler, I can check the underflow bit myself and do what needs to be done. I'm mystified why I still can't branch to my IWDT handler, but I do have a path forward now. Thank you all.
↧