Situation: an FPGA drives the IRQ6 input, the portF pin. It's set up to use another chip select as the autovector generator. This mostly works. The FPGA pulls it low, the ISR runs, and the ISR pokes the FPGA to release (raise) the interrupt request pin before it exits.
But sometimes the firmware wants to shut down the associated subsystem so tells the fpga to clear the interrupt request, which de-asserts (pulls high) the port F pin. Then the cpu traps with a "spurious interrupt" error and *everything* is trashed. It takes a full powerdown/powerup to recover.
I can fix this by never disabling the interrupt signal from the fpga, and ignoring any unwanted interrupts in the isr. But it's curious. Seems like requesting the interrupt, then changing your mind, is really a bad thing to do.
John