Hi,
The behavior you are seeing is real and the interface difference you noticed (Multilink Universal affected, Cyclone Pro not) is a useful observation. Let me give you some practical guidance based on the type of interrupts involved, because the options differ.
For timer-based interrupts
If your application uses timer overflow, output compare, or similar peripheral-generated interrupts, you do not actually need the "Disable maskable ISR's when stepping" feature to handle them during debugging. Most S12X timer and peripheral modules have a FRZ bit that freezes the module when the device enters BDM active mode, which happens automatically during any halt or single step. With FRZ set, the timer stops counting during stepping and cannot generate an interrupt between steps. This works at the hardware level, independent of the debugger interface, and is the cleaner solution for this class of interrupt.
For external interrupts and keyboard interrupts (KBI)
This is where the situation is more limited. External IRQ and KBI interrupts are asynchronous external signals and there is no hardware freeze mechanism for them. The BDM interface cannot suppress them at the hardware level. The "Disable maskable ISR's when stepping" feature exists precisely to cover this gap by masking them through the CCR I bit during each step.
Since this feature is not restoring the I bit correctly with the USB Multilink Universal in your setup, the most reliable option for debugging code that uses these interrupt sources is to use the Cyclone Pro, which you have already confirmed behaves correctly.
Worth trying
It is also worth testing with CodeWarrior 5.2 if you have access to it. There were some debugger-side changes in that version and the I-bit restoration behavior under the Large memory model with the USB Multilink Universal may be improved, though this cannot be guaranteed.
Hope this helps clarify the options.
Ladislav
