> How it manages to not hang on a real PC is puzzling...

Certain INTs unconditionally reenable the interrupts on a real BIOS. This is partially missing in DOSBox.

Interesting... so they return with something like STI/RETF 2 instead of IRET? I could hook the INTs and force such a condition to test the game with... which ones?

EDIT: I tried it with INT 13, and it does prevent the hang. A simple TSR that hooks INT 13 and does:

1pushf
2call oldint13
3sti
4retf 2

The INT 13 emulation code for the crack handles certain operations, but passes some through to BIOS. The ones passed through could be making up for the handled ones that disable interrupts. I guess a BIOS disassembly could "prove" the theory...

Somebody had made some tests: Testing utilities needed