On a MMC3 C++ project I was trying to use the irq-related API (based on nesdoug.h), and spent a long time trying to figure out why the irq wasn't "irqing", until someone pointed out to me that, if cli only happens inside nmi (like the first call to irq_parser possibly does, when the first instruction is a scanline counter), the flag state will be reverted by the nmi's rti. Which means that if the interrupts aren't enabled beforehand during initialization, they will never be.
For now I'm doing an asm volatile("cli") at the start of my main myself.
I think maybe either:
interrupts should start enabled for MMC3, or
if such magic is possible with llvm-mos, they should start enabled if irq functions are being used, or
set_irq_ptr should enable it, or
mapper.h should have a comment telling the user to enable it beforehand
On a MMC3 C++ project I was trying to use the irq-related API (based on nesdoug.h), and spent a long time trying to figure out why the irq wasn't "irqing", until someone pointed out to me that, if
cli
only happens inside nmi (like the first call toirq_parser
possibly does, when the first instruction is a scanline counter), the flag state will be reverted by the nmi'srti
. Which means that if the interrupts aren't enabled beforehand during initialization, they will never be.For now I'm doing an
asm volatile("cli")
at the start of my main myself.I think maybe either:
set_irq_ptr
should enable it, ormapper.h
should have a comment telling the user to enable it beforehand