MMC3 IRQ Issues about olcnes HOT 1 OPEN

Hi! I'm also following this wonderful series and I'm trying to a step further, to have a more complete emulator.
I also encountered the same bug as you, and from my understanding it comes to the IRQ signal.
Those games (like Kirby and SMB3) are using IRQ to know at which scanlines the PPU is currently at to change the rendering behavior. For Kirby/SMB3 it's the status bar, but in SMB3 there is also the Spade game, where you need to match 3 lines.

After diving a bit deeper on the internals of the CPU, I noticed that when the IRQ is requested by the Mapper, the I flag in the the status register in the CPU was always 1, and therefore IRQ were ignored.

In Kirby, it seems the PPU will trigger an NMI, and in the NMI function, it will clear the I flag.
But in the olc implementation, there is something not right done. And rightfully it seems someone noticed:
#33

By moving the set flag after writing to the stack is solving the problem.
I'll open a Pull request to fix this :)

PS: Didn't see it was from 2020, sorry if I chimed in too late :D

PS2: It's here: #42

from olcnes.