EclaireXL

Nir tested it on the latest core for the mist. It worked! I disabled PBI then it worked on the Eclaire. So something related to an antic write or read completing in the 2nd colour clock of a cycle I guess.

What do I know so far...
pbi disabled -> works
pbi enabled -> breaks

Tried instead by delaying the cpu enable 20 cycles -> breaks
Tried to delay cpu writes to antic alone by 20 cycles -> works...

Surprised by that last result because my plan had been to run two antics against each other - one delayed and the other not delayed. I have that now but my test case doesn't exhibit the problem. Trying to track down why, perhaps its gtia?? Or perhaps I messed this delay up?

Its a plain antic 4 display. Seems to be a kernel that moves pmgs around and sets colours also changes character set. I guess this is rasta converter stuff?

Anyway seems to be for some hscrol values it ends up reading the wrong chars, but only when running in later cycles.

Scrolling can be frozen by replacing sub at acfb with nops. i.e. in freezer press D for debug, then enter BS AC7B,AC7D,EA, then enter q to return. This isn't freezing scrolling pmgs but should be nice and stable for checking logic analyzer for what chars are read...

Wired up the pbi enable/disable to the menu, so I can turn it on/off and see the result on the frozen screen. Enough for now, but getting closer!

There is an extra dma fetch at colour clock 0, investigating what is going on...

Nasty edge case around dma disabling for the last few cycles of the scanline and cpu cycle completion timing, combined with the later cpu cycle completion due to pbi. Think its fixed, just checking!