Program doesn't work

[FerdinandoPH]

Rom works fine on wabbitemu (it's a ti 84 se rom), but I open it on this emulator and this happens: I can exit pressing esc, but that's about it. I'm using a ti nspire cx ii cas with ndless 5.3

[adriweb]

pinging @Vogtinator just in case

[Vogtinator]

Yeah, this looks like it's the usual HW-W-like LCD format mismatch. The issue with the CX II LCD in particular is that there's extra HW which performs rotation, so when writing to the existing framebuffer it's like a 320x240 one, but when changing to a different location it's actually 240x320... That unfortunately needs specific code to handle the is_cx2 case here.

[FerdinandoPH]

Is this project maintained?

[adriweb]

No (read the author profile bio...), a fork would be needed to update the code to use the latest SDK stuff

[Noahmatada]

There is already an is_hww clause in the lcd code that seems to rotate the framebuffer. How would an is-cx2 differ from this, and what would need to be done to implement it?

[Vogtinator]

There is already an is_hww clause in the lcd code that seems to rotate the framebuffer. How would an is-cx2 differ from this, and what would need to be done to implement it?

IIRC it's almost the same. It might have to do an X/Y flip in addition to the 240x320 conversion, but I'm not entirely sure.

You could try

    if(is_hww) {
        return x * 240 + y;
    } else if(is_cx2) {
                return (319-x) * 240 + 239 - y;
        } else {
        return y * 320 + x;
    }

[unilock]

@Vogtinator I've implemented such a check with https://github.com/unilock/nspire-z80/commit/01af73c098837b99aaf863cc36d7c7dc2a2ba5f3, but I'm still running into pretty much exactly the same issue as in the original post.

It seems the screen dimensions are referenced in a few other locations:

• https://github.com/unilock/nspire-z80/blob/01af73c098837b99aaf863cc36d7c7dc2a2ba5f3/src/lcd.c#L132-L147
• https://github.com/unilock/nspire-z80/blob/01af73c098837b99aaf863cc36d7c7dc2a2ba5f3/src/lcd.c#L207
• https://github.com/unilock/nspire-z80/blob/01af73c098837b99aaf863cc36d7c7dc2a2ba5f3/src/lcd.c#L225-L287

Unfortunately my knowledge of C isn't decent enough to know what exactly needs to be changed (nor is my ARM9 assembly, especially).

[Vogtinator]

Try

static void n_set_84_pixel(int x, int y, uint8_t gray, uint8_t *buf){
    for(int dx = 0; dx < 3; ++dx)
        for(int dy = 0; dy < 3; ++dy)
            buf[xy_to_fbo(C_XO + x + dx, C_YO + y + dy)] = gray;
}

Might be a bit slow, but should produce a usable picture.

[unilock]

@Vogtinator Done here: https://github.com/unilock/nspire-z80/commit/4ae848f0d45c373dde92073d2aa95801c711417a

Unfortunately, I still get more or less the same problem. The calculator also crashes and reboots a few seconds after starting the program (this isn't new; even the unmodified version of nspire-z80 did this on my CX2).

Maybe fb_setup(uint8_t *buf) has to be updated for CX2 hardware as well?

[Vogtinator]

That sounds like something else is wrong then

Maybe fb_setup(uint8_t *buf) has to be updated for CX2 hardware as well?

FWICT that just clears the screen and draws a border with an icon. It should still be usable if the border is corrupted.