stb_truetype: stbtt__fill_active_edges_new() dy=nan

[MWP]

I won't pretend to know what's going on in this function, but it's creating an ARM FPU fault.

I can trace through the code, and have found the following results in a div by 0.

               // check if final y_crossing is blown up; no test case for this
               if (y_final > y_bottom) {
                  y_final = y_bottom;
                  dy = (y_final - y_crossing ) / (x2 - (x1+1)); // if denom=0, y_final = y_crossing, so y_final <= y_bottom
               }

In this case, y_final=-14 y_crossing=-14 x2=21 and x1=20.

The ARM FPU fault then occurs on:

               step = sign * dy * 1; // dy is dy/dx, change in y for every 1 change in x,

If it matters, the font is: https://www.fontrepo.com/font/4274/painting-with-chocolate

Thanks!

[MWP]

I've temporarily prevented the fault from happening by:

               // check if final y_crossing is blown up; no test case for this
               if (y_final > y_bottom) {
                  y_final = y_bottom;
                  int div = x2 - (x1+1);
                  if (div == 0) div = 1;
                  dy = (y_final - y_crossing ) / div;
               }

I'm guessing this isn't the correct way to go about fixing this though?

[paul-lalonde]

I've just tripped over the same bug. Working fine for many fonts and scales, but hits the divide by zero.
So it looks like the precondition doesn't hold.
Edit: Yes, this appears to be related to the fix for #1057.
Also, if useful I can package up a reproducer. Painfully (I've, um, "ported" this to Plan 9), so really only if it can be useful.