Take the conic $X: z^2-xy=0$ in $\mathbb P^2$. On the patch with $z=1$ and coordinates $x, y$, $X$ is cut out by $1-xy=0$. On the patch with $y=1$ and coordinates $u, v$, $X$ is cut out by $v^2-u=0$.
For an affine plane curve $C$ cut out by $f(x, y) = 0$ in $\mathbb A^2$, we have $$ \frac{dx}{f_y} = -\frac{dy}{f_x} $$ (because $f$ is the zero function on $C$ so $f_x \, dx + f_y \, dy =0$).
So in our situation, on the patch $\mathbb A^2_{x, y}$, we have $$ \frac{dx}{-x} = \frac{dy}{y} $$ and on the patch $\mathbb A^2_{u, v}$, we have $$ \frac{du}{2v} = \frac{dv}{1} $$. The coordinate changes from $\mathbb A^2_{u, v}$ to $\mathbb A^2_{x, y}$ is given by $x=u/v$ and $y=1/v$. I am trying to show that a Kähler differential on the projective curve are Kähler differentials on the affine curves that paste correctly. Namely, the Kähler differentials above should agree. We have $$ \frac{dy}{y} = \frac{d(1/v)}{1/v} = v \cdot \frac{-1}{v^2} dv = -\frac{1}{v} dv $$ but this should agree with $dv$. Certainly I'm making some very dumb mistake. There is an example of this type of calculation here where it does work, on the second page with a cubic. Please save me from my misery.