This question comes from Boas, Mathematical Methods in the Physical Sciences, chapter 4, section 11: "Change of variables". Exercise question 3.
Suppose that $w=f(x,y)$ satisfies $$ \frac{\partial^2 w}{\partial x^2}-\frac{\partial^2 w}{\partial y^2}=1. \tag{1} $$ Put $x=u+v$, $y=u-v$ and show that $w$ satisfies $\dfrac{\partial^2 w}{\partial u\partial v}=1$. Hence solve the equation.
I can obtain an answer by a combination of the two new variables, \begin{align} u&=\dfrac{x+y}{2},\\ v&=\dfrac{x-y}{2}. \end{align} Because then I can write $$\dfrac{\partial w}{\partial x}=\dfrac{\partial w}{\partial u}\dfrac{\partial u}{\partial x}+\dfrac{\partial w}{\partial v}\dfrac{\partial v}{\partial x}=\frac{1}{2}\left(\dfrac{\partial w}{\partial u}+\dfrac{\partial w}{\partial v}\right)~,$$ and thus it also follows that $$\frac{\partial^2 w}{\partial x^2}=\frac{1}{4}\frac{\partial^2 w}{\partial u^2}+\frac{1}{2}\frac{\partial^2 w}{\partial u\partial v}+\frac{1}{4}\frac{\partial^2 w}{\partial v^2}.\tag{2a}$$ If you do the same for the other equation, $$\frac{\partial^2 w}{\partial y^2}=\frac{1}{4}\frac{\partial^2 w}{\partial u^2}-\frac{1}{2}\frac{\partial^2 w}{\partial u\partial v}+\frac{1}{4}\frac{\partial^2 w}{\partial v^2},\tag{2b}$$ you can combine them with the given equation to find the desired relationship.
But I'm not very satisfied with the answer: we were given $w(x,y)$ and we use that that is equal to $w(x(u,v),y(u,v))$. However, the chain rule as I applied it above is applied to a function of the form $w(u(x,y),v(x,y))$. It thus seems to me that equation (1) is: \begin{align} \frac{\partial^2 w(x(u,v),y(u,v))}{\partial x^2} - \frac{\partial^2 w(x(u,v),y(u,v))}{\partial y^2}=1, \end{align} while a subtraction of eq (2b) from (2a) gives: \begin{align} \frac{\partial^2 w(u(x,y),v(x,y))}{\partial x^2} - \frac{\partial^2 w(u(x,y),v(x,y))}{\partial y^2} = \frac{\partial^2 w(u(x,y),v(x,y))}{\partial u\partial v}.\tag{3} \end{align} It is not clear to me that the left-hand side in equation (3) equally satisfies equation 1, because the argument of $w$ is in reverse order.
Does my thinking go wrong somewhere? Or is my proposed answer wrong?
(My question also applies to changes of variables in general -- it is still quite unclear to me what the exact rules are.)
Your difficulty stems from an abuse of notation which is very common in these circles: The same letter $w$ is used for two different functions, one in the $(x,y)$-world, and one in the $(u,v)$-world.
At the start you are given a function $$(x,y)\mapsto f(x,y)\ ,$$ satisfying a certain pde. Now you introduce a new function of other variables $u$, $v$ by putting $$(u,v)\mapsto F(u,v):=f(u+v,u-v)\ .$$ Of course you have some physical interpretation, e.g., $w:=f(x,y)$ could be the temperature at the point $(x,y)$, and you introduce new coordinates $(u,v)$ in the $(x,y)$-plane. You still call the temperature at the points of the plane $w$, and say "$w$ is now a function of $u$ and $v$". But the chain rule does not know about such sentiments. It says that $$F_u(u,v)=f_x(u+v,u-v)\cdot1+f_y(u+v,u-v)\cdot1\ ,$$ and that $$\eqalign{F_{uv}(u,v)&=f_{xx}(u+v,u-v)\cdot1+f_{xy}(u+v,u-v)\cdot(-1)\cr &\qquad+f_{yx}(u+v,u-v)\cdot1+f_{yy}(u+v,u-v)\cdot(-1)\ .\cr &=f_{xx}(u+v,u-v)-f_{yy}(u+v,u-v)\cr &=1\ .\cr}$$