You need a map $\mathbb T \rightarrow \mathbb T$, not just a line segment in the torus. I am not sure what criterion you intend to invoke with the diagonal.
Let’s consider an easier example. Let us show that the circle $\mathbb S^1$ does not have the fixed point property. Geometrically it is clear why: we can just rotate the circle a bit and this will not have a fixed point. Let $r:\mathbb S^1 \rightarrow \mathbb S^1$ be such a rotation.
Now one can identify the torus $\mathbb T$ up to homeomorphism with the topological space $\mathbb S^1 \times \mathbb S^1$. Since $r$ doesn’t have a fixed point, the continuous map $$r\times\operatorname{id}:\mathbb S^1\times \mathbb S^1 \rightarrow \mathbb S^1 \times \mathbb S^1$$ doesn’t have a fixed point.
If you insist on defining the torus via gluing the sides of a square and don’t want to use the identification above, we can still write down essentially the same map as I just constructed. The idea is to just translate everything in the square a tiny bit to the right (points which map out of the right bound will be mapped into the left part of the square again). But this way we have to explicitly show that the map is welldefined and continuous, which is not hard but tedious.