There're are excellent answers by Asaf and Brian here. I have my own questions on this:
Asking for an alternative proof approach: I am following Halmos's Naïve Set Theory, and he gives this as an exercise just after proving well ordering theorem, and not in the previous section dealing with Zorn's lemma. It's as if he intends to use well ordering theorem to prove this. And I am just scratching my head how well ordering might be used here to give a proof. Everywhere, Zorn's lemma seems to be used for this.
Asking for a proof explanation: Brian's answer to the above linked question uses Zorn's lemma on the set $\mathscr L$, partially ordered by inclusion, of all the total orders extending the given partial order on the original set, say $P$. But the answer invokes Zorn's lemma by showing that any chain $\mathscr C$ in $\mathscr L$ has an upper bound in $\mathscr L$. The answer mentions to take the obvious choice for this upper bound, which I assume is $\bigcup \mathscr C$. But then this doesn't seem to work if $\mathscr C = \emptyset$. What am I missing? Note that what is to be proven is that $\mathscr L$ is nonempty.
drhab (in comments) has resolved the second question: Brian in his answer, means the restriction of $\le$ to $X\times X$, when he writes $\le\upharpoonright(X\times X)$. Hence $\mathscr L$ is the set of all the total orders on subsets $X$ of $P$ such that they extend the restriction of $\le$ on $X\times X$.