How can teachers warn students about common mistakes without causing the student to make the mistake?

Whether you mention the common mistake or not, it is likely to be made by some student. I think the best approach is to mention this mistake but you need to be very clear about why it is wrong, and show the thought process that arrives you at the correct answer. Many educators concern about the consequences of introducing the idea of a wrong answer because it may concern students. But in my experience when I was a student, often in my study I would come across these wrong answers and it would have been helpful to be able to talk myself out of them being right, by applying the same thought process that my teacher provided me with.

I constantly present wrong answers, identified as such, in my class. My strategy for dealing with exactly the issue you present is that wrong statements, and only (intentionally) wrong statements, appear in red. I hope that this is a convenient and reliable visual cue. Students hate it, though.

One approach that has given me reasonable success, and that students, while not praising, at least don't complain about, is this: when I know students will make a common mistake, I present them with an exam question saying:

Your friend thinks that $$int frac{mathrm dx}{1 + x^2} qquadtext{equals}qquad ln(1 + x^2).$$

(a) What mistake(s) did your friend make?

(b) What is the correct solution?

Examples of this sort have to be chosen with some care—it is not always reasonable for all errors to expect the student to be able to answer (a) beyond "it's wrong"—but I think that this particular one fits well.

(This is quite similar to @ElizaWilson's suggestion. It avoids possible issues with fear of public speaking, but maybe just substitutes another kind of pressure.)

You could perhaps use examples and non-examples.

This is less about explicitly telling students what the common mistakes are, but more about where the theorem or result applies and where it doesn't.

At high school, this might be showing students what is meant by angles in the same segment and what is not meant by angles in the same segment, for example.

In the case of your example, it might be about giving examples of what looks like $frac{kf'(x)}{f(x)}$ and what does not. Discussion can then be focused about why thinking it would be a log integral is wrong and that students should consider other techniques.

Here's another approach when there is a common pitfall that you wish the students to avoid. After teaching the correct reasoning: present the error to the class and ask a student to identify, explain, and correct the mistake. Being able to correct others' mistakes shows a high-level of understanding, and students who make that same mistake might be able to realize and fix it after seeing how.

Then, the students are forming an understanding of why a mistake is wrong, rather than thinking that it is actually correct.

This question is just old wine in new bottles. The situation referred to is true across the board in cognition. The mechanics of the situation is that mls always trumps gls, ‘mls’ standing for the ‘momentary life situation’, and ‘gls’ standing for the ‘general life situation’, these terms having been introduced by Kurt Lewin, the founder of Social Psychology. Therefore anyone who gives any thought to this situation realizes, at least intuitively, the problem with foregrounding an error. As is often the case, humor can make the issue very clear, as in the well-know joke/insult: “Your house burn down and all your family perish in the fire, God forbid.” In Language Arts, when it is felt necessary to cite an erroneous item, it is customary to precede it by an asterisk. There is another phenomenon at play here: any foregrounded condition is taken as necessary. (That is why definitions, even in a rigorous discipline like Mathematics, are given simply in ‘if’ form (as opposed to ‘if, and only if,’ form) – the ‘only if’ (necessary) condition is universally understood by the mere fact of foregrounding.) Thus we come to an error inadvertently introduced by mathematics teachers, namely, the belief that the logarithm (function) doesn’t exist. By heavily foregrounding the fact that n is not equal to -1 in the well-known integration formula, students internalize that n not being equal to -1 is a necessary condition, and therefore the integral of f(x) = x raised to the -1 power ‘does not exist’, all the more so because this integral is typically not dealt with until a significant amount of time later in the course. All this is due to operating at the object-level. There is not problem if, as in discussions like this, you clearly move to the meta level – but moving to the meta level on the fly in a classroom presentation and quickly transitioning back to the object level can be difficult. Meta probably has to be done in discussion mode, not merely lecture mode.

I don't think you have anything to be afraid of here, unless you spend all of your time playing what-if games with common pitfalls, or peppering the board with false statements and not clarifying them as such. As a class conversation about mistakes-to-avoid or a way to encourage students to always check their answers, I think bringing these up can be really helpful.

One of the reason why I avoid warning of common mistakes is because I can imagine how a student could see the intuitive appeal in the answer, even though the logic is flawed.

These are common errors, whether you introduce them or not. You want to teach your students to apply some slow thinking to check an answer or analyze an argument, and having some examples ready that test their ability to do this on the fly could benefit them.

For my part, I always end a conversation of this type by writing "NO" or "FALSE" or "DON'T DO THIS" above the common pitfall we were discussing, so as to not leave something misleading on the board.