Are there any reasons to avoid using tmpnam() to get a name for a temporary file?

Question

I was going to use tmpnam() to name a temporary file that will be renamed later, not deleted. But I found the following documentation and I'm now intrigued about the part that says

"...naive programmers may think it a suitable name for a temporary file."

From http://man7.org/linux/man-pages/man3/tmpnam.3.html

The tmpnam() function returns a pointer to a string that is a valid filename, and such that a file with this name did not exist at some point in time, so that naive programmers may think it a suitable name for a temporary file. If the argument s is NULL this name is generated in an internal static buffer and may be overwritten by the next call to tmpnam(). If s is not NULL, the name is copied to the character array (of length at least L_tmpnam) pointed to by s and the value s is returned in case of success.

Can someone explain why does the documentation say that?

I assume "it" refers to the name generated, and that only naive programmers would think it is adequate to name a temp file, otherwise why mention it? If it isn't an appropriate solution, why does it exist?

I would greatly appreciate if someone could clear it up.

Answer 1

Another program may create a file with exactly that name in the meantime. To be sure you can actually get a file with that filename the creation of filename and file itself should be an atomic, indivisible operation.

It's a bit theoretical as the time between the tmpnam() and actual creation is likely to be very short. But still it's possible.