As far as I remember, fair use is one of these really fuzzy things that give you a lot of leeway.
Lecturers at universities often copy examples and diagrams for a book, to illustrate their point. This is considered 'fair use'.
Simply taking a code section from a book, to illustrate the solution to a question on PM would also be considered 'fair use', as, you've purchased the book as a reference, which it is, and used it in a correct manner to aid in problem solution.
This would be the same as cribbing from this book to gain a solution in code used for a company you're doing work for.
However, posting the whole chapter of the book that contained the code fragment would be overkill, and would definately be a little 'unfair'.. Thus opening you up to copyright violation charge.
I believe that posting an application complexity chunk of code (not a quick demonstration fragment), then, you'd be infringing on fair use.
As a rough guide (and, I believe that's all there really is to go on), small fragments of code used to indicate a solution are fair use, especially if these fragments are commonly available, and heavily used.
Posting large sections of the book are a no-no.
Bear in mind that here, a lot of good Perl work is bandied around. If you make a reference to the author/book you're referred to, then it would almost certainly be considered 'fair use' by the publishers, as it's a strong commendation of their product, making it more likely that others will hear, and thus purchase it themselves.
If you really want the absolute answer in each case, it'd most likely be a case of calling the publishers, as different people are likely to call unfair use in many different circumstances.

Cheers,

Malk