I'm writing a book, but I liked all its source files distributed on derivative works.
What are the advantages and disadvantages of licensing non-executable content under GPL when compared with CC-BY-SA?
As has already been discussed here, the main reason for not using CC licences for software is that they fail to distinguish between the source and executable ("object code") forms of a work. There are a number of ways of making this distinction, but one I find useful is that the source form is easy to edit, but hard to use, while the executable form is easy to use, but hard to edit.
You have written a book which you intend to distribute in a format that is easy to edit (eg TeX source, oowriter file, some kind of markup language) but hard to read, in additional to the usual form (eg PDF, ebook, paper book) which is easy to read (or use on a reading device) but harder to edit.
Firstly, good for you! Secondly, this distinction is very similar to the one that copyleft software licences try to make. GPLv3 defines the difference thus, in s1:
The “source code” for a work means the preferred form of the work for making modifications to it. “Object code” means any non-source form of a work.
This is compatible with the distinction you and I have both made, above, and moreover contains nothing that implies it's referring to software. I would say that, with a bit of helpful labelling (where you make it clear that your source files are "the preferred form of making modifications" for your book) you could use GPLv3 verbatim, and benefit from the requirement that people using your work downstream would need to maintain the distinction you have started, and keep both parts free.