Thoughts on bug 15884

[Paul Eggert]

Leonhard Holz wrote:
> Maybe there are other use case ideas or input sources? And should I vary
> the collation? Actually C.UTF-8 is choosen.

I suggest also using lists of file names, since they're often compared 
using strcoll.  For example, run the following command in the glibc 
source tree:

find . -printf '%f\n'

For locales, I suggest "C" and "en_US.utf8" since they're the 
most-commonly used.  Also, if the source text is (say) German, I would 
also use the German locale for that test.