Fix p_secstodate overflow handling (bug 22463)

[Joseph Myers]

On Wed, 22 Nov 2017, Florian Weimer wrote:

> So maybe we should just fix the generic/default case to use unknown RR type
> syntax and just dump those records in that format.  I'll send patches for that
> once my current round of docs/syscall wrapper work is done (too many patches
> in flight right now).

We still need to fix the build of p_secstodate so that regressions with 
mainline GCC / binutils can be reliably detected again, which is not 
currently the case and is meaning that new build failures keep appearing 
while old ones such as this one remain unfixed.

-- 
Joseph S. Myers
joseph@codesourcery.com