Actually allow the patchfile committed in r497171 to take effect, to
fix the build on GCC-based architectures.

The failure was not detected during my own testing due to a stale copy
of the same patchfile in the masterport (not committed to svn). The
masterport apparently was able to build even with the patch in the wrong
location.

MFH to be determined.

Pointyhat to: linimon
Approved by: portmgr (tier-2 blanket)