Seems not only does sed on BSD use different regexp parameters but has a
different pattern set as well. The output HTML may look nasty without
cleaning but if the cleaning goes badly it can make the visual end-product
look worse. In this case spurious 'n' characters around the page.
Dropping the sed-based cleanup entirely in favor of usable pages.
cd $$lang; \
for f in `ls -1 ../templates`; do \
echo "po2html -i ../$$lang.po -t ../templates/$$f"; \
- po2html -i ../$$lang.po -t ../templates/$$f \
- | sed -E s/\>\ \ ?\</\>\\n\</g >$$f || exit 1; \
+ po2html -i ../$$lang.po -t ../templates/$$f >$$f || exit 1; \
done; \
cd ..; \
done