bug#21766: 25.0.50; delete-trailing-whitespace sometimes deletes non-whitespace

[martin rudalics <rudalics@gmx.at>]

 > Now, the question is, should skip-syntax-backward preserve the
 > match data, or must delete-trailing-whitespace be coded defensively?

The latter.

 > The
 > usual policy has been that code that needs the match data preserved should
 > make sure itself that it is so.

Shouldn't we remove that "\\s-$" rigmarole?  Something like the largely
untested

       (let ((end-marker (copy-marker (or end (point-max))))
	    old-bol new-eol new-bol)
         (goto-char (or start (point-min)))
         (while (re-search-forward "\\\n" end-marker t)
	  (setq new-bol (point))
	  (goto-char (setq new-eol (match-beginning 0)))
           (if (or (zerop (skip-syntax-backward
			  "-" (or old-bol (line-beginning-position))))
		  (looking-at-p ".*\f"))
	      (goto-char new-bol)
	    (delete-region (point) new-eol)
	    (forward-char))
	  (setq old-bol (point)))

It's still not very clean (should line-beginning-position be allowed to
go before START?) so if someone wants to polish it up ....

martin