Tests that cause problems, but are "legal" regexes.  These all have a '*' or '+'
following a regex that can be zero-length.  These will raise an exception while searching,
though not while compiling.
()*	-	c	-	-
a**	-	c	-	-
(a*)*	-	c	-	-
(a*)+	-	c	-	-
(a|)*	-	c	-	-
(a*|b)*	-	c	-	-
(^)*	-	c	-	-
(ab|)*	-	c	-	-
((a*|b))*	-	c	-	-

These are similar to the above, but don't raise an exception, since there's only
a maximum of one possible match of the subexpression.  So these are safe to use,
though probably shouldn't be legal.
^*	-	c	-	-
a*?	-	c	-	-

Wrong-order ranges in charsets.  They are legal, don't cause exceptions, but are shady.
It will consider the first character as in the set, but since it's a negative range,
no other characters are considered part of the range.  In the tests below, the first
will match, but the other three won't.
a[c-a]	ac	y	-	-
a[c-a]	ab	n	-	-
a[c-a]	aa	n	-	-
a[c-a]	a-	n	-	-