3.0.x: user-password length decoding sometimes wrong?

[Alan DeKok]

On Nov 30, 2015, at 11:00 AM, Arran Cudbard-Bell <a.cudbardb at freeradius.org> wrote:
> Better to do something more explicit with a regex, in case we get around to fixing string expansion :)

  Regexes on some platforms don't deal well with embedded zeros.

  I'm happy with leaving %{string:..} as being explicitly not binary safe.

  The alternative is:

	update request {
		Tmp-Octets-0 := &User-Password
	}

	#  Some systems send User-Password = "\000\000foo"
	#  which is stupid
	if (Tmp-Octets-0 =~ /^0x00/) {
		reject
	}

	if (Tmp-Octets-0 =~ /^(0x(..)+)00/) {

		# get the text BEFORE the zeros
		update request {
			Tmp-Octets-0 := "%{1}"
		}
		update request {
			User-Password := "%{string:Tmp-Octets-0}"
		}
        }


  Alan DeKok.