Fix a bug in header parsing in http.lua. After reading a block of

non-whitespace characters, the position counter was advanced one past
where it was supposed to be. This didn't have any bad effect when the
server used CRLF to separate header fields, because it ate the CR and
still recognized LF as ending the field. But it concatenated multiple
header fields when the server only used LF to separate them.

nselib/http.lua

@@ -521,7 +521,6 @@ local function parse_header(header, response)
         pos = pos + 1
       end
       words[#words + 1] = string.sub(header, s, pos - 1)
-      pos = pos + 1
       pos = skip_lws(header, pos)
     end