When text nodes are parsed with <cr> <lf> line endings the line endings in the result are normalized to line endings with <lf> only.

If the input buffer of such a text node was read up to the <cr> (<lf> is not yet read in input buffer), then the resulting text in the text node will contain an additional <lf>.

I've created a workaround for this i parser.c which works, although not sure if this is the correct way to fix this.

In function "xmlParseCharData" at the location <cr> characters are skipped when followed by <lf> character I've made a small change to test if this <cr> character is at the end of the input buffer and as such not skip it but force reading next part in input buffer:

	    if (*in == 0xD) {
		in++;
		if (*in == 0xA) {
		    ctxt->input->cur = in;
		    in++;
		    ctxt->input->line++; ctxt->input->col = 1;
		    continue; /* while */
		}
		if(*in == 0x0) {
			in--;
			return;
		}
		in--;
	    }