Both PHP's native implementation and the function of ericth at NOSPAM dot pennyworth dot com (below) contain an feature/error (usage for SMTP mismatch) which has been solved in my adjustment below.

The error causes a text containing period character(s), when passed through these functions, end up with encoded lines starting with a period character, the first period character on that line will be discarded when it is transported over SMTP.

Solution: add (another) leading '.' to period characters which would end up as first character on a line when encoded.

See http://stackoverflow.com/a/13949483: "This is a dirty artifact of your transport layer... SMTP is the most probable culprit (see the call for caution in the mail function documentation) but there may be other low level mecanisms that behave similarly. For example if you tweaked the sendmail_path setting or use a buggy sendmail program you may experience similar woes."

<?php
define('PHP_QPRINT_MAXL', 75);

function leading_dot_fixed_php_quot_print_encode($str)
{
    $lp = 0;
    $ret = '';
    $hex = "0123456789ABCDEF";
    $length = strlen($str);
    $str_index = 0;
    
    while ($length--) {
        if ((($c = $str[$str_index++]) == "\015") && ($str[$str_index] == "\012") && $length > 0) {
            $ret .= "\015";
            $ret .= $str[$str_index++];
            $length--;
            $lp = 0;
        } else {
            if (ctype_cntrl($c) 
                || (ord($c) == 0x7f) 
                || (ord($c) & 0x80) 
                || ($c == '=') 
                || (($c == ' ') && ($str[$str_index] == "\015")))
            {
                if (($lp += 3) > PHP_QPRINT_MAXL)
                {
                    $ret .= '=';
                    $ret .= "\015";
                    $ret .= "\012";
                    $lp = 3;
                }
                $ret .= '=';
                $ret .= $hex[ord($c) >> 4];
                $ret .= $hex[ord($c) & 0xf];
            } 
            else 
            {
                if ((++$lp) > PHP_QPRINT_MAXL) 
                {
                    $ret .= '=';
                    $ret .= "\015";
                    $ret .= "\012";
                    $lp = 1;
                }
                $ret .= $c;
                if($lp == 1 && $c == '.') {
                    $ret .= '.';
                    $lp++;
                }
            }
        }
    }

    return $ret;
}

?>

One will like to know and clearly read that RFC2045 specifies a line shall not exceed 75 characters.
Accordingly, quoted_printable_encode() splits line at this limit.

Two bugs:

1) your linebreak is wrong

                $linebreak = "\r\n";

2) continuation of lines with no whitespace is broken

                                        /*
                                         * the text after the whitespace will have to 
                                         * be read again ( + any additional characters
                                         * that came into existence as a result of the
                                         * encoding process after the whitespace)
                                         *
                                         * Also, do not start at 0, if there was *no*
                                         * whitespace in the whole line
                                         */
                                        if (($i + $addtl_chars) > $whitesp_diff) {  
                                                $output .= substr($cur_conv_line, 0,
                                                    (strlen($cur_conv_line) - $whitesp_diff)) .
                                                    $linebreak;
                                                $i = $i - $whitesp_diff + $addtl_chars;
                                        } else {
                                                /* emit continuation --mirabilos */
                                                $output .= $cur_conv_line .
                                                    '=' . $linebreak;
                                        }

I have re-written the PHP 5.3.8 function for quoted_printable_encode into PHP for use with PHP < 5.3.  Tested with PHP 5.2.11.

<?php
define('PHP_QPRINT_MAXL', 75);

function php_quot_print_encode($str)
{
    $lp = 0;
    $ret = '';
    $hex = "0123456789ABCDEF";
    $length = strlen($str);
    $str_index = 0;
    
    while ($length--) {
        if ((($c = $str[$str_index++]) == "\015") && ($str[$str_index] == "\012") && $length > 0) {
            $ret .= "\015";
            $ret .= $str[$str_index++];
            $length--;
            $lp = 0;
        } else {
            if (ctype_cntrl($c) 
                || (ord($c) == 0x7f) 
                || (ord($c) & 0x80) 
                || ($c == '=') 
                || (($c == ' ') && ($str[$str_index] == "\015")))
            {
                if (($lp += 3) > PHP_QPRINT_MAXL)
                {
                    $ret .= '=';
                    $ret .= "\015";
                    $ret .= "\012";
                    $lp = 3;
                }
                $ret .= '=';
                $ret .= $hex[ord($c) >> 4];
                $ret .= $hex[ord($c) & 0xf];
            } 
            else 
            {
                if ((++$lp) > PHP_QPRINT_MAXL) 
                {
                    $ret .= '=';
                    $ret .= "\015";
                    $ret .= "\012";
                    $lp = 1;
                }
                $ret .= $c;
            }
        }
    }

    return $ret;
}

?>

Note that this function returns the quoted-printable string with Windows / RFC822 CRLF line breaks. If you're passing the output of this function into mail(), you may need to convert its line breaks to Unix LF style first.

A function that QP-encodes an input string (written for PHP < 5.3) and
wordwraps it at the same time, in order to avoid classification according to the MIME QP LONG LINE rule of SpamAssassin.  Thanks for Matt Jeffers to point out errors in the below quoted_printable script!

<?php
function quoted_printable_encode($input, $line_max = 75) {
   $hex = array('0','1','2','3','4','5','6','7',
                          '8','9','A','B','C','D','E','F');
   $lines = preg_split("/(?:\r\n|\r|\n)/", $input);
   $linebreak = "=0D=0A=\r\n";
   /* the linebreak also counts as characters in the mime_qp_long_line
    * rule of spam-assassin */
   $line_max = $line_max - strlen($linebreak);
   $escape = "=";
   $output = "";
   $cur_conv_line = "";
   $length = 0;
   $whitespace_pos = 0;
   $addtl_chars = 0;

   // iterate lines
   for ($j=0; $j<count($lines); $j++) {
     $line = $lines[$j];
     $linlen = strlen($line);

     // iterate chars
     for ($i = 0; $i < $linlen; $i++) {
       $c = substr($line, $i, 1);
       $dec = ord($c);

       $length++;

       if ($dec == 32) {
          // space occurring at end of line, need to encode
          if (($i == ($linlen - 1))) {
             $c = "=20";
             $length += 2;
          }

          $addtl_chars = 0;
          $whitespace_pos = $i;
       } elseif ( ($dec == 61) || ($dec < 32 ) || ($dec > 126) ) {
          $h2 = floor($dec/16); $h1 = floor($dec%16);
          $c = $escape . $hex["$h2"] . $hex["$h1"];
          $length += 2;
          $addtl_chars += 2;
       }

       // length for wordwrap exceeded, get a newline into the text
       if ($length >= $line_max) {
         $cur_conv_line .= $c;

         // read only up to the whitespace for the current line
         $whitesp_diff = $i - $whitespace_pos + $addtl_chars;

        /* the text after the whitespace will have to be read
         * again ( + any additional characters that came into
         * existence as a result of the encoding process after the whitespace)
         *
         * Also, do not start at 0, if there was *no* whitespace in
         * the whole line */
         if (($i + $addtl_chars) > $whitesp_diff) {
            $output .= substr($cur_conv_line, 0, (strlen($cur_conv_line) - 
                           $whitesp_diff)) . $linebreak;
            $i =  $i - $whitesp_diff + $addtl_chars;
          } else {
            $output .= $cur_conv_line . $linebreak;
          }

        $cur_conv_line = "";
        $length = 0;
        $whitespace_pos = 0;
      } else {
        // length for wordwrap not reached, continue reading
        $cur_conv_line .= $c;
      }
    } // end of for

    $length = 0;
    $whitespace_pos = 0;
    $output .= $cur_conv_line;
    $cur_conv_line = "";

    if ($j<=count($lines)-1) {
      $output .= $linebreak;
    }
  } // end for

  return trim($output);
} // end quoted_printable_encode
?>