Fix the dumb way I was passing through the output of cvsweb.cgi . Do

it with popen and fpassthru instead. We still have to read in the first
few lines to look for headers produced by cvsweb, but we can pass the
rest straight through.

Fixes tarball download of the whole source tree.

www/cvsweb/cvsweb.php3

@@ -42,41 +42,25 @@ $name = preg_replace("/ /","\\ ",$name);
 $agent = preg_replace("/ /","\\ ",$agent);
 $encoding = preg_replace("/ /","\\ ",$encoding);
 
-$output = `env PATH=./cvsweb/ QUERY_STRING=$query PATH_INFO=$path SCRIPT_NAME=$name HTTP_USER_AGENT=$agent HTTP_ACCEPT_ENCODING=$encoding $script`;
+$fp = popen("env PATH=./cvsweb/ QUERY_STRING=$query PATH_INFO=$path " .
+            "SCRIPT_NAME=$name HTTP_USER_AGENT=$agent " .
+            "HTTP_ACCEPT_ENCODING=$encoding $script",'r');
 
 #
 # Yuck. Since we can't tell php to shut up and not print headers, we have to
-# 'merge' headers from cvsweb with PHP's. And, since preg_match returns
-# totally unhelpful results, we have to split it up into lines and iterate
-# through them. Again, yuck!
+# 'merge' headers from cvsweb with PHP's.
 #
-$array = split("\n",$output);
-$i = 0;
-for ($i = 0; $i < count($array); $i++) {
-	#
-	# A blank line signifies the end of headers
-	#
-	if (!preg_match("/\w+/",$array[$i])) {
-		#
-		# We're done with the headers, we can stop and finish printing
-		# in the loop below
-		#
-		break;
-	} else {
-		#
-		# It's a header, we use the PHP header() function to add it
-		# to the list of headers that PHP maintains.
-		#
-		header($array[$i]);
-	}
+while ($line = fgets($fp)) {
+    # This indicates the end of headers
+    if ($line == "\r\n") { break; }
+    header(rtrim($line));
 }
 
 #
-# Just print the rest of the output (the $i++ skips the blank line that
-# seperate the headers from the body of the document)
+# Just pass through the rest of cvsweb.cgi's output
 #
-for ($i++; $i < count($array); $i++) {
-	echo "$array[$i]\n";
-}
+fpassthru($fp);
+
+fclose($fp);
 
 ?>