queue/qmail-queue optimization

[Jose Luis Martinez]

Hi,

	Just wanted to know what the qp gurus think about applying the 
following patch to queue/qmail-queue

	Instead of writing the message line by line (depending on perl doing 
command-queueing for the prints), it just does it in chunks. That way 
the code doesn't have to scan for the newlines, saving what looks like 
lots of CPU cycles in calls and string scanning (very rough eye-estimation).

	Will doing this affect neagtively the other Transaction.pm body_xxx 
methods?

	I have tried this plugin on a dev forkserver, and messages have been 
queued without problems. Not tried in production machines, though.

	Also: I chose 4KB chunks because that seems the buffer size of lots of 
devices. Any pointers on the recommended way of retrieving  a better 
buffer size from body_fh, for example?

Best Regards,

Jose Luis Martinez
[EMAIL PROTECTED]
CAPSiDE

--- qmail-queue 2008-05-08 14:06:22.000000000 +0200
+++ qmail-queue_opt     2008-05-20 12:59:55.000000000 +0200
@@ -58,9 +58,17 @@

     $transaction->header->print(\*MESSAGE_WRITER);
     $transaction->body_resetpos;
-    while (my $line = $transaction->body_getline) {
-      print MESSAGE_WRITER $line;
+    my $input = $transaction->body_fh;
+    my $buf;
+
+    while (read($input, $buf, 4096)) {
+        print MESSAGE_WRITER $buf;
     }
+
+#    while (my $line = $transaction->body_getline) {
+#      print MESSAGE_WRITER $line;
+#    }
     close MESSAGE_WRITER;

     my @rcpt = map { "T" . $_->address } $transaction->recipients;