I have found that v2.3 is leaking temporary .m64 files. Not always, but
often. Why not always? Because the current behavior implicitly relies upon
finalize to run. I have made the following patch, which I propose to
include.
Please review.
If there are no objections, I will commit this later. I now go to apply
similar changes to trunk, which I will commit in short order.
--- Noel
Index: v2.3/src/java/org/apache/james/smtpserver/SMTPHandler.java
===================================================================
--- v2.3/src/java/org/apache/james/smtpserver/SMTPHandler.java (revision
427862)
+++ v2.3/src/java/org/apache/james/smtpserver/SMTPHandler.java (working
copy)
@@ -400,36 +400,37 @@
//handle messages
if(mode == MESSAGE_RECEIVED_MODE) {
- getLogger().debug("executing message handlers");
- List messageHandlers = handlerChain.getMessageHandlers();
- int count = messageHandlers.size();
- for(int i =0; i < count; i++) {
-
((MessageHandler)messageHandlers.get(i)).onMessage(this);
- //if the response is received, stop processing of
command handlers
- if(mode == MESSAGE_ABORT_MODE) {
- break;
+ try {
+ getLogger().debug("executing message handlers");
+ List messageHandlers =
handlerChain.getMessageHandlers();
+ int count = messageHandlers.size();
+ for(int i =0; i < count; i++) {
+
((MessageHandler)messageHandlers.get(i)).onMessage(this);
+ //if the response is received, stop processing of
command handlers
+ if(mode == MESSAGE_ABORT_MODE) {
+ break;
+ }
}
- }
- }
-
- //do the clean up
- if(mail != null) {
- if (mail instanceof Disposable) {
- ((Disposable) mail).dispose();
- }
+ } finally {
+ //do the clean up
+ if(mail != null) {
+ if (mail instanceof Disposable) {
+ ((Disposable) mail).dispose();
+ }
- // remember the ehlo mode
- Object currentHeloMode = state.get(CURRENT_HELO_MODE);
+ // remember the ehlo mode
+ Object currentHeloMode =
state.get(CURRENT_HELO_MODE);
- mail = null;
- resetState();
+ mail = null;
+ resetState();
- // start again with the old helo mode
- if (currentHeloMode != null) {
- state.put(CURRENT_HELO_MODE,currentHeloMode);
+ // start again with the old helo mode
+ if (currentHeloMode != null) {
+ state.put(CURRENT_HELO_MODE,currentHeloMode);
+ }
+ }
}
}
-
}
theWatchdog.stop();
getLogger().debug("Closing socket.");
Index: v2.3/src/java/org/apache/james/smtpserver/DataCmdHandler.java
===================================================================
--- v2.3/src/java/org/apache/james/smtpserver/DataCmdHandler.java
(revision 427862)
+++ v2.3/src/java/org/apache/james/smtpserver/DataCmdHandler.java
(working copy)
@@ -278,6 +278,23 @@
mail.setAttribute(SMTP_AUTH_USER_ATTRIBUTE_NAME,
session.getUser());
}
session.setMail(mail);
+ } catch (MessagingException me) {
+ // if we get here, it means that we received a
+ // MessagingException, which would happen BEFORE we call
+ // session.setMail, so the mail object is still strictly
+ // local to us, and we really should clean it up before
+ // re-throwing the MessagingException for our call chain
+ // to process.
+ //
+ // So why has this worked at all so far? Initial
+ // conjecture is that it has depended upon finalize to
+ // call dispose. Not in the MailImpl, which doesn't have
+ // one, but even further down in the
MimeMessageInputStreamSource.
+
+ if (mail != null) {
+ mail.dispose();
+ }
+ throw me;
} finally {
if (recipientCollection != null) {
recipientCollection.clear();
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
