From: SourceForge.net <no...@so...> - 2004-12-23 20:16:07
|
Patches item #1088651, was opened at 2004-12-20 15:36 Message generated for change (Comment added) made by nosnilmot You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=300235&aid=1088651&group_id=235 Category: None Group: None >Status: Closed >Resolution: Accepted Priority: 5 Submitted By: Felipe Contreras (revo) Assigned to: Stu Tomlinson (nosnilmot) Summary: Yet another MSN fix Initial Comment: If we send a message, and the ACK takes more than one minute then we consider it a time out, and destroy certain data. If the ACK actually comes after a minute then we crash. This fixes that, althought probably it will leak data, but just in the case the ACK takes more than 10 minutes. ---------------------------------------------------------------------- >Comment By: Stu Tomlinson (nosnilmot) Date: 2004-12-23 15:16 Message: Logged In: YES user_id=309779 As discussed in #gaim - I reset the timeout to 1 minute, and removed session->pend_swboards (using session->switches to check for the switchboard instead) Thanks! ---------------------------------------------------------------------- Comment By: Felipe Contreras (revo) Date: 2004-12-23 14:32 Message: Logged In: YES user_id=158337 Now messages are sent even if we close the conversation. I added a new function and I think it's a better design. ---------------------------------------------------------------------- Comment By: Stu Tomlinson (nosnilmot) Date: 2004-12-23 13:33 Message: Logged In: YES user_id=309779 Why do we need another list of switchboards? isn't session->switches sufficient for what session->pend_swboards does? (my minimal testing by removing all references to pend_swboards and adjusting the g_list_find in got_swboard didn't show up any problems) With the latest patch, messages sent before closing the conversation window are silently dropped. Wouldn't it be better to process queued messages before sending the OUT and destroying the switchboard in msn_convo_closed and msn_chat_leave? Do we still need to increase the timeout to 10 minutes given the other changes to how ACKs are now referenced? Sorry for all the questions :-) ---------------------------------------------------------------------- Comment By: Felipe Contreras (revo) Date: 2004-12-23 05:53 Message: Logged In: YES user_id=158337 Fixes for bugs in 1086093. If a conversation window is closed after we receive the XFR answer, but before we connect to the switchboard, a crash happened. Now it shouldn't. Also fix for bug 1085594, so now the name appears in the info dialog. ---------------------------------------------------------------------- Comment By: Felipe Contreras (revo) Date: 2004-12-23 02:47 Message: Logged In: YES user_id=158337 Ok, it seems the changes to the leak helped to return to a more usual ref/unref usage. Also a fix for when the conversation window is closed before the server returns a switchboard. Guys you should tell me about this kind of stuff so I can fix them. I check the tracker but not that much and I've not been following them for a while. If someone finds a reproduceable bug I can probably fix it. ---------------------------------------------------------------------- Comment By: Felipe Contreras (revo) Date: 2004-12-22 17:51 Message: Logged In: YES user_id=158337 This one (if works) will not leak. Also a fix for a bug marv reported. It happened because we handled a switchboard closing/error as a NAK. So we tried to send a SLP message part again, through an inexistant switchboard, so it crashed. Now it just dies without trying to resend anything. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=300235&aid=1088651&group_id=235 |