From: Dima K <di...@ya...> - 2003-01-05 13:11:59
|
--- Jon Keating <jo...@li...> wrote: > > I want to make a new release of Licq shortly. Many > bugs have been fixed, lots > of memory leaks has been addressed, and new features > have been added. In > otherwords a release is overdue. If there are any > outstanding CRITICAL bugs, > please let me know about them in the next day or > two. > > Jon Hi, Now that you mention it, there are a few accesses to uninitialized/deallocated variables (i've been running valgrind and fixing them 1 by 1). The 2 major ones though are the fact that ProcessRunningEvent_Server_tep doesn't really guarantee the events are sent in order like it claims to, the thread can easily pick an event and then sleep while the next thread picks the next event and sends it ahead of that 1st one. The 2nd one is also in ProcessRunningEvent_Server_tep and ProcessRunningEvent_Client_tep and has to do with cancelling an event. This can be best demonstrated if you try to send a message directly to a user behind a firewall, the thread will block trying to connect to the user.. if you cancel the event at this point, the event instance will be deleted, and a cancel signal sent to the thread, but the thread does not act on the cancel signal since cancelation has been disabled.. Once the connection fails, the now deleted event is dereferenced by the thread and possibly even deleted (again) in ProcessDoneEvent I am working on fixing those issues btw, but i don't really have an ETA yet. PS. sorry i sent it to your private mail Jon.. i meant to send it to the ml __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com |