Are you handling notifications through your mainline or through a callback function? If the latter, we lost almost a month trying to track down spurious errors until we realized our whole program was being moved when NVFS got full and this was invalidating callbacks we'd previously registered with PalmOS to handle notifications.
We couldn't find a recommended safe mechanism to lock down our own primary (0) code resource, so we ended up moving all the notification handlers in three of our programs either to launch code notifications (where possible) or to separate code resources that we could lock down manually to keep them from moving. PalmSource, (Ben?) can you confirm this behavior or offer a better solution? Also, if PalmOS knows the addresses of all notification handlers that programs have registered, why doesn't it just automatically mark any memory blocks containing registered callbacks as non-movable in the DBCache? -Vince PS: Anybody else confirmed the similar bug where DmCurAppDatabase can return invalid values during Hotsync start/finish notifications on the T5? The bug only seems to occur if more than one app is registered for those notifications and it happens to the app with lower priority. Using DmGetNextDatabaseByTypeCreator works around the problem. (this cost us another month of development) ----- Original Message ----- From: "Jan Slodicka" <[EMAIL PROTECTED]> To: "Palm Developer Forum" <[email protected]> Sent: Tuesday, March 15, 2005 11:29 AM Subject: Notifications buggy on NVFS as well? >there are several indications that Palm notification mechanism (that seems to be LocalID-based) can easily fail (i.e. cause a crash) when the DbCache is full. ... -- For information on using the Palm Developer Forums, or to unsubscribe, please see http://www.palmos.com/dev/support/forums/
