For those who missed the old discussion the main problem is that once the DbCache is full, the DmQueryRecord fails frequently (crash with Free handle or Invalid chunk ptr message, crash without any message or - mostly - dmErrMemError). Our experiments showed the bug frequency above 0.1%. The funny thing was that in case of dmErrMemError 1-20 retries always helped.
The Tungsten E2 is running NVFS code that's very similar to what is on the updated Sprint Treo 650. I think there may be a couple of bug fixes in the E2 that weren't in the Treo, but they weren't major changes.
I'm not sure if your issue is fixed. I have had extensive discussions this week with the engineering teams that have been working on NVFS, including getting access to their extensive issue list. Most of the issues on their "known NVFS bugs since launch" list have been resolved in the builds used in the E2 and updated 650; some issues with exhausting the DBCache require more extensive code changes which are currently being implemented.
At a fundamental level, the way that records are managed in the DBCache will be changing from one where caching is done at the record level to where it's done at the MemHandle level.
I plan on having more definitive information for my session at the DevCon next month. I've got a lot of internal documents to go through, presentations to build, and material to get reviewed by both palmOne and PalmSource review.
-- Ben Combee, Senior Software Engineer, palmOne, Inc. "Combee on Palm OS" weblog: http://palmos.combee.net/ Developer Forum Archives: http://news.palmos.com/read/all_forums/
-- For information on using the Palm Developer Forums, or to unsubscribe, please see http://www.palmos.com/dev/support/forums/
