Steuart Dewar on T5 and Databk5

This is what Steuart Dewar had to say about the T5 and Datebk5:

From: "C. E. Steuart Dewar" <dewar@P...>
Date: Wed Oct 20, 2004 5:31 pm
Subject: Re: [pdbk] Re: T5 and Datebk5

It's pretty clear the T5 was never tested much (if at all) with any third party apps. I've found a really horrible flaw when running DateBK5 on the T5: just try tapping around on blank timeslots, and you'll get a spurious "VFSDBCache.c, Line:6761, DeleteRecord() called on invalid record" crash.

Then again, I tried running their "Favorites" application in the simulator (this with the official release ROM) and get a "Odd aligned read from 60434345" error. So who knows? Maybe the simulator is just flaky? Maybe the device is flaky??

Unfortunately, the T5 simulator is so slow (The Tungsten T5 Palm OS takes over 3 minutes just to boot up on my 3.4Ghz Pentium4) that I can't look at most problems - the Code Warrior debugger refuses to connect because it just times out the device. I tried running the gremlins automated tester and the simulator croaked after only executing 20 operations (!).

I have heard of this long delay when exiting Datebk5, but it doesn't seem like everyone is reporting it. I have a bad feeling that this device is going to be a real dog as it sure seems to have an awful lot of problems.

And they have some pretty horrible problems in their own code too as Theseus mentioned - just set the built-calendar to use the monthly startup view as a default and you're completely hosed (you'll have to kill out the SAVED PREFERENCES database or do a hard reset and not copy that over to recover). BTW - there is obviously no parallel problem in DateBk5 - this was just something they never tested with their own Datebook app. I'm concerned the Treo650 is going to have all these same problems as well :( - may not sit that well with Treo600 users who've gotten used to my Datebook+ application).

As for the delay, I don't think it would be the flash memory unless they put the mirrored PalmSource databases in flash memory. But as near as I can tell, that is not the case: the PalmSource database is in the main RAM - at least when I look in the "internal drive" there's nothing listed there and I see DatebookDB in the normal RAM area where you would expect it to be. However, the VFS error is interesting (why would VFS be invoked here when there are no VFS databases being accessed? Did the Palm OS just go beserk and crash into the VFS routines? Or are they doing some really fiddling in the DM Patch and just screwed up implementing something?

About the only thing I can suggest for now is for some enterprising user here with a T5 to remove the Datamanager patch and see if all the performance problems disappear - that would be interesting.

I guess it's another $400 down the drain for yet another useless (for me) device to try and figure out what new and interesting bugs PalmOne's introduced into this device since their desktop simulator's performance makes it of only academic interest - it's completely useless for debugging (it takes a full 3 1/2 minutes just to boot up the Palm OS and runs about 5-10 times slower than the simulator for other Tungsten devices :(

Just performing these few tests on the T5 simulator has taken over three hours (most of it just staring at the simulator going through it's long bootup process).

Cheers!
CESD,
Pimlico Software, Inc.