D5 deployed apps - early observations

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

D5 deployed apps - early observations

Bill Schwab-2
Hello all,

It was a long story, but my D5 rollout happened only a couple of weeks ago.
On balance the deployed apps are MUCH smaller, and the installation was
remarkably unremarkable :)   That was a pleasant surprise because I changed
a lot more than I would ordinarily be willing to do in one shot; events
outside my control gave me little choice.  It worked very well though.

Here are two things that have changed, one might point to a memory leak, and
the other might be explained by improvements to D5.

Possible leak
=========
One app is showing signs of a memory leak (the machines seem to start
swapping after several days) that I don't recall seeing with the D4 version.
This app runs 7x24 on ME boxes which have a surprisingly good service record
considering the OS.  Anybody else seeing anything that might point to a
leak, maybe a stray #detach???  Obviously, it could be my fault, and I'll
proceed under that assumption.  Any favorite debugging tools for such a
thing?  I have another set of apps that run on win2k and (based on D4
experience) should run for months between reboots; no signs of stress there
yet, but the machine is faster, has more RAM, lots more disk space, and
obviously a much more reliable OS.


Printing/shutdown problem seems to be gone
==============================
On the positive side, I had a problem with a different D4 app that, once
moved to win2k, would "randomly" fail to shut down completely.  All views
would close, but the OS process was still running.  The app is set to run
only one instance, so it would sometimes see its former self and refuse to
run until one either rebooted the machine, or clobbered the old process via
the task manager.  FWIW, that latter step didn't always work, at least not
immediately, and win2k would often complain that "the operation does not
apply to this process" (or something like that), and then kill it anyway.

With that said, I have yet to see this shutdown problem with the D5 app.
There is a chance of under-reporting by users, but I check for it myself,
and have not seen it since installing the D5 version.  The GUI of the app in
question was a big part of why it took me so long to deploy D5 apps, so the
change could be in my code vs. Dolphin.  Still, there is the overlapped call
bug fix in the VM (the bug that caused D4 apps to yield relentlessly to CPU
hogs like Access), and no doubt other improvements that could explain it.
FWIW, I suspected that the problem had something to do with printing,
especially when it didn't go terribly smoothly.

Comments anyone?

As an aside re that Byte article, how could they fail to mention something
as slick as Dolphin!?!?!?!?!?!?

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]