Crash coredll with 1 line of code

G

GrahamS

Hi,

More than 6 months ago I tried (unsuccessfuly) to get support for a random
problem where coredll bombed without going into my catch handler. I could not
easily demonstrate this as it seemed to be 'data dependent'. It may even have
nothing to do with the following - apart from the unprocessed exception!

I worked around this at the time by eliminating a specific piece of hardware
(usb->RS485 port), and using instead an external serial to RS485 adapter.

I now have a similar problem once again but this time I can prove it as
follows :

In a Windows CE Winform app add a button and make the button handler do the
following :
try
{
long l1 = 8169325;
MessageBox.Show(string.Format("Convert: {0}", (Single)l1));
}
catch (Exception ex)
{
MessageBox.Show("Exception: " + ex.Message);
}
}

The relevant piece of code is '(Single)l1' - as my app used this in a
different way, the dialog box is simply to demonstrate the BUG.

Running this on my CE device bombs with an Exception code 0xc0000005 in
coredll. This pops up a Quit dialog instead of calling my handler.

Now build and run the very same code under Windows and its fine.

NB I suspect that the actual number may not really matter - although it just
might.

Whilst I can work around the issue by declaring the variable as a Single in
the first place - I am still VERY worried that coredll can even throw up such
a dialog box. My application runs on an embedded box which has no user
interface - let alone a user to press the quit button. So my remote app is
effectively trashed.

Look forward to a review/comment from MS please :).

NB The underlying OS is not built using the latest release, as I already
have builds out there with an nk.bin built a year ago.

Thanks

Graham
 
P

Paul G. Tobey [eMVP]

It sounds like there's a bug in the floating point library. You're saying
that this happens on a CE device running any of the supported processors? Or
just the one you happen to use? What processor is that using?

I'm pretty confident that you won't hear from Microsoft in this group (this
is not a Microsoft-monitored support channel). You can use one of your
Platform Builder support incidents to report the problem directly to
Microsoft and, if they confirm a bug, request a QFE for the operating system
and rebuild the OS with the fix for your device.

Paul T.
 
M

Markus Humm

Hello,

just curious: what modell of USB to RS245 converter?
What do you control on the RS485 side?

Greetings

Markus
 
M

Markus Humm

Paul said:
It sounds like there's a bug in the floating point library. You're saying
that this happens on a CE device running any of the supported processors? Or
just the one you happen to use? What processor is that using?

I'm pretty confident that you won't hear from Microsoft in this group (this
is not a Microsoft-monitored support channel). You can use one of your
Platform Builder support incidents to report the problem directly to
Microsoft and, if they confirm a bug, request a QFE for the operating system
and rebuild the OS with the fix for your device.

If it turns out that it is actually a bug on MS side does he get a
refund for the used incident?

Greetings

Markus
 
P

Paul G. Tobey [eMVP]

Yes, Microsoft doesn't 'charge' if it's confirmed as a bug. I've never used
up my two PB5 incidents, but I've made a number of calls...

Paul T.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top