www.virtualacorn.co.uk/forum

.... and needs to close. We are sorry for the inconvenience."

"The instruction at 00x00b6e3f3 referenced memory at 0xe59ce38e which could not be read."

Any ideas ?

I have now narrowed the problem down to two of my own BBC basic applications. These two programs share data via two files.

- On a RiscPC 600 running RiscOS 3.50, The applications are working perfectly
- On a RiscPC 700 running RiscOS 3.60, the machine locks up eventually
- On a VRPC-SA running RiscOS 4.02, the virtual machine either just closes down or stops as above.
- VRPC-SE does the same thing and using ADFS instead of HostFS make no difference.

I realise the BBC basic programming fault is mine, but does the above help to locate it .....
..... and why does VRPC-SA just shut down?

Yes, I know there are better ways to share data and I am now re-writing the code to use Wimp_TransferBlock.

It would seem from the error that your application is trying to write to memory that it does not own. Beyond that it's difficult to know.

OK thanks. (I had thought the message came from Windows, not RiscOS)

But, I have changed the code and everything is stable again.

Glad you have got it working.

If you still have versions of the app that crash VRPC would you be able to forward them to Aaron so we can take a look? If not, could you produce a minimal app that causes the same effect?

Thanks,
Graeme

OK - give me a day or two.

Sorry. I can not reproduce the close down.

I have managed to make the system lock up - but a real RiscPC locks up just the same .....
..... and is nothing more than a program in a loop without calling Wimp_Poll

At the time VA close downs, I had lots of applications running:
Two of my own written in Basic.
!Igor - with a session connected from another PC
!WebJames - with page requests from yet another PC and from a Real RiscPC 700
!Perl - as it comes with !WebJames
!Deltanet - with just FTP active and connections from a Solaris Workstation.
(all of these communicating with each other using Wimp_TransferBlock and/or shared files)
!VNCserver
SocketFS module
!Printers+

Normally you can escape from a program that's stuck in a loop using ALT + BREAK. This applies to both real and Virtual machines.

Thanks for taking the time to try and reproduce the problem.

I appreciate that this is going to be hard to fix if I can not reproduce it in a controlled manner.

However, yesterday I had VRPC-SA running all the applications listed above with lots of network traffic. It ran for about 10 hours and then VRPC-SA just closed down with no warning.

I know it is a long shot, but two things occur to me. Just before the crash I noticed that a WebJames created application was appearing repeatedly in the RiscOS application tasks listing when the web client was only sending one page request once every minute. This had run for the ten hours without problem and it all also works perfectly when I use a real RiscPC 600/700. Perhaps a network buffer overflow somewhere?

The second long shot is that my application sometimes writes directly to the screen instead of using a window nicely. This is OK on a real RiscPC 600/700 and A410 but maybe the VA Windows Controller gets confused?

(As I was typing this, it has just occurred to me to try all this on a different PC with VRPC-SE)
We will see how it all gets on overnight - OK so far after 50 minutes


Edit: still OK after 10 hours - will post again only if it crashes.

Sorry. VRPC-SE crashes (and closees down) in just the same way.