Error reporting component

AlexStumm · Sep 22nd 2009

Hi,

sometimes (very rare) we get application crashes on our NetDCU6/8 boards and then customers become unsure and ask what happened (because the error message from WCE is not very useful).

To get an idea about the reason it would be helpful to use something like Dr.Watson logs and dumps, which can be added to the OS image as I found on different WCE sites.
This is not a major issue, but is it perhaps possible to add this feature later as an update package ? Or are there other similar methods already included ?

Thanks
Alexander Stumm

fs-support_ZU · Sep 23rd 2009

Hello,
yes it is possible to add "Error Reporting" (by kernel modification). But think about the required configuration especially in a customers enviroment!
Following may offer a better way to detect your problems (but presumes that errors occures in your testing phase):
1. Create a MAP File for your application (under VisualStudio200x goto: ProjectSettings -> Linker -> Debugging -> GenerateMapFile -> Yes).
2. Connect a terminal prg. to the serial debug port (38400,N,8,1).
3. In case of error you will get DLL name which causes exception and offset address.
4. This offsetaddresse you will finde also in the MAP File.
5. Now you have detailed information about the call which causes the error.
Please refer also MSDN for further details.

AlexStumm · Sep 24th 2009

Thanks...well I experimented a little with map files, but found it not sufficient in our case: it finds me possibly the function which causes a crash, but not the exact line number. Also it can't provide call stack information to help localizing the real error source.

As I mentioned the problems are very rare and we cannot reproduce the behaviour in a test environment. So how complicated would it be actually to do a kernel modification ? Do I have to exchange the whole image or is there an option to upgrade boards "in the field" without losing application data ?

fs-support_ZU · Sep 24th 2009

Hello,
you can not add "error reporting component" without kernel update.

I suggest use:

- map file methode, you will get the desired details (refer http://geekswithblogs.net/Bruc…e-of-a-data-abort-an.aspx and http://geekswithblogs.net/Bruc…ause-of-a-data-abort.aspx). Some platforms offer the possibility to redirect debug line to a file (refer registry under HKLM\Drivers\BuiltIn\mbd).

- several try-catch-blocks with your "own" message box.

You can display desired data which customer can send you for evaluation.

AlexStumm · Sep 28th 2009

Thanks again...but somehow I still cannot figure out the exact crash cause.

According to the map file it happens at appui3.cpp line 138. Is it possible for you to provide .cod files for NetDCU6 MIPSII MFC ?

fs-support_DK · Sep 29th 2009

Please post the debug output on serial line where the exceptions reports the return address (RA) and the program counter (PC).

AlexStumm · Sep 29th 2009

As I said this is not reproducible here...all I have is an image from the customer which shows the "fatal application error" box from Windows (exception C0000005 at address ...)

medele · Dec 8th 2009

Quote

Following may offer a better way to detect your problems (but presumes that errors occures in your testing phase):
1. Create a MAP File for your application (under VisualStudio200x goto: ProjectSettings -> Linker -> Debugging -> GenerateMapFile -> Yes).
2. Connect a terminal prg. to the serial debug port (38400,N,8,1).
3. In case of error you will get DLL name which causes exception and offset address.
4. This offsetaddresse you will finde also in the MAP File.
5. Now you have detailed information about the call which causes the error.
Please refer also MSDN for further details.

Display More

How can I do this procedure in Visual Studio 2005 an C# ?

How can I find the wrong code for this error message:
Ausnahmecode : 0xc0000005
Ausnahmeadresse: 0x03fc2c40
Lesen: 0x002df928

Share