Business Situation

In February of 2006, ACME Grocers had completed in-house testing of a field-service automation application that allowed their customers to automatically re-order products for their stores.  The application was designed to run on a specific piece of hardware – the Symbol PDT8100.  The application went live and was shipped pre-installed on devices to approximately 80 customers for a first deployment phase. 

Within days of deployment they started getting a high volume of support calls and a vast majority of the support incidents fell into one of two cases. 

In the first, the user would get a NullReferenceException in the application in the middle of a transaction and the application would have to be stopped and restarted.  While this was a nuisance, a customer could still use the application.

The second type of error however, was far more severe.  Customers were reporting that occasionally the application would generate a NativeException and the application would shut down.  On restart data was often lost, and in a few cases the entire mobile database was irreversibly corrupted.  This was a high-impact bug that needed resolution as soon as possible.

After spending a few days attempting to debug the application, ACME Grocers’ development team had been able to put together a test script that would usually reproduce the reported bugs, though it they did not happen 100% of the time.  Even with the ability to reproduce the error, their developers were unable to determine the root cause of the failures so their development manager contacted OpenNETCF Consulting for assistance.

Solution

OpenNETCF Consulting requested that ACME Grocers send the exact hardware that their customers were using in the field, and while the device was in route, they would begin familiarizing themselves with the application code and developing a plan for finding the bugs based on the repro test script.

Within hours of receiving the hardware, OpenNETCF Consulting was also able to reproduce the bug and began triaging the failures.  Within a day the bugs had been identified, the root causes described and multiple mitigation possibilities were presented to ACME Grocers. Additionally general best-practice recommendations were also made based on a general review of their code.

The root cause of both major bugs turned out to be a lack of understanding of how power management works on Windows CE devices and that the implementation of suspend and resume can vary from manufacturer to manufacturer (a separate white paper describing the fundamentals of this bug is available separately from OpenNETCF Consulting).

Multiple mitigations were provided as their implementation and impact depended on the end-user usage patterns as well as unknown variables such as typical and maximum database sizes.

Benefits

With extensive platform knowledge, OpenNETCF was able to resolve the bugs being reported by their end customers within a week and ACME Grocers was able to ship an application update.  This helped reduce the amount of time required by the development team to diagnose the issue and instead they were able concentrate on other projects instead of investigating platform specific issues with Windows Mobile.  The development team had gained invaluable device development knowledge that would be useful for the current project as well as future projects.