The Research And Implementation Of Embedded Software Compilation Test System Based On Naive Bayes Classification

With the increasingly popular of embedded system,more and more electronic devices or products are equipped with embedded chips,which has broad market prospects.Most of semiconductor manufacturer continuous launching new Microcontroller Unit(MCU)based on ARM core to win market share.The semiconductor manufacturer not only make and sell their MCU chips,but also provide the peripheral drivers as well as the example applications for the series of these chips,the software is called Software Development Kit,or SDK for short.In SDK,the peripheral drivers and example applications are designed as common module,they can be reused in different compilers and MCU products.The advantage is that it can reduce the effort of development and maintenance,but the disadvantage is that there are many similar compilation errors during build testing.It brings great challenges to embedded software compilation testing.In this thesis,it based on Naive Bayes classification algorithm,open source Natural Language Toolkit(NLTK)and Software Test Automation Framework(STAF),to design and implement an embedded compilation test automation system.It can analyze and classify the key error message in error logs automatically to greatly improve the efficiency of testing and reduce the cost of manpower and time.The achievement of this paper as follows:(1)Design and implement naive Bayes classifier in order to analyze the compilation error logs,support to extract the key error information from the error logs in practical test by training the collected build log data.(2)Design and implement the background automate compiling for the popular embedded IDEs IAR,Keil,ARMGCC and MCUXpresso,and implement the distributed parallel compiling system based on Software Test Automation Framework(STAF).(3)Design and implement the functionalities to query and browser the build result based on the Django Web framework,which can provide a friendly user interface to show the build results.