Principles in software debugging

Principles in software debugging

• 1.
  Principles in SoftwareDebugging Udacity CS259 Software Debugging class summary https://www.udacity.com/course/cs259 +PipatMethavanitpong @fulcronz27
• 2.
  Be Scientific CausesEffect 1. Reason an outcome 2. Make a hypothesis 3. Test the theory 4. Repeat 1. if not true
• 3.
  3 States ofError Defect Error in code •Sloppy writing •Corner cases Infection Error in program states •Step upon defect code •Carry undesired states Failure Error in execution •Infections become severe •Cannot maintain functioning
• 4.
  Program State Timeline Fail 3 States of Error Step on a defect code Chain reaction Some infections are handled More chain reaction Fatal infection happens
• 5.
  Program State Timeline Fail Chain Causes Step on a defect code Chain reaction Some infections are handled More chain reaction Fatal infection happens Then we can itch the right spots 
• 6.
  Chain Causes •Information can come in bulk • Core dump after crashed • Crash report submissions by users • Execution log • Not all information is relevant to error • Spend more time • Don’t know what to look for • Looking for dependencies • Control dependency – A statement that depends on a decision • If-else / Switch / For-loop / etc. • Data dependency – A statement that depends on data • Conditional Expression • Chain these dependencies together • Backward slice – All statements that influence a statement • Forward slice – All statements that depend on a statement S0 S1 S2 S3
• 7.
  Debugging Techniques •Assertion • An if-else like statement e.g. assert x == 9 • Use to confirm a state / Narrow down possiblities • Included in debug mode • May be removed in build mode • Code Coverage • Record which lines are executed • Tracing • Record how a program runs • Granularity of what are recorded depends on implementation • Python has these trace events fired: “call”, “return”, “line”, etc. • https://docs.python.org/2.7/library/sys.html?highlight=settrace#sys.settrace
• 8.
  Picking the Relevance • Delta Debugging • Given a set of variables’ value, a program fails • All variables may not involve with the failure • Find the smallest subset that can fail the program • To use these values • Initial values • Injected in an interested program state • Phi Scoring • Pair code coverage and program’s outcomes (pass or fail) • Obtain statistics of 4 values : n11, n10, n01, n00 Fail Pass Covered n11 n10 • Compute phi value from the equation 푛11∙푛00−푛01∙푛10 푅표푤1∙푅표푤2∙퐶표푙푢푚푛1∙퐶표푙푢푚푛2 • High value = More relevant Row1 = n11+n10 Not covered n01 n00 Row2 =n01+n00 Column 1 = n11+n01 Column 2 = n10+n00
• 9.
  Managing Bugs ina Project Removing bugs is important Managing removing bugs is also important
• 10.
  Problem Life Cycle Unconfirmed •A bug report is submitted New •The bug is confirmed as new by a committee Assigned •Responsible developers are assigned to fix Resolved •The developers say the bug is fixed Verified •The committee confirms the fix Closed •The bug is officially dead • Record a bug state properly • Fix in timely manner • Skill matching
• 11.
  Defect Map Bug Database Version Database Bug ID Bug problem Bug detail Revision ID Revised code Revision meta e.g. fixed bug ID Give a relationship between code and a bug
• 12.
  Defect Map (2) Bug Database Version Database Bug ID Bug problem Bug detail Revision ID Revised code Revision meta e.g. fixed bug ID Components Directories Files Get statistics of which are likely problematic
• 13.
  This does notcover the whole story If you are interested, please make a visit to the course website https://www.udacity.com/course/cs259 Have a good day