2017 - Jouella Fabe - Fix Bugs Without Falling Into a Rabbit Hole

How not to fall into a rabbit hole when debugging mysterious bugs in a large open source library.

Almost everyone uses some open source library, but when that library fails to do what you want - i.e. decoding only 2 pages out of a 4 page pdf file or it just straight up raises an exception.

Many would dive deep into the code. In this talk, I will try to convince you *not to do that* by sharing my anecdotal experiences such as finding an elusive unicode character encoding difference between two popular OS's from a large open source project.

Video: https://youtu.be/nAhjroo374I

Transcript

1. Fix bugs without falling into a rabbit hole @jmfabe
2. Outline • Why this talk is relevant • Anecdote on debugging a pdf decoder library • What makes debugging in a large open source project difficult • Steps on how to tackle the beast • Closing remarks
3. Why is this talk relevant?
4. A tale of two bugs
5. First bug Passing a 4 page pdf and only decoding the last 2 pages
7. Second bug Library method raises exception when it calls unichr
9. Why is it hard?
15. Run through a list first
16. Some ideas - Replicate the problem in a different environment or in a different interpreter - Play around with the libraries tests - Write your own unit tests to see what happens
17. Understand how the problem came to be Come up with a list of questions
18. Case study: decoding last 2 pages out of 4 How does the library read the pages?
19. • Try to go through a check-list • Come up with a list of questions • Do some research about the problem • Go back and step through a debugging session • Talk to someone else or take a break
22. • Debugging mysterious bugs can expand your knowledge • Debugging is a skill, exercising it makes you a stronger engineer