Software Surprises

No program is perfect.

Even if one were, you would not understand it perfectly, so sometimes it would surprise you.

Pleasant surprises require no special treatment. It will be appreciated, however, if you take a moment to thank the programmer(s) for their hard work on your behalf when you encounter such a surprise.

When a program unleashes an unpleasant surprise on you, the first thing to do is stop.

If the surprise is not pleasant, your instinct will be to do something - try the action again to see if it works this time, or restart the program, or the computer.

If you do any of those things, you may lose a bug to the mists of time by destroying the only evidence anyone has ever had that a bug exists.

Once you have prevented yourself from damaging the computer's state, write down the sequence of events leading up to the surprise, ending with the surprise itself and an explanation of why you found it surprising.

Your audience for this explanation is someone who has never used the program before. Writing for them will push you towards the specifics that are crucial for reproducing tricky bugs.

Once you have written the explanation, you may find you are no longer surprised because you have gained insight into the program's behavior. Sometimes such an insight will come before the explanation is even finished.

If the insight is "It works as intended but I misunderstood the program", then you should mention the incident to the programmer, so they know the interface was not immediately obvious to you. Other people may have had similar problems, and if enough have, the interface or manual should be improved. In this case, you have no other responsibilities.

If instead the insight is about the nature of the problem, or a theory about why the surprise happened, write it down.

Next collect evidence the surprise happened. A screenshot is a good place to start (especially for graphical programs), as are log files or (for a batch command-line program) a copy of the terminal session.

Once you have collected all the evidence you can, decide whether the surprise is a problem or not. Check program documentation and issue database to see if your situation has been encountered before, and whether there is a workaround or a fix for the issue available.

If you are not sure whether the surprise is a problem, create an issue report to find out what the programmers think.

If the project documentation has instructions on bug or issue reports, follow them. Some programs have built-in bug reporting functionality, which can make your job simpler.

The first step in writing a bug report is to reproduce the problem. Your explanation of the surprise should guide you in the right direction, and can be refined to the minimum steps needed to reproduce the issue.

If there is a program available to record your screen, use it while you try to reproduce the bug. If you capture the bug on video, include that video as part of your bug report.

If you cannot reproduce the bug, the bug may be one that's hard to reproduce. Simply use your explanation and evidence for the report, and note that you have not been able to reproduce it.

The highest form of reproduction is a program that demonstrates the defect when run. It is not always feasible to create one, but for command-line programs it often is. If your initial reproduction attempt is successful, and you can see how to quickly write a program to reproduce the surprise, it may be worth doing so and including it in your bug report.

Alongside your reproduction steps, include any additional information that may be useful. Some things that are almost always helpful are the program's version number and the operating system you ran it on. Many other things could be relevant depending on the context - trust your instincts and lean slightly towards too much information rather than too little.

Be prepared for the programmers to ask further questions - they will often want to know things you did not think to include. If you can, answer them.

Other Articles In 'workflow'

  1. Code Review

    First drafts are wrong, but the author can't see where.