The process of defect tracking is an extremely important process in software engineering. This article describes how to use two classic analysis models to control the process of defect tracking. These two models are called "Active bug Trend Map", "Bug Open Close Diagram".
In addition, the article also mentions two concepts: "Bug Convergence", "0 bug Bounce", the specific meaning will be described in the introduction.
First look at a picture, this is the analysis of two models of pictures, integrated in a coordinate inside. Active bug trend is a line, bug open Close is bar chart, X axis is time. Let's talk about the meanings of the two models in detail.
Let's say a few noun explanations:
1. Number of active bugs. The state is not the total number of all bugs that are closed. Active bug refers to the bug that you need to pay attention to in the project, some bug management tools also have invalid, duplicate state, these are not the active bug, but later bug, belong to the active bug.
2. Open the number of bugs and close the number of bugs. Each new 1 bug or reopen a bug, the number of open will be added one. Each close one bug, the number of closures will be added one.
Illustrate these concepts, the above two models are better understood. Each point on the active bug trend curve indicates how many active bugs exist in the software today. The larger the number, the worse the quality of the software. And the bug open close diagram, every day there will be red, blue a total of two pillars, indicating that day open, close the number of bugs, if the day these two numbers are very high, indicating that the handling of the bug is very active, the software is very unstable. Note that the unit of the active bug is "a", and the Open closed unit is "secondary", so we use the line chart and the column chart to represent each.
Let's talk about the usage of the model. The general software testing process, there are 3 stages, from the above diagram can be clearly seen.
Phase 1: Test group to start a thorough test of the system, open the bug is significantly faster than the speed of closing the bug, the number of active bugs rise rapidly, when the completion of all the test case execution, the number of active bugs to reach the maximum;
Phase 2: The development team is fully committed to fixing bugs, testing teams to verify bugs, and a small range of regression tests to verify the surrounding features of the bug. At this point, closing the bug is faster than opening the bug, and the number of active Bugs falls back. When the number of active bugs has just begun to fall, it is called "Bug convergence." Eventually, the active bug drops to a very low position, sometimes hitting a "0 bug", but that doesn't mean the project can be published.
Phase 3: The test group once again performed a complete regression test of the software system. In this process, some bugs are also opened, but the number is very small, which is called "0 bug Bounce." After completing this round of regression, the software really stabilized and went into the release candidate process.
Therefore, we can pass these two models, to check whether the test progress of the project is normal, the quality of the software is stable, the check method is as follows:
- If the second stage has already started, but the active bug continues to rise, does not fall down, the open bug is still very high speed, may be the first phase of the use case execution has not completed, or the development group fixed bug is low;
- If the end of the second phase, the active bug did not fall back to a low level, indicating that a large number of bugs need to be repaired, software quality is low;
- If the third stage, the number of open and close the bug is many, it indicates that the bug activity is frequent, the system stability is poor.
As a result, the normal project test should be that the active bug goes up first, then falls back, and finally a small oscillation in the low position, and the number of turns off is very small. With these two analytical models, we are more confident that we can control the progress of the project.
Reprint: http://www.uml.org.cn/Test/201001147.asp
Two classic analysis models for defect tracking