6. Consider other automated test types
At the Lawst meeting, mainly focused on the GUI level, so this article is mainly written in this regard. Before the meeting, we participated in the meeting mainly described our experience in testing automation. Some people have a vivid successful report. A larger success is to cooperate extensive cooperation with people who write test programs. In these stories, this type of test tool has been used in a variety of cases, which reflects the benefits obtained from different test tools.
There are a lot of scams, expectations and hopes to apply GUI-based recession test. They can fantasize the test override without a separate tool, which causes a series of problems that can put its energy more than the design and maintenance of more bugs.
These test tools are available, but they need meaningful investments, detailed plans, reliable training, and careful warnings.
appendix:
About discussion scope
In the last three molecule one, we wrote some views on the white version, and the issues that cannot be unified adopted by voting. The goal of this is to achieve every point of view that meets experienced testers. In some cases, we don't need to vote, neither because of lack of special vote experience, not because we think this framework has problems (we skip a lot of such ideas and views).
If you try to train and perform automation costs and risks, vote is to solve this discussion best data.
General principle
These views cannot be changed. In the automation plan, many efforts are needed, and what problems need to be solved by clear understanding of the awareness. (Unanimously agree)
GUI test automation is an important software development, and his success needs to be based on systems, standards and regulations. General principles apply to software design and implementation applicable to automation design and implementation (unanimously consent)
For practical and maintenance, we first need to develop an architecture that contains many variables to handle changes in the software. We should develop GUI-based automation to meet stability characteristics (unanimously consent)
Some people have a hunch that should change with the company's automation. General resolution (7 agreement 1 No), more automation achievements will develop in the absence of automation experience.
Fall playback failed. Whether we have a small amount or form component (the simplicity of the object) becomes irrelevant.
Failed to apply individual programming tests (individual test case code does not comply with General Standard Nuclear establishment sharing libraries)
It is necessary to develop shared libraries. These libraries may contain recorded scripts or data-driven test scripts.
The second general resolution (10 passes, 1 disagree): General automation fail is the reason:
Establish a test case to use the principle of pocked return visit.
Application has no standard test permanence (such as the general standard in the test sample code, no sharing library)
Apply design defective frame. This is a general problem
The bug rate discovered directly in the multiplicity of direct playback is very low (unanimously passed)
Once the program passes the test, it does not mean that the future test can be executed again without failure. These leaders have attracted some views (no people vote): Automated test is very dangerous because he gave us a false fuzzy program can't fail, even if the program will not run today, it is not to say that tomorrow will continue to succeed. There are many reasons why the procedure failed. But you don't need to find bugs, unless you think so.
BUG 60% -80% of the automated test is found in the development test. Unless you create and run the test from the start of the automation tool, more questions are found in manual testing (unanimously pass).
More test teams are usually not used in the start of automation tools. In the traditional case, manual test is started, and he can join him in the automation kit after the program test passes. In any case, if you decide not to rely on the previously gainful output, you will use the test tool for the program test success or test failure. For example: running different system versions or configures the same series of tests. You may never test programs in a special environment, but you know how to make him run.
Run the same test. In this case, you run both programs and enter the same sequence. If the result of the program is the same, the program test is successful.
The code is tested, so some series of programs run log catalog, such as abnormal state, state transition, memory management, heap space, or other exceptions, or additional program errors. Test tools are random drivers through a lot of status transitions, and record the commands that the test tool can run. The next day, testers and developers are tracked by looking for BUG and triggering their environments. This is a simple simulation example. If you are working in a environment in a development team, then the test case you build using the test tool will be more widely used and effective than the script that is created by manually (the team can find new bugs per week).
When you develop Hooks, interfaces, and debug output together, the automated test will be more successful. (Unanimously pass)
How to collaborate: do not rely on GUI automation test tools, simple use tools - facilitating test drivers, without simple GUI decline test examples. The form provided at the Lawst meeting is fascinating and heard a story about the success of the automation. In most cases, most of the development team collaborates, most of them have achieved dramatic success, excluding traditional (any use) based on GUI-based recession test.
We may study the writing relationship between design and development in the later period of Lawst.
Most code is a script that is unmatched and has no long-term values by tools. In any case, write test cases with test tools because it can record a series of recent events, it is useful. The script created by the recording tool can be given some tips when you write your own code. (Unanimously pass)
We can't use the screen area to record everything, because it was wasted at that time. (In fact, we don't use the screen area and use them when you need it. We found problems in comparison screens. Sometimes we have to compare the screen area, maybe because of the area of the area we test, it is too big. We should compare the local end, not BMP pictures). (Unanimously agree)
Do not lose test automation verification points. Adding a rush in the script is much easier than looking for bugs. (Unanimously pass)
Test design
Automated simple scripts are not correct policies. (Unanimously pass)
If you start from the establishment of a lot of simple test cases, then you have finished using a powerful test case. In a large number of examples, the test cases that are easy to pass can look more rigorous than manual tests, but an experienced hand-type tester may run a complex test to be more stable than the program.
Group and tests can find new bugs (summary than partial points) (unanimously pass)
