The woman is difficult to go out of software testing
Author: Liu Lijun, general manager of Beijing Kaidi four-dimensional software
Sending time: 2004.06.04
Source: Software World
With the launch of information technology in different industries in China, people no longer doubt the huge role of software for social operations. However, as people's improvement of the software role expectations, there have been more and more people to transfer their attention to the reliability of the software, so the domestic software testing company or the evaluation center appears like the spring bamboo shoots.
Software test is not universal medicine
In the process of developing software test market, we found such a problem: Many companies believe that software test is really important, so I will put forward: I will execute the program (or have a complete user manual) for you, you Tel to me ("Sorry, the code cannot be given, because of the issue of property rights"); if the test is passed, the user should no longer ask questions. If the end user makes a problem, the company will find a software test company: "Look, how do you engage in, the user puts forward, why can't you find a problem with testing?".
We also encountered this situation, and the software developers and users have repeatedly caused by software quality. So the company found us say: "Since you are software testing hipster, you will do test, as long as the test cost is controlled within 20,000 yuan, we introduce you to your business." In the end we didn't dare to bear such business, Because we are worried that they will fall into the situation in the valence. It can therefore also see that there are some misunderstandings for the understanding of software testing.
For the highest level of software among the aviation industry, in order to protect its reliability, the work content of the test includes grammar rules inspection and program analysis, conditional coverage, boundary coverage, statement branch coverage, demand coverage, strongness, functionality and input and output Test, eventually all passed, and only guarantees the defect probability of 10-9.
Therefore, software testing is an important part of improving software quality and reliability, but does not mean that there is a software test, and software does not have problems. If simulated a simple trial, the verification utility of software quality is worse.
You may wish to do a class ratio, if a project is acceptable, the external decoration is extremely compliant, giving people a good feeling, can we determine this job is a good quality of quality? The actual situation is often, and the golden steel jade outside the tofu. Of course, you open the faucet, open the light bulb, there will be no problem, if there is a fire, big wind, this project is ok? do not know. Why not know? Because it doesn't see if the construction process is in line with the standard; the construction process is qualified, it does not know if the material is qualified.
Therefore, software testing is not a universal medicine that guarantees software reliability.
Software test is layered
If only the user manual, the user acceptance test to be made is simply tested with a proportional special case. Experienced testers are just more scientific systems for test cases, and in addition to add some strong testing and pressure testing. For a safe reliability requirement, it is not enough.
However, we know that our country is currently engaging in e-government projects represented by "12th Gold". Part of these projects involving financial and taxes and high-tech fields such as telecommunications, finance, insurance, aviation, aerospace or high software reliability requirements, their testing of software is not far from ways. As you can simply imagine that the probability of a defect is 10-9, can you meet the requirements by the previous test?
And so strict requirements, the human and financial resources that investment will be very huge. Generally speaking, at least 3 to 5 times the development cost, and requires a very standardized development process.
Overall, we don't agree with the way users simulate testing, because this approach is under system and complete.
I personally think that the acceptance test is to complete the following work: function traversal, link test, interface test, stability test, data interface test, security test, performance test, load test, pressure test, platform test, browser test, strong Sexual test, etc. If the problem is found during the test, you should isolate the problem to the part according to the relevant design documentation. For core modules, such as functional core or primary control parts, the white box test of the module level is required.
Test should be connected to the development process
Many developers or users pay attention to software quality also pay attention to software testing, but because their development process is not standardized, it is often caused by test, especially the module-level black box and white box test difficult to carry out. The reason is very simple, is the lack of detailed design documents and flow charts and interface relationships corresponding to each module code. The results of the results are like the blind touch - only the agent cannot see if the program is consistent with the design ideas, the correctness of the software input and output.
Thus, to perform software testing, especially strict software testing, software development process should not only conform to general specifications; not only, the complete document is complete, and the degree of meticization should be quite high. In order to ensure the smooth development of test effects and regression testing, the configuration management of development processes should also be strictly effective.
"Make bricks without straw". As a professional software test company, we hope that through our efforts through developers and users, improve and improve the process control and development documentation process control and development documents, so that test work can better improve software reliability.
