Lenny

Network of states and transitions between them consists of four characteristics of bonds and general communication, not to be questioned. Each link connects the four characteristics of state-a starting point, the state-endpoint, the fact that the right of the command transfer and the need to implement this transition. Example of work - click on the button, we decide that you need to click, first, the application looked one way, then we expect that it will look the other way, provided that a click has been executed correctly. Characteristics of the correctness of execution - is the presence of an internal error in the performance may object, on which there is no need to do the surgery. These links are targeted. Recalculation of bonds is only one step. Relationship is not to be questioned binds all the states and controls that the sum of the probabilities of all states is unity. This connection works with a network of lower level before running the command and then execute the command. 

The network, which makes the decision about what operation to produce - is a network consisting of three links and the characteristics of a connection not to be questioned. Relationship is not to be questioned, binds all the characteristics - the need to perform some operation and ensures that the sum of the probabilities of all these characteristics is unity. Connection of three characteristics associated generalized characteristics of the sub, or characteristic of a successful test, the characteristics - the need to perform some operation and characterization of the state. 

The network, which takes into account the purpose of testing consists of links between the characteristics - sub-goals and a description - a successful test. Characteristics of sub-goals are the presence or absence of errors in the application. Connection between the characteristic-error - it non-directional communication between two double-digit performance and imply the presence or absence of this error now. If the purpose of testing achieved 100%, further on these connections, you can synchronize the system to record defects. 

Recalculation of the network begins with the host-target. The purpose of conversion - Definition percent goal. If the goal is achieved in full, it means that the application is consistent with our understanding of it. The advantage of this approach is that this feature does not mean that the application does not contain any errors. Any program in the development process usually has an error. The task of testing - consideration of these errors, which is why achieving the same response decreases when new bugs and fix old ones. The decrease of this characteristic - an indicator for the tester has a job. 

To calculate this characteristic is necessary to check its associated operations and sub-goals. Recalculation of the network determines which sub-goals, defects, you must check first, what condition for this to be achieved in the first place, what actions it requires to produce in the first place. The result of this conversion are the characteristics of what an operation must be performed in the first place. 

To calculate what operation to perform first the network is necessary to calculate, and hence to determine in what state is currently under application. The decision about what operation to perform first direct review pursuant to the objectives after each operation over the application. The basis for the recalculation is new information that the application was transferred to a new state. Problem of determining the status of the application activates the function of signs that the application in a particular state. 

The result of the operations of the application is to determine the state of sub-goals. Determining the status of sub-means that has been made a test phase, and although further operations may challenge the result, after a sufficient number of operations on the application, the characteristics of sub-goals will vary less. Determining the status of the primary goals of testing leads to the end of testing. 

Assume, for verification of some functions of the program to check a certain operation, but this operation only works in 80% of cases. Suppose the first test showed that the function works on the same operation has caused another subgoal for their review and action did not occur. In this case, the success of the first sub-contested. Based on this decision, whether peretestirovat previous subgoal, as confidence in the operation efficiency is reduced. If this affects the set of already computed sub-goals, plans for testing will change. If confidence in the operation due to a detected performance or malfunction does not change the overall picture because of the already accumulated a large amount of statistics for this operation, or for Retesting need to do much work, then check the current sub-goal will continue. 

Consequences of the new architecture of the system testing. 

Retesting is not endless testing, because an additional operation of the application has been steadily increasing statistics, so the new test changes the prevailing state of the network is less and less. The system produces a continuous assessment of appropriateness of certain actions, evaluating the whole situation, reducing the total time of testing. 

Test result is to reduce the likelihood that the goal of testing is reached, communication emanating from this feature in a chart shows the reasons for reducing this probability. The resultant effect, a report on testing was not reached on the basis of specialized analysis program, but on the basis of a specific description of the problem area, test the program. Completeness of the coverage tests warrant this description. 

The presence of an intermediate layer between the sub-goal of testing and testing operations allow for the possible errors in the test program, describe the attainability of certain states, and hence the completeness of testing under the condition of defects. 

Test results of any function of the program is presented in the form of statistics, its running and not divided on separate, unrelated tests. There is an opportunity to compare different versions of the application through the comparison of statistics on the test functions. Redundant way of describing the interface provides the reliability of testing, even a partial change in the interface design, or program functions. For example, a function of the program was run differently, the system is capable of testing to determine the unexpected state of the application and continue testing, adjust subsequent operations under the condition found specificity. 

The disadvantage of this architecture test system is that to implement the test is necessary to implement the related goals require redundancy description states, the complexity of converting the network compared with the performance test program. 

The advantage of this architecture is the ability to synthesize knowledge, recorded with a code generator, comparing the recorded actions and observable application state, in the process of writing new tests, and as a result of self-implementation. 

Ability to effectively use the code generator makes it possible to reduce the complexity of test generation. Create new tools for making tests, where the tester does not require programming skills, and to answer the logical questions that arise during testing. 

Questions to the tester can be varied. Correct recognition interface, for example, found that the new controls, which did not exist, or the absence of some expected signs. It should be noted that the correct description of the controls can be checked by the system, and analyze the reasons for the presence or absence of these elements is the task of the tester. Changing some features of the program, the stability of their work can be evaluated testing system, and the causes of this instability allotted to man. The tester determines the purpose of testing, registration errors in the program, the definition of error, in turn, the logic of the system reduces to the determination of controversies in the description, in the presence of contradictions in terms. 

A person can run in real time during testing, analyzing the intermediate results, responding to the controversy, which found the system. On the other hand the system is able to independently resolve conflicts, if the work program does not change drastically, reducing the complexity of testing.