KR20190089615A - Bug fixing system and bug fixing method - Google Patents

Abstract

According to one embodiment of the present invention, provided is a bug correction system, which comprises: a search unit for receiving new bug information and searching for past bug information similar to the new bug information from past bug information stored in bug storage; an abstract syntax tree (AST) conversion unit for converting bug correction codes of the past bug information, similar to a bug code of the new bug information, into ASTs, separately; a GP operation unit for generating a new tree by applying genetic programming (GP) to the ASTs, calculating fitness a new tree, and outputting a new tree having fitness equal to or more than a predetermined threshold value; and a correction code output unit for outputting the new tree having fitness equal to or more than a predetermined threshold value as a patch code.

Description

{BUG FIXING SYSTEM AND BUG FIXING METHOD}

This disclosure relates to a genetic programming based bug correction system and a bug correction method.

In recent years, the structure of software installed in a terminal such as a PC or a smart phone or executed in a terminal has become more complex, and the size of the software has been further increased. This increases the number of bug cords that cause errors and malfunctions in the code that makes up the software.

Research is underway to automatically correct these bugs. For example, a method of automatically generating a patch using genetic programming and a method of creating a proper patch by creating a template of a bug pattern and its correcting action through a preliminary investigation are mainly studied.

Among these approaches, the conventional method of utilizing genetic programming acquires information for genetic programming from program code including bug code, so that if there is no information about the correction code of the corresponding bug code, the patch can not be successfully generated . The template-based approach is efficient for bugs with identified patterns by specifying precise correcting procedures, but there is a problem that templates for bugs and their correction patterns must be written in advance.

Embodiments are intended to provide a bug correction system and a bug correction method that reduce software maintenance costs.

Embodiments are also intended to provide a bug correction system and a bug correction method that reduce the amount of debugging work.

In order to achieve the above or other objects, a bug correction system according to an embodiment of the present invention includes a search unit for receiving new bug information and searching past bug information similar to new bug information among past bug information stored in the bug storage, The AST transformer converts the bug correction code of the old bug information similar to the bug code into an abstract syntax tree (AST), and generates a new tree by applying genetic programming (GP) to abstract syntax trees A GP operation unit for calculating a fitness of a new tree and outputting a new tree having a fitness value equal to or higher than a predetermined threshold value and a correction code output unit for outputting a new tree having a fitness value of a predetermined threshold value or more to a patch code.

Upon receiving a new bug report as new bug information, the search unit identifies a new bug report and a corresponding topic received based on the modeled topic, searches past bug reports with the topic identified in the bug repository, The bug correction code of the past bug report similar to the bug code included in the new bug report can be determined by extracting the bug code included in the new bug report and the bug correction code of the searched past bug report and calculating the code block similarity.

The searching unit calculates the following equation (1)

, Where Src1 and Src2 may denote the bug code included in the new bug report and the bug correction code of the past bug report, respectively.

When the search unit receives a new bug code as new bug information, it can search past bug code similar to the new bug code in the bug repository.

The GP operation unit can generate an initial population using abstract syntax trees, and apply genetic programming by selecting two operation subject trees in the initial population.

The GP operation unit performs the intersection operation using the abstract syntax trees, performs the transition operation on the first tree generated by the intersection operation, calculates the fitness of the first tree on which the transition operation is performed, Adding the performed first tree to the initial population, and if the number of trees in the initial population is equal to or greater than a predetermined first number, generation of the initial population can be completed.

When the generation of the initial population is completed, the GP operation unit selects two operation subject trees from the initial population using the fitness value, performs the intersection operation using the two operation subject trees, 2 tree, calculates the fitness of the second tree on which the mutation operation is performed, and outputs the second tree on which the mutation operation has been performed as a patch tree if the calculated fitness is equal to or greater than the threshold value.

If the calculated fitness is smaller than the threshold value, the GP operation unit adds the second tree in which the mutation operation is performed to the population including the initial population, determines whether the number of trees in the population is equal to or greater than a predetermined second number, If the number is greater than or equal to the predetermined second number, it is determined whether the population is the last-generation population, and if the population is not the last-generation population, the next-generation population can be created.

The GP operation unit can generate a next generation population by removing a tree less than a predetermined fitness from the population.

The GP operation unit can set the target nodes in the two operation target trees and then perform the intersection operation by rearranging the subtrees from the target node to each other.

The mutation operation includes (1) a deletion operation of deleting only a target node after deleting a target node in one tree, or deleting all subtrees from the target node, (2) determining a target node in one tree, An insertion operation of copying and inserting some nodes of one tree at a position of a target node, and (3) an exchange operation of determining two target nodes in one tree and then reversing the positions of two target nodes .

According to an exemplary embodiment of the present invention, there is provided a method for correcting a bug, the method comprising: receiving new bug information, searching past bug information similar to new bug information among past bug information stored in the bug storage, Transforming the correction code into an abstract syntax tree (AST), generating a new tree by applying genetic programming (GP) to the abstract syntax trees, calculating a fitness of the new tree , Outputting a new tree having a fitness greater than or equal to a predetermined threshold value, and outputting a new tree having a fitness greater than or equal to a predetermined threshold to the patch code.

The step of searching for past bug information similar to the new bug information may include receiving a new bug report as new bug information, identifying a topic corresponding to the new bug report received based on the modeled topic, Searching a past bug report having a topic identified in the repository, and extracting the bug code included in the new bug report and the bug correction code of the searched past bug report to calculate the code block similarity, Determining a bug correction code of a past bug report similar to the bug code.

을 이용하여 코드 블록 유사도를 계산하고, 여기서 Src1, Src2는 각각 새로운 버그 리포트에 포함된 버그 코드 및 과거 버그 리포트의 버그 정정 코드를 의미하는 것인, 버그 정정 방법.To calculate the code block similarity, Equation (1)

Wherein Src1 and Src2 refer to the bug code included in the new bug report and the bug correction code of the past bug report, respectively.

The step of searching for past bug information similar to the new bug information may include receiving a new bug code as new bug information and searching for a past bug code similar to the new bug code in the bug repository.

According to the embodiments, it is possible to reduce the cost and the work time required for the software maintenance work.

According to the embodiments, it is possible to improve the quality of bug correction.

1 is a schematic block diagram of a bug correction system in accordance with one embodiment.
2 is a flowchart of a bug correction method according to an embodiment.
3 is a flowchart specifically showing an example of a bug code search step of the bug correction method of FIG.
4 is a diagram showing an example of a new bug code, a past bug code, and a correction code of a past bug code.
5 is a diagram showing an example of a new bug code converted into an abstract syntax tree (AST) and a correction code of a past bug code.
FIG. 6 is a flowchart specifically illustrating an example of a genetic programming application step and a fitness evaluation step of the bug correction method of FIG.
7 is a diagram showing an example of a new bug code correcting code.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The present invention may be embodied in many different forms and is not limited to the embodiments described herein.

In order to clearly illustrate the present invention, parts not related to the description are omitted, and the same or similar components are denoted by the same reference numerals throughout the specification.

Also, throughout the specification, when an element is referred to as "including" an element, it is understood that the element may include other elements as well, without departing from the other elements unless specifically stated otherwise.

Hereinafter, a bug correction system and a bug correction method according to an embodiment will be described with reference to the drawings.

1 is a schematic block diagram of a bug correction system in accordance with one embodiment.

Referring to FIG. 1, a bug correction system 100 according to an exemplary embodiment of the present invention includes a repository 110, a search unit 120, an AST transformation unit 130, a GP operation unit 140, And a code output unit 140.

The storage 110 stores information on bugs that have occurred in the past. For example, the storage 110 may store information on bug codes that have occurred in the past, bug report information on bug codes, information on developers who have corrected the bug reports, correction history of developers who have corrected bug reports, Information may be stored.

The bug report stored in the storage 110 basically records an error present in the program source code file in various fields. The bug report includes a report of the bug report, the current status, the severity of the current bug report, Ranking, and so on.

The search unit 120 receives information 10 about a new bug (for example, a bug code or a bug report), and retrieves past bug information similar to the new bug information 10 . ≪ / RTI > The search unit 120 may transmit the new bug information 10 and similar past bug information to the AST conversion unit 130.

The AST conversion unit 130 converts the received codes into an abstract syntax tree. Through this process, the codes can be transformed into a tree form in which each node has information. The abstract syntax tree will be described later with reference to FIG. 2 and FIG.

The GP operation unit 140 applies genetic programming to the transformed abstract syntax trees, evaluates the fitness of the patches generated by genetic programming, and outputs a patch having a fitness value equal to or greater than a predetermined threshold value. Genetic programming is performed by repeating initial population generation, selection, crossover, and mutation operations, which will be described later with reference to FIGS. 6 and 7. FIG.

The correction code output unit 150 converts a patch having a fitness value of a predetermined threshold value or more into a patch code 20 and outputs the patch code.

Next, a bug correction method of the bug correction system 100 will be described with reference to FIG.

2 is a flowchart of a bug correction method according to an embodiment.

Referring to FIG. 2, the search unit 120 searches past bug information similar to new bug information 10 among past bug information stored in the storage 110 (S110).

For example, when a new bug report 10 is received, the search unit 120 searches for a past bug report corresponding to a new bug report and searches for similar past bug information. This will be described together with reference to FIG.

3 is a flowchart specifically showing an example of a bug code search step of the bug correction method of FIG.

The search unit 120 identifies a new bug report 10 and a corresponding topic received based on the modeled topic (S112). For example, the search unit 120 identifies a topic corresponding to a new bug report, using the frequency of the topic term appearing in the received new bug report. The search unit 120 searches for a past bug report having the topic identified in the storage 110 (S114).

Next, the search unit 120 extracts the bug code included in the new bug report 10, extracts the bug correction code included in the found bug report, and calculates and determines the code block similarity (S116).

The code block similarity can be calculated by the following Equation (1).

Here, Src1 and Src2 represent two source code blocks. The result is a pair of scores, where (1.0, .221) means that the first code block contains 100% of the second code block.

The search unit 120 determines a bug correction code having a code block similarity degree that is equal to or larger than a threshold value that is extracted from the new bug report 10 as a similar past bug code (S118), and transmits the bug correction code to the AST conversion unit 130 .

In another example, when a new bug code 10 is received, the search unit 120 searches for a past bug code similar to the new bug code 10 in the storage 110. For example, the search unit 120 may be implemented by T. Kamiya et al. (IEEE Transactions on Software Engineering, pp. 654-670, 2002), which is proposed by K. Kiyoung, S. Kusumoto and K. Inoue, "CCFinder: A Multilingual Token-Based Code Clone Detection System for Large Scale Source Code, Etc. can be used to identify similar past bug codes. Then, the search unit 120 extracts a bug correction code of a similar past bug code and transmits the received new bug code 10 and the extracted bug correction code to the AST conversion unit 130.

Referring to FIG. 4, the exploration of the bug code in step S100 and the correction code extraction of the bug code will be described.

4 is a diagram showing an example of a new bug code, a past bug code, and a correction code of a past bug code.

The code of FIG. 4 (a) represents a new bug code. The new bug code constructs a program that takes a string of strings as input and prints the corresponding checksum value. The cause of the defect in the code is the 7th line and the 14th line. However, line 7 is a line that can cause run-time errors due to uninitialized lines. The 14th line that affects the output of the actual task case is the fault, since the 14th line adds the wrong number (22). That is, the 14th line corresponds to the new bug syntax (B1).

4B shows a pseudo-past bug code searched by the search unit 120. The pseudo-past bug code shown in FIG. Similar past bug codes include the use of the while statement, the point where the scanf () function and formula are contained in the while statement, the use of the modulo operation and the addition operation (+) in the process of finding the expression, similar. The twentieth line corresponds to the bug syntax (B2).

The code of FIG. 4 (c) shows a correction code corresponding to the pseudo-past bug code. The 21st line is used as the correction phrase C1 of the past bug code as the corrected bug code.

The new bug syntax (B1) does not have an integer constant value corresponding to a space character. Therefore, the bug can not be corrected unless the correction syntax C1 of the similar past bug code is used.

Next, the AST conversion unit 130 converts the received codes into an abstract syntax tree (S120). When codes are translated into abstract syntax trees, it is easy to manipulate node connectivity or information within nodes and can be used as inputs to genetic programming.

The GP operation unit 140 applies genetic programming to the transformed abstract syntax tree (S130) and evaluates fitness (S140). With reference to abstract syntax tree transformations, genetic programming, and fitness evaluation, FIG. 5 will be described together.

5 is a diagram showing an example of a new bug code converted into an abstract syntax tree (AST) and a correction code of a past bug code. FIG. 5 shows an abstract syntax tree for a new bug code and a portion of the past bug code.

As shown in FIG. 5 (a), the components of the new bug code (B1 in FIG. 4) are represented as nodes of the abstract syntax tree.

Similarly, as shown in FIG. 5B, the components of the correction code (C1 in FIG. 4) of the past bug code are expressed as nodes of the abstract syntax tree.

If genetic programming is applied to the abstract syntax tree, the same tree as in Fig. 5 (c) can be obtained. Each operation is applied according to a predetermined probability before execution, and it is repeated several times. The tree can know that the target node N1 of the source tree (Fig. 5 (a)) has been changed to the node N2.

Hereinafter, steps S130 and S140 will be described in detail with reference to FIG.

FIG. 6 is a flowchart specifically illustrating an example of a genetic programming application step and a fitness evaluation step of the bug correction method of FIG.

First, the GP operation unit 140 selects two operation subject trees (S131). The GP operation unit 140 can initially select a new bug code tree and a corrected code tree of the past bug code as the operation target tree.

The GP operation unit 140 performs a cross operation using the two operation subject trees (S132). Specifically, the GP operation unit 140 sets the target nodes in the two operation target trees, and then changes the subtrees from the target node to each other.

At this time, the GP operation unit 140 tracks from the parent node whether the two nodes are interchangeable node types. One of the two trees that is the parent in the intersecting operation can be exchanged only when it is determined that only the node judged to be a bug exists and another tree is confirmed to be the node type exchangeable with the former for all the nodes existing in the tree. This allows the trees of the two codes to blend into one another and extract them from the source code, turning some of the program's source code back and forth.

That is, in step S132, the GP computation unit 140 can generate the first tree in which a part of the new bug code tree is exchanged as a part of the corrected code tree.

The GP operation unit 140 performs a shift operation on the first tree in which the intersection operation is performed (S133).

For example, the GP operation unit 140 may delete the target node, add a new child node to the target node, or exchange information of two nodes in the target tree except for the parent-child node information, do.

In addition, when the comparison operator used in the conditional statement is selected as the target node, the GP operation unit 140 modifies the comparison operator to correct a defect that the conditional statement may have.

Specifically, the mutation operation can include three operations as follows.

(1) The delete operation is an operation for determining a target node in one tree, and then deleting only the target node or deleting all subtrees from the target node. This can be expected to eliminate the source code that causes the bug.

(2) The insert operation determines a target node in one tree, and then copies and inserts some nodes of the tree at that position.

(3) The exchange operation reverses the position of two target nodes after determining two target nodes in one tree.

The GP operation unit 140 can perform the intersection operation and the disparity operation according to the predetermined probability, and the probability value can be adjusted. For example, the crossover is performed with a probability of 0.95, and the crossover can be performed with a probability of 0.20.

Next, the GP calculation unit 140 calculates the fitness (S134). The newly generated tree through the genetic programming operation corresponds to one proposed patch program.

In order to calculate the fitness, the test is performed through a predefined test case, and the higher the number of passes through each test case, the higher the fitness is considered.

For example, the GP operation unit 140 performs testing through a positive test case and a negative test case.

A positive test case is a test case for checking whether a program containing code corresponding to the patch tree contains necessary functions. If the test case passes all the test cases, the program to be tested is considered to operate normally when returning the correct value.

A negative test case is a test case for checking whether a program containing code corresponding to a patch tree performs an unexpected operation by deliberately performing an operation that may cause an error.

That is, the positive test case checks whether or not an essential function is included, and the negative test case detects an error by performing an operation with a possibility of error.

The GP operation unit 140 can calculate the fitness through the positive and negative test cases using the following equation (2).

,

, 및

는 각가 성공한 포지티브 테스트 케이스의 개수에 대한 가중치, 실패한 네거티브 테스트 케이스의 개수에 대한 가중치, 및 성공한 네거티브 테스트 케이스 각각의 개수에 대한 가중치이며, ptc는 성공한 포지티브 테스트 케이스의 개수이고, ntc는 실패한 네거티브 테스트 케이스의 개수이며,

는 전체 포지티브 테스트 케이스의 개수이고,

는 네거티브 테스트 케이스의 개수이다. 결과값이 클수록 적합도가 더 높은 패치이며, 1.0 이면 경우 패치 생성에 성공했음을 나타낸다.

,

, 및

의 값을 변경하여, 적합도 평가 시 필수 기능 요소 포함 여부와 결함 발생 동작 여부에 대한 중요도를 바꿈으로써 적합도 평가의 기준을 조절 가능하다.Here, I is an evaluation target program including a code corresponding to a patch tree,

,

, And

Ptc is the number of positive test cases that succeeded, ntc is the number of failed negative test cases, and ptc is the number of successful positive test cases. The number of cases,

Is the total number of positive test cases,

Is the number of negative test cases. The larger the result value, the higher the fitness. If the result is 1.0, the patch generation is successful.

,

, And

, It is possible to adjust the criteria of the fitness evaluation by changing the importance of whether the essential functional elements are included in the fitness evaluation and whether the defect occurrence operation is performed.

The GP operation unit 140 adds the tree to which the fitness is calculated to the initial population (S135), and determines whether the number of trees in the initial population is greater than or equal to a predetermined first number (S136). For example, the GP operation unit 140 may generate a tree through intersection and mutation operations until the number of trees included in the initial population reaches 50, and add the tree to the initial population.

If the number of trees in the initial population is smaller than the predetermined first number, the GP operation unit 140 again selects two operation subject trees from the population (S131).

If the number of trees in the initial population is greater than or equal to the predetermined first number, the GP operation unit 140 selects two operation subject trees from the initial population having the first number or more trees (S137). At this time, each tree may be determined using the fitness value calculated in step S134. For example, the GP operation unit 140 may select two trees having a fitness value of the highest level as the operation target tree by using the fitness value, or use the roulette wheel method (each tree of the fitness of the fitness By using the probability of fitness of the target tree as a probability).

The GP operation unit 140 generates a new patch tree by performing the intersection operation (S138) and the variation operation (S139) using the two operation subject trees, and the GP operation unit 140 generates a patch tree The fitness of the tree (S141) is evaluated.

The GP computing unit 140 evaluates whether the fitness of the patch tree generated by the computation is equal to or greater than a threshold value (S142).

If the fitness of the tree generated by the operation is smaller than the threshold value, the GP operation unit 140 adds the generated patch tree to the population (S143) and determines whether the number of trees in the population is equal to or greater than a predetermined second number (S144) do.

If the number of trees in the population is less than the predetermined second number, the GP operation unit 140 generates a new patch tree by applying a genetic programming operation to the population again (S137 to S139), calculates a fitness (S141) (S142).

If the number of trees in the population is equal to or greater than a predetermined second number, the GP computing unit 140 determines whether the population is the last-generation population (S145).

If the population is not the last-generation population, the GP computing unit 140 generates a next-generation population (S146). For example, the GP operation unit 140 generates a population of the next generation by removing a tree below a predetermined fitness value in the current population.

Then, the GP operation unit 140 generates a new patch tree by applying a genetic programming operation to the population of the next generation (S137 to S139), calculates fitness (S141), and evaluates the fitness (S142).

If the fitness of the tree generated by the calculation is equal to or larger than the threshold value, the correction code output unit 150 generates a patch code from the patch tree (S150). Please refer to FIG. 7 together with the patch code.

7 is a diagram showing an example of a new bug code correcting code.

As shown in FIG. 7, the bug syntax B1 in FIG. 4A is changed to the new revision syntax C2 by the bug correction system 100 and the bug correction method according to the embodiment.

That is, according to the bug correction system 100 and the bug correction method according to the embodiment, by using the correction code of the past bug similar to the code to be corrected, even when no information for debugging is included in the target code at all, It is possible to create a patch.

In addition, according to the bug correction system 100 and the bug correction method according to an embodiment, it is possible to deal with various types of bugs by determining a bug similar to a given source code without requiring a complicated predefined operation.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It belongs to the scope of right.

Claims (14)

A search unit for receiving new bug information and searching past bug information similar to the new bug information among past bug information stored in the bug storage,
An AST conversion unit for converting the bug code of the new bug information and the bug correction code of the similar past bug information into an abstract syntax tree (AST)
A GP operation unit for generating a new tree by applying genetic programming to the abstract syntax trees, calculating a fitness of the new tree, and outputting a new tree having a fitness greater than or equal to a predetermined threshold;
A correction code output unit for outputting a new tree having a fitness value equal to or higher than the predetermined threshold value,
A bug correction system. The method according to claim 1,
The search unit, when receiving a new bug report as the new bug information, identifies a topic corresponding to the new bug report received based on the modeled topic, and searches for a past bug with the identified topic in the bug repository Extracting a bug code included in the new bug report and a bug correction code of the searched past bug report to calculate a code block similarity so as to calculate a code block similarity of a past bug report similar to the bug code included in the new bug report To determine the bug correction code,
Bug Correction System. 3. The method of claim 2,
The searching unit searches,
[Equation 1]

To calculate the code block similarity,
Here, Src1 and Src2 denote the bug code included in the new bug report and the bug correction code of the past bug report, respectively.
Bug Correction System. The method according to claim 1,
Wherein the search unit searches the bug storage for a past bug code similar to the new bug code when receiving the new bug code as the new bug information,
Bug Correction System. The method according to claim 1,
Wherein the GP operation unit generates an initial population using the abstract syntax trees, selects two operation subject trees from the initial population, and applies the genetic programming,
Bug Correction System. 6. The method of claim 5,
The GP operation unit,
Performing a crossover operation using the abstract syntax trees, performing a crossover operation on the first tree generated by the crossover operation, calculating a fitness of the first tree on which the crossover operation is performed, Adding the performed first tree to the initial population, and completing the generation of the initial population if the number of trees in the initial population is greater than or equal to a predetermined first number,
Bug Correction System. The method according to claim 6,
The GP operation unit,
When the generation of the initial population is completed, the two operation subject trees are selected from the initial population using the fitness value, the cross operation is performed using the two operation subject trees, and the cross- And calculating a fitness of the second tree in which the mutation operation is performed, and if the calculated fitness is greater than or equal to a threshold value, the second tree in which the mutation operation has been performed is performed as the patch tree Outputting,
Bug Correction System. The method according to claim 6,
Wherein the GP operation unit adds the second tree in which the mutation operation is performed to the population including the initial population if the calculated fitness is smaller than the threshold value and determines whether the number of trees in the population is greater than or equal to a predetermined second number Determining if the population is a last-generation population if the number of trees in the population is greater than or equal to a predetermined second number; and if the population is not a last-generation population,
Bug Correction System. 9. The method of claim 8,
Wherein the GP operation unit removes a tree less than a predetermined fitness from the population to generate the next generation population,
Bug Correction System. The method according to claim 6,
Wherein the GP operation unit sets the target nodes in the two operation subject trees and then performs the cross operation by rearranging the subtrees from the target node to each other,
Bug Correction System. The method according to claim 6,
The shift operation may include:
(1) a deletion operation for deleting only the target node after determining the target node in one tree, or deleting all the subtrees from the target node,
(2) an insert operation for determining a target node in one tree and then copying and inserting some nodes of the one tree at the position of the target node; and
(3) an exchange operation of determining two target nodes in one tree and then reversing the positions of the two target nodes,
Bug Correction System. Receiving new bug information, searching past bug information similar to the new bug information from past bug information stored in the bug repository,
Transforming the bug code of the new bug information and the bug correction code of the similar past bug information into an abstract syntax tree (AST)
Generating a new tree by applying genetic programming to the abstract syntax trees,
Calculating a fitness of the new tree,
Outputting a new tree having a fitness greater than or equal to a predetermined threshold value, and
Outputting a new tree having a fitness level equal to or higher than the predetermined threshold value to a patch code
/ RTI > 13. The method of claim 12,
Searching for past bug information similar to the new bug information,
Receiving a new bug report as the new bug information,
Identifying a topic corresponding to the new bug report received based on the modeled topic,
Searching past bug reports with the identified topic in the bug repository, and
A bug correction code of the past bug report similar to the bug code included in the new bug report is determined by extracting the bug code included in the new bug report and the bug correction code of the searched past bug report ≪ / RTI >
How to fix bugs. 14. The method of claim 13,
Calculating the code block similarity,
[Equation 1]

To calculate the code block similarity,
Wherein Src1 and Src2 are respectively the bug code included in the new bug report and the bug correction code of the past bug report,
How to fix bugs.

Images