Home

- HSR - Institutional Repository

1. A Environment Set up This appendix describes the hardware and software components that support us in reach ing our project goals We give detailed installation and configuration instructions and highlight problem areas to be aware of when setting up a similar environment A 1 Hardware We use a virtual server to host different kinds of software that support us in our daily project tasks The virtual server is hosted by the HSR We have full root access and can connect to the server by VPN if we are outside of the HSR LAN The server runs with Ubuntu 10 04 TLS on 1GB RAM The host name is sinv 56013 edu hsr ch A 2 Project Management Software To support our project management tasks we decided to use Redmine The latest release at the start of our project was version 1 2 1 We use Redmine for the following tasks e issue tracking e time tracking e meeting records e Git repository browsing login required e Gather and share thoughts Our Redmine instance is publicly and read only available at http tiny cc metriculator A 2 1 Set up amp Configuration We followed the set up instructions on red11 to install and configure Redmine using Passenger Basically we used the following commands sudo aa complain usr sbin mysqld sudo apt get update sudo apt get install redmine redmine mysql sudo apt get install libapache2 mod passenger sudo ln s usr share redmine public var www redmine sudo a2enmod passenger Ue
2. required to resolve cdt core tests dependencies of test project in eclipse we use the project testing project that provides required jar files gt lt url gt http download eclipse org tools cdt updates indigo lt url gt lt layout gt p2 lt layout gt lt repository gt lt repository gt lt id gt indigo lt id gt lt url gt http download eclipse org releases indigo lt url gt lt layout gt p2 lt layout gt lt repository gt lt repository gt lt id gt updates lt id gt lt url gt http download eclipse org eclipse updates 3 7 lt url gt lt layout gt p2 lt layout gt Ueli Kunz Julius Weder 33 December 22 2011 metriculator CDT metric plug in lt repository gt lt repository gt lt indigo swtbot release not available yet 14 nov 2011 gt lt id gt swtbot lt id gt lt url gt http download eclipse org technology swtbot helios dev build update site lt url gt lt layout gt p2 lt layout gt lt repository gt lt repository gt lt id gt zest lt id gt lt layout gt p2 lt layout gt lt url gt https hudson eclipse org hudson job gef zest integration ws org eclipse zest repository target repository lt url gt lt repository gt lt repositories gt Listing A 7 Maven repositories Caution Note that at the time of writing the latest swtbot release available is the one for helios Target Platform Since we want Maven to be able to build on Windows ma
3. lt checker Class ch hsr ifs cdt metriculator checkers NumberMembersMetricChecker id ch hsr ifs cdt metriculator classMembers problem category ch hsr ifs cdt metriculator MetricCategory defaultEnabled true defaultSeverity Warning description Number of Members per Type id ch hsr ifs cdt metriculator classmembers messagePattern Number of members of type is above specified maximum of name Number of Type Members gt lt problem gt lt checker gt class ch hsr ifs cdt metriculator checkers ExampleMetricChecker id ch hsr ifs cdt metriculator exampleChecker gt lt problem category ch hsr ifs cdt metriculator MetricCategory defaultEnabled true defaultSeverity Warning description Example Description id ch hsr ifs cdt metriculator exampleproblem messagePattern Example value is above specified maximum of name Example Metric rd Figure D 2 Part of the plugin xml file The highlighted element is the newly added checker 2 Create a new metric class which inherits from AbstractMetric class a If your new metric requires a non default metric value aggregation override the aggregate method 3 Create a new metric checker class that inherits from AbstractMetric Checker a b Define a problem ID Create the name description and preferences strings Add the name description and preferences strings to the MetricLabels properties file in the package resources Add
4. pendencies also come from the AbstractMetricChecker class Metric checkers register itself as job listener to be notified when the Codan command is about to run or has finished This events are mandatory to trigger for example post execution actions like value aggregation BB views BB checkers 4 E E metriculator 186 BS 11 i 29 129 San Wy 2 model 1 N v EH resources Figure B 1 Dependency graph of metriculator generated by Structure101 B 4 Test Coverage We used the EclEmma plug in for Eclipse emm11 to analyse the code coverage after running all unit test The coverage results are shown in Figure B 2 The left column contains the packages and files analysed the right column numbers the test coverage in percent The elements of our interest are expanded and will be explained in this section Ueli Kunz Julius Weder 37 December 22 2011 metriculator CDT metric plug in B 4 1 UI Code The packages mainly dedicated to the view have a low test code coverage This is because we do not have ui test code Following packages heavily contain code that is related to the view views resources tagcloud The AbstractNode implementations in the package model nodes all have a coverage of over 50 The reason for this low coverage in some classes is because they all implement the AbtractNode getIconPath method which is only called from the view Also the AbstractMetricChecker has 7 out of 15 methods that are on
5. Conclusion 23 5 l Performance Tuning 325g 9 4 Regn RO RR Eo a ae 23 glkA imine Issues iuis A T 23 5 1 2 Memory Issues llle 24 5 2 Known Issues xu oho dE ee hoe eed I a aoe Eee 25 5 2 1 Performance s 6 ux Saa ee ee ee 25 5 2 2 Zest Cloudio Integration 2 a 25 5 2 3 Installation via Composite Update Site 25 5 2 4 Merging of Function Declarations in Anonymous Namespaces 25 Did Future Worlds gti nee Y ORE ee Be ee eee Rk aE DG 25 A Environment Set up 28 AL Hardware zc co xU Bae Baek Pee Re eS a 28 A 2 Project Management Software oaoa a a 2 0000 eee 28 A 2 1 Set up amp Configuration ls 28 A 3 Version Control System Git 22e 29 A 4 Development Environment 30 A 5 Continuous Integration Server 2 2 2 30 A 5 1 Set up amp Configuration les 30 A 6 Build and Deployment Automation lens 32 A 6 1 Set up amp Configuration for Windows 32 A 6 2 Set up amp Configuration for Ubuntu 04 4 32 A 6 3 Maven XML Configuration es 33 A 6 4 Jenkins Maven Integration les 34 B Metriculator Metrics 36 B 1 Finding Problems using FindBugs llle 36 B 2 Static Source Code Analysis 222 36 B 2 l Warnings og rrr ee E d SES Sex 36 B 2 2 metriculator Metrics Applied to Itself 36 B 3 Dependecy Analysis a 37 B 4 Test Coverage se 37 B
6. Density of Comments 1 2 1 1 Tag Cloud Integration Tag clouds are an excellent way to visualise weighted data The Sourcecloud plug in for Eclipse sou11 generates a tag cloud of all words in source code files The view of that plug in could be integrated in our plug in to visualise weighted data of one metric For instance the LSLOC values of all functions A higher LSLOC value results in a bigger font size of the function name Ueli Kunz Julius Weder 2 December 22 2011 metriculator CDT metric plug in 1 3 Project Duration The semester thesis starts on September 19th and has to be finished until December 23rd 2011 Ueli Kunz Julius Weder 3 December 22 2011 metriculator CDT metric plug in 2 Requirements This chapter describes the functional and non functional requirements of the metriculator plug in 2 1 metriculator view This view is able to show the data of all metrics for different scopes It visualises the underlying data in three different views hybrid view is the default view and shows the projects and its content as a mix of physical and logical representation in a tree This is the default view and shows a tree view representing the file system structure of the investigated files and folders where the files further contain the logical tree structure that is a simplified form of the abstract syntax tree AST of the file contents The following node types are displayed e project e fol
7. 2011 Codan is a lightweight code analysis framework for the eclipse cdt platform http wiki eclipse org CDT designs StaticAnalysis 2011 Iso iec 14882 PDF 2011 Eclipse bug report Problems view updating too slow https bugs eclipse org bugs show_bug cgi id 349869 2011 jprofiler product page profiling tool for the java virtual machine 2011 Eclemma homepage java code coverege tool for eclipse http www eclemma org 2011 Jez Humble and David Farley Continuous Delivery Reliable Software Re leases Through Build Test and Deployment Automation Addison Wesley M nchen 1 aufl edition 2010 A brief history of static analysis http www coverity com library pdf Coverity White Paper SAT Next Generation Static Analysis pdf 2011 see p 2 chapter A Brief History of Static Analysis Jenkins set up https wiki jenkins ci org display JENKINS Installing Jenkins on Ubuntu 2011 Java ist auch eine insel explanation of the equals method in ger man http openbook galileocomputing de javainsel javainsel 03 007 html dodtp60b4b83b b280 4e66 950f 1adf13899e67 2011 jprofiler product page profiling tool for the java virtual machine http www ej technologies com products jprofiler overview html 2011 Project homepage of the clang project http llvm org 2011 Lsloc counting standard http sunset usc edu research CODECOUNT 2011 mavlla Maven download http maven apache org download
8. Otherwise the Jenkins build will fail because no password has been set A 6 3 Maven XML Configuration Maven uses pom xml for build instructions We have one pom xml in the root directory of all Eclipse projects root pom and one in each subdirectory project pom To automatically generate an initial version of the pom files follow the instructions in chapter A 6 1 All pom files are checked in to the VCS as well This chapter only highlights the changes applied to the generated pom files Version The version tag in all pom files should specify the same version as your plug in manifest does We use lt version gt 0 0 1 qualifier lt version gt in the project pom files and Bundle Version 0 0 1 qualifier in our plug in manifests All project pom files refer to the root pom file The reference contains a version tag as well The version tag of the root pom and the version referenced in the project pom files have to match We use lt version gt 0 0 1 SNAPSHOT lt version gt Repositories Although we imported some CDT projects into our Eclipse workspace see chapter A 4 there is no need for Maven to build them all The only modules that Maven has to build are the metriculator projects Maven knows how to resolve the dependencies of our projects and may download them from remote repositories We only need to specify the repositories to search in our root pom lt repositories gt lt repository gt lt id gt cdt indigo lt id gt lt
9. about software metrics http en wikipedia org wiki Software metric 2011 Git set up on windows http help github com win set up git 2011 The versions of the documents referenced to in this bibliography that we used are stored in our VCS Ueli Kunz Julius Weder 57 December 22 2011
10. html 2011 Ueli Kunz Julius Weder 56 December 22 2011 metriculator CDT metric plug in mavllb Maven set up on ubuntu http lukieb wordpress com 2011 02 15 mcc11 met11a met 11b mys11 PDE11 red11 soull swt11 upd11 Vog11 wik11 win11 installing maven 3 on ubuntu 10 04 1ts server 2011 Mccabe counting standard http www verifysoft com de cmtpp mscoder pdf 2011 see p 40 chapter Die zyklomatische Komplexit t von McCabe Java metrics plug in for ecplise http metrics sourceforge net 2011 list of metric tools http testingfaqs org t static html 2011 Mysql password reset https help ubuntu com community MysqlPasswordResetitAnother way 2C purge 2011 Official help documentation for the eclipse plug in development en vironment http help eclipse org galileo index jsp topic org eclipse pde doc user guide intro pde overview htm 2011 Redmine set up using mod passenger http www redmine org projects redmine wiki HowTo Install Redmine in Ubuntuit Ubuntu 1004 and 10041 using Passenger 2011 Sourcecloud plug in for eclipse https github com misto Sourcecloud 2011 Article about subclassing in swt http www eclipse org swt faq php subclassing 2011 metriculator p2 update site http sinv 56013 edu hsr ch updatesite 2011 Lars Vogel Comprehensive tutorials for eclipse developers http www vogella de 2011 wikipedia article
11. is 57 hours 2496 above the expected 238 hours E 2 2 Mean Time to Fix The mean time to fix Figure E 4 shows how long it took to fix a bug after it was reported The second bug has the biggest MTTF This bug was reported because the Jenkins server was not accessible from outside the HSR LAN Since this circumstance was not critical for us to go on working we did not prioritize it All other bugs were fixed within maximum nine days E 3 Personal Impression From the very beginning of this project we both worked consequent and targeted to create a highly useful and simple to understand plug in In the sub chapters of this section each team member writes about his personal impressions during this project But first of all we would like to thank our advisor Prof Peter Sommerlad for his 18 ECTS 30 hours per ECTS 14 weeks Ueli Kunz Julius Weder 49 December 22 2011 metriculator CDT metric plug in 35 30 25 20 mum Veli Kunz 15 EH Julius Weder Expected 10 5 0 123 4 5 6 7 8 9 10 11 12 13 14 Figure E 3 Time spent per member working on the project Mean Time to Fix in days EN MTTF Poly MTTF High High High Normal Normal High High Normal Figure E 4 Number of days it took to fix bugs after they were reported X axis is sorted ascending by date Ueli Kunz Julius Weder 50 December 22 2011 metriculator CDT metr
12. of the NodeInfo instance to the existing scopeName Following example describes the identifier types used in AbstractNode and is illustrated in Figure 4 6 e scopeName of function f int i is f int i e scopeUniqueName of an anonymous namespace is just a number hash code e hybridId of the function f int i therefore is Workspace ProjectX identifier test cpp 2470869 A1338645 f int 1 2730475 4 2 2 2 CompositeValue The CompositeValue is used by AbstractNode objects to store the calculated values for each AbstractMetric As shown in Figure 4 5 the CompositeValue is implemented as a pure value object Ueli Kunz Julius Weder 16 December 22 2011 metriculator CDT metric plug in Scope 4 projectX 4 identifier test cpp 4 anonymous QA f int i O anonymous o f0 Figure 4 6 Sample tree structure The nodeValue field stores the current value of the node The aggregated Value field is calculated by the nodeValue of it self plus the sum of all nodeValues of its descendant nodes The aggregated Value is calculated once when the Codan job finishes Calculation is done for all nodes in the HybridTreeBuilder See chapter Performance Tuning 5 1 for more details An example which describes the Composite Value is illustrated in Figure 4 7 4 Examplel cpp 0 7 4 Qn 1 5 o funcl int i 2 2 9 classi 2 2 2 2 ni classi imemberFunc bool b Figure 4 7 Example of CompositeValue the values written in b
13. the metriculator hybrid view Screenshot of the metriculator filter view listing all functions Generated tag cloud that visualises the LSLOC value of all functions Hybrid sew sia ok a Meu RO E Soe ae Boe ae ke x7 Logica Vie Wii As e Ru E a ee On A HR ond Eiltet Views uere un AAA eet Sv it ine eee fe us dece uk az Tag clo d Views raties REI e eL pid req pas ae Be ne ee oe Architectural representation of the components metriculator relies on Package diagram with packages inside the metriculator plug in Class diagram with classes related to MetriculatorPlugin Activator Class diagram with classes related to TreeBuilder Class diagram with classes related to AbstractNode Sample tree structure 2s Example of CompositeValue the values written in blue on the left are the node values and the other values on the right are the aggregated values Sample output of the TreePrinter class eee ee Relations between the MetriculatorPluginActivator AbstractMetric and AbstractMetricChecker classes 22e Class diagram with classes related to AbstractMetricChecker Class diagram with classes related to ScopedAST Visitor High level sequence diagram of the processing of metriculator Dependency graph of metriculator generated by Structurel01 Screenshot of the emma code coverage analysis after running all unit tests Use th
14. the name description and preferences strings to the MetricLabels class Create a new instance of the metric in this checker Register the new metric at the MetriculatorPluginActivator singleton in stance Implement the reportProblemsFor method Implement the processTranslationUnit method See chapter 3 for further information about visitors Ueli Kunz Julius Weder 46 December 22 2011 metriculator CDT metric plug in 4 Create a new ScopedAst Visitor a Define the key string for this metric b Override the shouldVisitaxx fields to visit the desired AST nodes c Override and implement the v sit method for the desired AST nodes to be visited could be one or more visit methods D 3 Writing Checker Tests Each checker has its own test class When implementing a new checker test you should be aware of the following e Add your test class to the test suite by modifying AllTests java e You can use the TreePrinter class to print a tree structure to the console e Use the loadCodeAndRun getAboveComment methods provided by Codan to define the C code your checker under test will use e Reset the model after each test method has been ran Ueli Kunz Julius Weder 47 December 22 2011 metriculator CDT metric plug in E Project Management This chapter provides an overview to project management related tasks such as project planning and spent time analysis E 1 Project Plan The project lasts fourte
15. your command line change to your projects root directory e Run run mvn org codehaus tycho maven tycho plugin generate poms DgroupId ch hsr ifs cdt metriculator Dtycho targetPlatform C Programme eclipse Where targetPlatform should point to your Eclipse installation folder This command will generate or update pom xml files in all directories in your project root Chapter A 6 3 explains how to customize this pom xml files A 6 2 Set up amp Configuration for Ubuntu Follow the steps on mav11b to install Maven 3 on Ubuntu Basically these are the steps sudo mkdir usr local apache maven cd usr local apache maven sudo wget http ftp heanet ie mirrors www apache org dist maven binaries apache maven 3 0 2 bin tar gz tar xzvf apache maven 3 0 2 bin tar gz Listing A 5 Maven set up commands After that make sure to configure the environment variables as described at mav11b Our plug in unit tests require Maven to execute headless builds That for we have to install Xvnc with a X11 window manager Use the following commands to install Xvnc as virtual screen buffer and metacity as window manager Ueli Kunz Julius Weder 32 December 22 2011 metriculator CDT metric plug in sudo aptitude install vnc4server metacity sudo su jenkins vncserver enter password manually Listing A 6 Set up xvnc After the install command we have to manually set the vncserver password for the Jenkins user
16. 0 0 Figure B 2 Screenshot of the emma code coverage analysis after running all unit tests Ueli Kunz Julius Weder 39 December 22 2011 metriculator CDT metric plug in C User Manual This user manual explains developers familiar with Eclipse and C how to use met riculator It assumes that Eclipse CDT is already installed C 1 Installation 1 Install the Zest framework via update site from https hudson eclipse org hudson job gef zest integration ws org eclipse zest repository target repository see Figure C 1 8 instal lia Available Software Check the items that you wish to install Work with https hudson eclipse org hudson job gef zest nightly ws org eclipse zest repository target rep v Find more software by working with the Available Software Sites preferences type filter text Name Version 4 Y 000 Graphical Editing Framework Zest Visualization Toolkit v g Graphical Editing Framework Zest Visualization Toolkit SDK 2 0 0 201111252109 Figure C 1 Use the wizard at Help gt Install New Software 2 Install metriculator via update site from http sinv 56013 edu hsr ch updatesite site see Figure C 2 Available Software Check the items that you wish to install Work with http sinv 56013 edu hsr ch updatesite site Find more software by working with the Available Software Sites preferences type filter text Name Version a Y 000 metri
17. EA ULbDGode sso Rep She a eR te 38 Bid 2 Model Code 5 xi uus dome x Poh X Elea 38 C User Manual 40 Lo Installation 3 2 ac eo IN ee utbs edle Boi ge 40 C 2 Start code analysis with metriculator 41 C 3 Configuration of metriculator a o 41 C 4 Working with the Views lA 42 ULT View Types dus uum eek rites Lp oe So deme t a ih 42 C 5 Problem Reporting and Markers llle 42 Ueli Kunz Julius Weder ix December 22 2011 metriculator CDT metric plug in D Developer Manual 45 Dil Set UDs ars aena AX ERE a Say ee Be Ret ae 45 D 2 Adding a new Metric 2A 46 D 3 Writing Checker Tests 2e 4T E Project Management 48 EL Project Plans m tye te RE A Be arae eg 48 E 2 Time Schedules ces oo Ee s Rods d ee ee 49 E 2 1 Spent Time per Project Member 49 E22 Mean Time to Fix e u eGo eed Pak Ee Oe DS PE 49 E 3 Personal Impression eh 49 Eds 7Ueli I amp unzzova out Sheth a a V RERENSSEIGQEEu ILS e 51 E 3 2 Julius Weder 2e mpm m eoe e y RE A XP Rev bo ag 51 F Nomenclature 55 Experience is a hard teacher because she gives the test first the lesson after ward Vernon Law Ueli Kunz Julius Weder x December 22 2011 metriculator CDT metric plug in 1 Introduction The investigation of source code has ever since been an important task in software development his11 Static sou
18. INSTITUTE FOR SOFTWARE E HSR HOCHSCHULE F R TECHNIK re Bi RAPPERSWIL FHO Fachhochschule Ostschweiz HSR University of Applied Sciences Rapperswil Institute for Software Semester Thesis metriculator CDT metric Plug in Ueli Kunz ukunz hsr ch Julius Weder jweder hsr ch http sinv 56013 edu hsr ch Supervised by Prof Peter Sommerlad December 22 2011 metriculator CDT metric plug in Abstract This thesis aims at statically analysing software written in C using known software metrics Software metrics are used to obtain objective reproducible and quantifiable measurements of source code This measurements may support various tasks such as performance optimization quality assurance testing or software debugging wik11 Metriculator is programmed in Java and integrates in Eclipse as plug in that depends on the Codan cod11 framework which is part of the C C Development Tooling platform CDT CDT11 Codan is a code analysis framework that offers a mechanism to add new code analysis features Each metric in metriculator is implemented as an extension to Codan First and foremost it is about extracting meaningful information out of C source code using different software metrics After analysing the code the results are shown in the metriculator view which provides different representations of the same underlying data Additionally the metriculator view helps finding problems and moving to their problema
19. PLAY DISPLAY THIS readlink f 0 BUNDLE_ROOT dirname THIS cd BUNDLE_ROOT mvn e clean install Listing A 9 Jenkins build script Ueli Kunz Julius Weder 35 December 22 2011 metriculator CDT metric plug in B Metriculator Metrics This chapter describes the results of a metric analysis applied to the metriculator plug in source code B 1 Finding Problems using FindBugs FindBugs did not find any problems B 2 Static Source Code Analysis We used the Eclipse metrics plug in met11a to analyse the metriculator source code B 2 1 Warnings The analysis found two warnings both are related to the McCabe metric The method LSLOCScopedAST Visitor visit LAS TStatement has a McCabe value of 16 which is 6 above the threshold of 10 Since this method has already been refactored heavily and reviewed by our supervisor we decided not to change the current implemen tation The method Nodelnfo equals Object has a McCabe value of 14 which is 4 above the threshold of 10 Since we have many fields to compare the McCabe value is exception ally high Based on the best practice patterns jia11 we used in our implementation we do not think it would make much sense to further refactor the method B 2 2 metriculator Metrics Applied to Itself This section outlines the metric results of the metrics that both the metrics plug in and metriculator implement The metrics plug in does not always use the exact sa
20. Some count the number of statements others count the physical lines of source code with or without comments or even the number of byte code instructions We decided to use the term LSLOC Logical Source Lines of Code because the Univer sity of Southern Carolina defines a standard for LSLOC counting rules for C Isl11 As long as not otherwise noted metriculator respects the rules of this standard Because this standard does not define rules for some of the new concepts of the C 11 cpp11 standard especially lambdas we defined new rules where required 3 1 1 Lambda Expressions int z x auto square z x 1 one lambda capture two captures O 1 1 function body return z z 1 statement 1 assignment statement Listing 3 1 Lambda expression code snippet with LSLOC of 4 As illustrated in listing 3 1 the following rules apply to lambda expressions 1 A lambda capture counts one regardless of the number of captures it contains 2 Anonymous function declaration counts one as any other function declaration does 3 Statements in the lambda body count as usual Ueli Kunz Julius Weder 7 December 22 2011 metriculator CDT metric plug in 4 The assignment of the lambda expression to square counts one as any other as signment 5 Direct calls of lambda expressions no assignment count one In contrast to other function declarations lambdas may have a lambda capture that
21. an command on the UI Codan automatically calls all registered checkers The metriculator component in Figure 4 1 can further be divided into the packages shown in Figure 4 2 Figure 4 1 Architectural representation of the components metriculator relies on The metriculator package is the core of this plug in Thanks to the PluginActivator we know when the Codan command is started and finished This allows us to initialise the model each time before the checkers start working After the work is done the Plugin Activator notifies the views to update their data via observer As the checkers run they build a tree model that represents the analysed files and their content The bidirectional dependency between the model and the metriculator component ex Ueli Kunz Julius Weder 12 December 22 2011 metriculator CDT metric plug in Figure 4 2 Package diagram with packages inside the metriculator plug in ists because some model classes use a singleton inside of the metriculator component Singletons are used due to constraints of the Eclipse framework 4 1 1 Plug in Activator The plug in activator activates the plug in when it is about to run Every Eclipse plug in inherits from the AbstractUIPlugin of Eclipse The MetriculatorPluginActivator holds the singleton instance of this plug in The constructor of the MetriculatorPluginActiva tor registers itself at a job listener which notifies metriculator when Codan starts the co
22. ar gitrepo chgrp group var gitrepo chmod gts var gitrepo create a git repository shared all users of the group must have access bare no working directory it is just the repository cd var gitrepo Ueli Kunz Julius Weder 29 December 22 2011 metriculator CDT metric plug in mkdir repository git cd repository git git init bare shared group acces via ssh with public key authentication every user has to generate a key linux generates a private and a public key in ida_rsa pub ssh keygen t rsa b 2048 f ssh id_rsa the public key of the users has to be installed on the server umask O77 mkdir v home user ssh cat id_rsa_ user pub gt gt home user ssh authorized_keys chown user user home user ssh tt ida_rsa_ user pub ist he public key of the user cloning the git repository git clone ssh user host var gitrepo repository git Listing A 3 Git repository set up Caution For using Git on Windows and how to set up the required SSH access see winll A 4 Development Environment The plug in was developed in Eclipse Indigo using the plug in development environment PDE plug in To test metriculator we used sample C4 4 11 source code To force the compiler to build according to the C 11 standard add the std gnu 0x flag to the following field Project Properties gt C C Build gt Settings gt GCC C Compiler gt Miscellaneo
23. as the C C Development Tooling CDT plug in D 1 Set up 1 Checkout sources from the Git repository at sinv 56013 edu hsr ch var gitrepo metricular git 2 In Eclipse Import Existing Projects into Workspace select the repository checked out from point 1 3 Set the missing baselines to ignore in Eclipse gt Window gt Preferences gt Plug in Development gt API Baselines 4 Open the target file in the package metriculator and set it as target platform illustrated in Figure D 1 5 Update all the locations illustrated in Figure D 1 JBledttargettarget 52 2 Te Target Definition Set as Target Platform 18 Target Name Entera nameforthis target Restored Target Locations The following list of locations will be used to collect plug ins for this target definition ES Siwerkspace locj testing project 3 plug ins available Add S http download eclipse org eclipse updates 3 7 390 plug ins available A http download eclipse org releases indigo 133 plug ins available S https hudson eclipse org hudson job gef zest integration ws org eclipse zest repository target repository 164 plug ins available Show location content Figure D 1 Target file 6 Clean all projects if there are still errors Ueli Kunz Julius Weder 45 December 22 2011 metriculator CDT metric plug in D 2 Adding a new Metric 1 Add a new checker with a problem in the plugin xml illustrated in Figure D 2
24. ator shows its view in Eclipse Ueli Kunz Julius Weder 20 December 22 2011 metriculator CDT metric plug in add IScopeListener leave visit Figure 4 11 Class diagram with classes related to ScopedAST Visitor Ueli Kunz Julius Weder 21 December 22 2011 metriculator CDT metric plug in A metriculator lt lt return gt gt run X jobaboutto run reset model lt lt return gt gt foreach checker process translation unit visitor for nodes show view lt lt return gt gt X X Figure 4 12 High level sequence diagram of the processing of metriculator Ueli Kunz Julius Weder 22 December 22 2011 metriculator CDT metric plug in 5 Conclusion Overall we reached our basic objectives defined in chapter 2 The plug in performs well in medium sized projects It has a solid design and is easily extensible The following chapter describes analysis and problem finding processes related to per formance issues After that we describe the unresolved issues of the project At the end we describe ideas to improve and extend metriculator 5 1 Performance Tuning In week nine we set up the last step in our deployment pipeline HF10 That is the continuous and automated publishing of the latest plug in build The plug in was made available through a publicly available p2 update site upd11 The update site allowed us to install t
25. atorPluginActivator See Figure 4 9 for more details Y MetriculatorPluginActivator Figure 4 9 Relations between the MetriculatorPluginActivator AbstractMetric and Ab stractMetricChecker classes Beside the checker reference AbstractMetric implementations are responsible to imple ment a value aggregation strategy The AbstractMetric class provides a default aggrega tion strategy as you can see in Listing 4 1 public int aggregate AbstractNode node CompositeValue metricValue node getValueOrDefaultOf getKey metricValue aggregatedValue 0 for AbstractNode child node getChildren metricValue aggregatedValue aggregate child Y metricValue aggregatedValue metricValue nodeValue return metricValue aggregatedValue Ueli Kunz Julius Weder 18 December 22 2011 metriculator CDT metric plug in Listing 4 1 Code snippet of the default value aggregation strategy implemented in Ab stractMetric 4 3 1 Tasks Each AbstractMetricChecker implementation is responsible for the following tasks e Visiting of the AST using a ScopedAST Visitor see 4 3 3 e Initialisation of its profile preferences e Problem creation and reporting The following tasks are taken care of by the AbstractMetricChecker class e Getting the hybrid tree model see 4 2 1 e Creating the file and folder nodes for each processed file and adding it to he hybrid tree model see also 4 4 e Resetting its problems all tree
26. ber of Parameters per Function measures the number of parameters of a function See 3 3 for detailed description Number of Members per Type measures the number of members of a class struct or union See 3 4 for detailed description Efferent Coupling counts the number of foreign types a certain type depends on For instance many dependencies indicate highly coupled software See 3 5 for detailed description Compared to the specified objectives at the start of the project there are two metrics missing The Number of Nested Levels and the Number of Template Parameters metric were not implemented This is because we decided to implement the Efferent Coupling metric in advance and because of the unforeseen performance issues that must have been resolved prior to additional feature implementations See 5 1 for more details Metriculator is available as Eclipse plug in and can be installed from within Eclipse using the install wizard After installation users can run the Codan code analysis As part of this analysis metriculator will run as well The gathered metric data can be investigated using different views provided by metriculator Illustrated Example In this section we will explain the results shown in the metriculator views after analysing a sample project Default View Figure 0 1 shows a screenshot of the default metriculator view after the static code analysis ran The default view shows all analysed files and folders in a tree v
27. called NodeInfo that serves as pure value object holding only the relevant infos of an ASTNode that we really need to keep As a result memory allocation significantly decreased But there is still capability left to improve the memory allocation in Nodelnfo We hold an Binding reference that allows us to merge function declaration and function definition The merging task has to run after all checkers ran because function declara tions and definitions may be in different files Since the problematic memory allocation happens during the processing we can not release the Binding references and that still causes a significant amount of memory being allocated The second problem was caused by the jface TreeViewer component We used the TreeViewer component in the hybrid view as well as in all others views including the filter view But in the filter view we never displayed hierarchical data but very large lists of for instance FunctionNodes The TreeViewer component is not optimized to display flat lists Therefore we decided to use the TableViewer component in filter views That decision tremendously improved the performance of the filter view mode Now displaying and sorting thousands of FunctionNodes just takes a fraction of a second The introduction of the TableViewer component beside the existing TreeViewer com ponent forced us to create duplicated almost identical code Although their API is very similar TreeViewer and TableViewer do not s
28. ch means that cloning one node in fact cloned the node itself and all descendant nodes Before caching aggregation values this step was mandatory to show correct values in filter views Thanks to cached values we changed the deep clone Ueli Kunz Julius Weder 23 December 22 2011 metriculator CDT metric plug in implementation to a shallow clone implementation That is much more faster and saves even more memory Similar to the value aggregation problem was the path algorithm problem Each Ab stractNode has a unique path that represents its position in the tree The initial im plementation requested the path of a parent node every time a child was added Since the path building algorithm was recursively it was very time and also memory consum ing Building the path only once and only force recalculation if mandatory solved that performance issue 5 1 2 Memory Issues The main reason why metriculator crashed at the beginning was because the huge amounts of memory allocation The virtual machine run out of heap space and Eclipse crashed We identified two problem areas During the processing metriculator created ten thousands of char instances that al located mega bytes of memory We figured out that the instances were hold by C parser class instances that were used to create the AST Since AbstractNode referenced the AST Node instances they could not be removed by the garbage collector As a solution we introduced a new class
29. chines as well as on our Linux based CI server we had to define different target platforms in our root pom We use Tycho in version 0 10 0 lt groupId gt org sonatype tycho lt groupId gt lt artifactld gt target platform configuration lt artifactld gt lt version gt tycho version lt version gt lt configuration gt lt resolver gt p2 lt resolver gt lt environments gt lt environment gt lt os gt win32 lt os gt lt ws gt win32 lt ws gt lt arch gt x86 lt arch gt lt environment gt lt environment gt lt os gt linux lt os gt lt ws gt gtk lt ws gt lt arch gt x86_64 lt arch gt lt environment gt lt environments gt lt configuration gt Listing A 8 Maven target platform configuration A 6 4 Jenkins Maven Integration There is a Maven plug in available for Jenkins This plug in wraps the mvn command of Maven But it is also possible to use a custom build script that invokes the mvn command Since our build requires other actions being executed prior to the Maven Ueli Kunz Julius Weder 34 December 22 2011 metriculator CDT metric plug in build we use a custom build script Another advantage of such a script is that it is easier to maintain All build steps are saved in one place and easy to change bin sh export PATH usr local apache maven apache maven 3 0 3 bin PATH make DISPLAY from xvnc globally available so that other processes can access it export DIS
30. culator v E metriculator 0 0 1 201112211122 Figure C 2 Use the wizard at Help gt Install New Software Ueli Kunz Julius Weder 40 December 22 2011 metriculator CDT metric plug in C 2 Start code analysis with metriculator 1 Choose your C source code to be analysed Select one or more files and folders 2 In the Project Explorer right click on the selection This can be one or more projects folders and files Then choose to run the command Run C C Code Analysis This will start the static code analysis and metriculator will analyse the source code with the activated metrics After the analysis the metriculator view opens C 3 Configuration of metriculator 1 Open the Eclipse preferences via Window gt Preferences 2 Open the Code Analysis preferences for C C see Figure C 3 ES AAA O Cosme type filter text Code Analysis pes General E Function cannot be resolved amp 9 Error Ant E Invalid arguments Error C C E Method cannot be resolved G Error Appearance E Field cannot be resolved O Error Build E Abstract class cannot be instantiated G Error Code Analysis E 4 7 Security Vulnerabilities Code Style Format String Vulnerability amp Warning Debug 4 E Metric Problems E Editor Y McCabe Cyclomatic Complexity Warning File Types v LSLOC amp Warning Indexer V Number of Parameters Warning Language Mappings Number of Type Members Warning 3 New CDT P
31. de analysis and when Codan has done its job Before the metriculator plug in starts its job it has to reset all model data The job listeners also reset the caches of the metric values clear the views and models to deref erence all old objects from a previous execution of metriculator See 5 1 for further information All metric checkers has to register themselves at the MetriculatorPluginActivater After Codan has finished its job the MetriculatorPluginActivator invokes the aggregation of all values of the activated metrics and shows the metriculator view MetriculatorPlug inActivator has a JobObservable that implements the Observable and is responsible for notifying his observers if the plug in is about to run or if its job is done Observers of the JobObservable are the metriculator view and all metric checkers 4 2 Tree Structure This sections describes the composite tree structure that is build by the AbstractMetric Checker instances during the processing described in Chapter 4 4 This tree structure holds all the data shown in the metriculator view Ueli Kunz Julius Weder 13 December 22 2011 metriculator CDT metric plug in MetriculatorPluginActivator Attributes metrics HashMap lt Class lt AbstractMetric gt AbstradMetric x a in send eim Y HatTreeBuilder B plugin MetriculatorPluginActivator Zi Operations A aboutToRun addCommandExecutionListener E gt registerMetric Abst
32. der e file e namespace type class struct union e function logical view shows the logical representation of the source code in one tree This view merges the logical trees of all files and thus does not show any file system infor mation The following node types are displayed e namespace e type class struct union e function filter view shows a flat representation of all hybrid tree nodes The node list can be filtered by the different types of nodes Following filters exist e file e namespace e type class struct union Ueli Kunz Julius Weder 4 December 22 2011 metriculator CDT metric plug in Figure 2 1 Hybrid view Figure 2 2 Logical view Function Function2 Function3 Figure 2 3 Filter view Ueli Kunz Julius Weder 5 December 22 2011 metriculator CDT metric plug in e function tag cloud view is another nice way of visualising metric values by generating a tag cloud that allows fast visual detection of potentially problematic items such as large functions or high coupled classes Figure 2 4 COLT main CQICUIOE prit view extract Measure getinfos vist poe filter reportProblem expand Collapse create complete Figure 2 4 Tag cloud view In all views the columns can be sorted ascending or descending by scope name or metric value The nodes in the tree structure can be expanded and collapsed fully or partially 2 2 Use Cases Brief Forma
33. different views to watch at the results That helps you detect problems and obtain an overview of the quality and size of the analysed source code Different metric ulator views show the metric values from different perspectives In all views all columns can be sorted Cells with metric values beyond the thresholds are highlighted this helps figuring out potential problems in your source code All scope labels for example a function name can be double clicked to open an eclipse editor with the file where the function is in and to select the function code Views with a tree view hybrid view and logical view can be easily expanded and collapsed with the commands in the upper right corner of the metriculator view C 4 1 View Types e hybrid view Figure C 5 e logical view Figure C 6 e filter view Figure C 7 e tag cloud Figure C 8 C 5 Problem Reporting and Markers Metriculator allows to report problems to the Problems View built in Eclipse This will also set markers in source code editors next to the problematic source code sections If the option Create problem markers in the metric problem preferences is activated see Figure C 4 metriculator reports the problems Example C 9 and sets the markers Example C 10 Otherwise the problems are just visible inside the metriculator view Cells with problematic values are highlighted anyway Hint Deactivating problem reporting can improve the speed of the analysis ecl11a Ueli K
34. e task Specification of metrics has been renamed to Specification amp Impl of metrics As we started with the implementation of the first metric 3 1 we noticed that there was no task on the project plan for metric implementation We did not create a new one instead we extended the existing specification task since specification and implementation can not strictly be separated especially because we follow iterative development cycles The second change is related to user interface tasks They were initially planned to start in week 43 two weeks before the release of the prototype In week 41 we gladly realized that we are on schedule and that it would be reasonable to start with the implementation of the view one week earlier as planed As we started with the view we decided to shift the markers and settings tasks to the end of the prototype phase because they were relatively easy to implement since the Codan framework provides a simple API to handle them E 2 Time Schedules This chapter evaluates the time spent during the project The first section describes the time spent per project member and week the next section the time spent on each reported bug E 2 1 Spent Time per Project Member Figure E 3 shows the time spent per member The semester thesis module is worth 8 ECTS This means that the expected work per week of an average student to pass the module is about 17 hours In average each of us worked about 295 hours in total which
35. e wizard at Help gt Install New Software o o o Use the wizard at Help gt Install New Software sn metriculator checker preferences a a Customise Problem a Hybrid view This is the default view It shows the projects and its content as a mix of physical and logical nodes in a tree structure Logical view Shows the logical representation of the source code in a tree structure The logical elements like namespaces or classes are decoupled from the physical location and merged together Filter view Shows a filtered representation of the hybrid view Possible filters are files namespaces types and functions 12 13 14 15 16 17 17 17 18 20 21 22 37 40 40 41 42 43 43 Ueli Kunz Julius Weder 52 December 22 2011 metriculator CDT metric plug in C 8 Tag cloud visualisation of the nodes for a specific metric The higher the metric value the bigger the font size of the name in the cloud This tag cloud was generated for the metric LSLOC 44 G 9 Problems View ena ee RoR a BPE i ge bo Ee et Es 44 C 10 Problem marker in the source code editor with a message explaining the A A epike a a a da he aT o i e AARAA ae nie te p aire nia dA a 44 Dl Target fe our rr a OY a a ee ee 45 D 2 Part of the plugin xml file The highlighted element is the newly added AA EE A E sd EE Lu ns seo e 46 E 1 Initial version of the p
36. eli Kunz Julius Weder 19 December 22 2011 metriculator CDT metric plug in k T a AbstractMetricChecker me Observer Attributes zi Operations initPreferences IProblemWorkingCopy leaving AbstractNode processAst IASTTranslationUnit a interface processRescource IResource IScopelistener reportProblem String problemId AbstractNode abstractNode Object messageParameters E Attributes update Observable observable Object state r visiting AbstradNode amp Operations process TanslationUnit IASTTranslationUnit leaving AbstradNod reportProblemsFor AbstradNode visiting AbstradNod Figure 4 10 Class diagram with classes related to AbstractMetricChecker 4 4 Processing The Codan framework invokes the metriculator checkers At the end of the processing of Codan the view of metriculator appears The whole processing is described in detail below and is illustrated in Figure 4 12 1 A user runs Codan with the Run C C Code Analysis command in Eclipse CDT 2 Metriculator is notified about the start of Codan 3 Metriculator resets its models 4 Codan starts processing the C resources to be analysed 5 For each resource each checker is called to process it 6 Each translation unit is processed 7 The metriculator checker visits the relevant AST nodes to calculate metric data 8 When Codan has finished metricul
37. ember function declarations and definitions 3 4 Number of Members per Type The Number of Members per Type metric is a software metric that counts the members of a type A types can be a class struct or union Our implementation is based on the official C 11 Standard cpp11 especially on the chapter Class Members As long as not otherwise noted metriculator respects the rules in this standard 3 4 1 Explanation of Member Types There are three different kinds of members 1 data members 2 member functions 3 nested types a class c struct union e a enum e typedef declaration Members are counted whatever visibility they have Nested types can be anonymous Members are not counted if they are friend classes or member functions Listing 3 6 illustrates an example class Example int i counts 1 public class NestedExamplei counts 1 struct NestedExample2 counts 2 itself as nested type and its member int i counts 1 enum a b c counts 1 typedef int myInt counts 1 void doIt counts 1 friend class MyFried does not count friend void doAnother does not count 3 void doNothing does not count Ueli Kunz Julius Weder 10 December 22 2011 metriculator CDT metric plug in Listing 3 6 Code snippet for Number of Members per Type 3 5 Efferent Coupling Efferent Coupling is a metric that numbers on how many distinct foreign units the
38. en weeks from September 19 to December 23 2011 The first version of the project plan E 1 was created in week one During the first seven weeks the initial project plan has experienced a few minor changes E 2 Wochen KW 38 KW39 KW 40 KW 41 KW 42 KW 43 KW 44 KW 45 KW 46 KW 47 KW 48 KW 49 KW 50 KW 51 19 09 11 26 09 11 03 10 11 10 10 11 17 10 11 24 10 11 31 10 11 07 1111 141111 211111 281111 05 1211 121211 191211 Task Projekt Setup Metric Analysis Specification of the metrics Use Cases Prototype Plugin Lines Of Code Metric Markers and Annotations Metric View detail view of metrics Metric Plugin Settings Configuration Delivery Presentation Figure E 1 Initial version of the project plan 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Wochen KW 38 KW39 KW 40 KW 41 KW 42 KW 43 KW 44 KW 45 KW 46 KW 47 KW 48 KW 49 KW 50 KW 51 19 09 11 26 0911 03 10 11 10 1011 17 10 11 24 10 11 2311011 071111 141111 211111 281111 051211 121211 191211 Task Projekt Setup Metric Analysis Specification amp Impl of the metrics Use Cases Prototype Plugin Lines Of Code Metric Markers and Annotations Metric View detail view of metrics Metric Plugin Settings Configuration Delivery Presentation Release Names at starting dates 2 E o o9 pA gt o o a x Figure E 2 Final version of the project plan green shoots deep blue sea vaikuntha Ueli Kunz Julius Weder 48 December 22 2011 metriculator CDT metric plug in Th
39. hare common base that we could use to create an abstract interface to interact with both of them Additionally subclass ing is not allowed in SWT swt11 Therefore we just duplicated the code used for the TreeViewer and replaced the types used Ueli Kunz Julius Weder 24 December 22 2011 metriculator CDT metric plug in 5 2 Known Issues There are some known issues that should be taken care of in future releases This chapter describes this issues and also contains a reference to the issues documented in Redmine 5 2 1 Performance As already described in detail in section 5 1 we optimized metriculator to perform better on large input data Currently metriculator can analyse about 300 000 physical lines of code if problem reporting is disabled and only the LSLOC metric is active Issue 117 at http sinv 56013 edu hsr ch redmine issues 117 5 2 2 Zest Cloudio Integration The tag cloud integration is not stable yet The code of this feature was taken from the project at soul1 after the author contacted us and suggested to integrate his work We just customized the code to work within metriculator Generating a tag cloud based on large lists for instance function nodes sometimes results in exceptions saying that there is not enough drawing space available Issue 176 at http sinv 56013 edu hsr ch redmine issues 176 5 2 3 Installation via Composite Update Site When installing metriculator multiple steps are requi
40. he plug in on an independent eclipse instance which is mandatory to test the performance According to the quantitative goals of this semester thesis 1 we ran our performance tests against the llvm project llv11 especially the clang sub project cla11 The first manual test runs were very disappointing The performance was very bad Analysing two thousand LSLOC did work Everything above was critical and most often eclipse ran out of heap space Also the operation took very long to finish We soon realized that for the remaining two weeks of development performance tuning will have highest priority The plug in must perform well otherwise it would be useless regardless of the number of features implemented Thanks to jprofiler jpr11 we quickly figured out the time consuming methods and memory bottle necks 5 1 1 Timing Issues We first optimized metriculator to take less time of execution This was mainly achieved by reducing the number of method calls by caching values instead of recalculating them on every request An AbstractNode is able to sum up all node values of its descendant nodes This process has been triggered very often by the tree viewer control since every cell displays such an aggregated value To fix that problem we now only aggregate the values once at the end of the code analysis before the data are bound to the view Another hot spot was the clone mechanism of an AbstractNode Cloning was imple mented as deep clone whi
41. ic plug in valuable time and competent advices Special thanks goes to Thomas Corbat who was always ready to generously assist us at technical problems E 3 1 Ueli Kunz At the start of this project I was very excited what will be the outcome I was highly motivated to deliver a well designed and stable plug in Although it was a small team I was also interested in gathering more project management experiences and applying the lessons learned from earlier term projects Getting to known with a new framework and contributing to it was also a motivating task Since I have already benefited from other open source projects I have been willing to give something back At the beginning we had some extra effort setting up the project environment soft ware Setting up Jenkins with Git and Redmine was worth the experience but I think next time it will not bother us that much any more To set up build automation I had to learn Maven from scratch Afterwards I do not regret it All in all I feel like we spent too much time working on tasks not directly related to the goals of this thesis I am proud of the overall result We reached our objectives and the plug in is good enough to be used in medium sized projects It has a solid design and is easily extensible I always had fun with Julius Weder also during the effective extreme programming sessions late in the evenings We often discussed and solved problems together that was a great experience to
42. iew Files can further be expanded to explore the source code they contain The values of each metric are displayed in a separate sortable column Each row represents one analysed scope Warnings As visible in Figure 0 1 func3 has seven parameters Because seven is above the threshold set in the metric preferences the cell is highlighted Filter View The filter view in Figure 0 2 shows all functions Other filters can be applied to see a list of only files types or namespaces Thanks to the filter view users are able to quickly identify large files for instance Data Visualisation Metric values can also be visualised in a tag cloud as illustrated in Figure 0 3 using the LSLOC metric values The bigger the font size of the function name the higher its metric value is Ueli Kunz Julius Weder iv December 22 2011 metriculator CDT metric plug in L Problems 4 metriculator 3 gt 3 Source Cloud amp ts ID ee HE B McCabe M LSLOC NbParams 16 32 16 16 a Example2 cpp 16 4n o func2 func3 int long bool int int long short 4 6 Examplel cpp s s On o funcl int i 9 classi nl classl memberFunc bool b a e Main cpp 9 main w N k e 27 20 20 6 18 1 0 1 7 1 5 1 2 1 2 1 2 1 5 1 3 Figure 0 1 Screenshot of the metriculator hybrid view 12 Problems metriculator ZA 3 Source Cloud amp k og Fo FILI Scope McCabe LSLOC NbParams 9 main 1 3 0 O f
43. increases the LSLOC value by one 3 1 2 Enum enum x vali val2 1 type definition Listing 3 2 Enum code snippet with LSLOC of 1 As illustrated in listing 3 2 the following rules apply to enums 1 An enum declaration counts one 2 Enum values do not count 3 2 McCabe Cyclomatic Complexity CC The McCabe metric also known as Cyclomatic Complexity mcc11 is a software metric for measuring the complexity of parts of a software This metric indicates whether a piece of code is still comprehensible to humans Simply expressed it is the number of binary branches plus one Our implementation is based on a paper published by Verifysoft Technology GmbH mcc11 3 2 1 Example namespace ni 1 2 1 2 initial funci func2 children int funci int i 1 1 if i lt 01 1 return 0 return 0 int func2 int 1 return 1 Ueli Kunz Julius Weder 8 December 22 2011 metriculator CDT metric plug in Listing 3 3 Code snippet example McCabe Every node has an initial Cyclomatic Complexity of 1 In listing 3 3 the function func1 has due to the if statement a Cyclomatic Complexity of 2 The parent node of the two functions has therefore a Cyclomatic Complexity of 2 To build the node value of a parent node all children node values are summed up Afterwards the sum of the parent is decremented by the number of children 3 3 Number of Parameters per F
44. li Kunz Julius Weder 28 December 22 2011 metriculator CDT metric plug in sudo chmod a x usr share redmine public sudo service apache2 restart Listing A 1 Redmine set up commands After these steps Redmine is up and running at http host redmine To enable Redmine repository browsing the user under which Apache runs has to have read access to the Git root directory By default Apache runs as www data Additionally you might have to refresh Redmines repository cache preferably gather all users that require git access in one group sudo adduser www data group with read access cd path to redmine amp amp usr bin ruby1 8 script runner Repository fetch_changesets e production Listing A 2 Redmine repository browsing set up Caution Redmine version 1 2 x requires rails version 2 3 11 with Ruby gems 1 3 7 We also had to reset our MySQL root password We followed the instructions on mys11 to accomplish this A 3 Version Control System Git To support our file version management we decided to use Git The latest release at the start of our project was version 1 7 6 We used the following commands to set up Git login as root add a group for the git users group git users groupadd group add a user user name_of_user adduser user add the user to the git group usermod aG group user shared directory for the git repository mkdir pv var gitrepo chmod 770 v
45. lue on the left are the node values and the other values on the right are the aggregated values 4 2 3 TreePrinter The TreePrinter class was mainly used for debugging purposes We use it in our test cases to print the generated tree to the console after a test has been ran Figure 4 8 shows a sample output Scope names are trimmed to a maximum length and the metric values in each column are left aligned The node values are in braces next to the aggregated values metric columns LSLOC testworkspace 670 0 testproject 670 0 filel 650 10 nsi 10 10 ns31 310 310 ns32 320 320 file2 20 20 Figure 4 8 Sample output of the TreePrinter class Ueli Kunz Julius Weder 17 December 22 2011 metriculator CDT metric plug in 4 3 Metric Checkers Codan offers extension points for third party tools to add new checkers Metriculator uses Codan s extension points to define one checker per software metric Every checker in metriculator inherits from the AbstractMetricChecker that again inherits from the AbstractInderAstChecker of Codan Metriculator uses the AbstractMetric class to encapsulate checkers and metric relevant data Each checker implementation is associated with one AbstractMetric implemen tation When the checker is created on application start up it associates itself with the corresponding metric instance A list of all AbstractMetric instances is accessible through the plug in singleton held by the Metricul
46. ly called if the view is run B 4 2 Model Code The checkers and model packages have a test coverage of over 80 Most units tests validate the correctness of the checker implementations which is reflected in a 95 cov erage of the checker package Ueli Kunz Julius Weder 38 December 22 2011 metriculator CDT metric plug in a 8 src EB ch hsr ifs cdt metriculator checkers 4 E3 ch hsr ifs cdt metriculator model J FlatTreeBuilder java J LogicTreeBuilder java J PreOrderFlatTreeVisitor java J PreOrderTreeVisitor java J PreOrderLogicTreeVisitor java J ScopedASTVisitor java D TreeBuilder java D TreePrinter java J AbstractMetric java D HybridTreeBuilder java J NodeFilter java J AbstractMetricChecker java 4 E3 ch hsr ifs cdt metriculator model nodes J CompositeValue java J AbstractNode java J FunctionNode java JJ NamespaceNode java J ProjectNode java J FolderNode java JJ WorkspaceNode java J Nodelnfo java D CompositeTypeNode java J FileNode java E3 ch hsr ifs cdt metriculator views E3 ch hsr ifs cdt metriculator resources E3 ch hsr ifs cdt metriculator E3 ch hsr ifs cdt metriculator tagcloud model MM E3 ch hsr ifs cdt metriculator tagcloud views MN 59 4 95 0 854 100 0 100 0 100 0 100 0 98 1 97 5 95 5 94 9 91 3 96 87 7 66 7 41 1 821 100 0 96 8 95 7 94 1 88 2 86 7 718 74 0 738 58 6 37 6 35 7 354 0 0
47. me metric names as metriculator does but they can easily be mapped to the names we used Method Lines of Code LSLOC Beside the code in the tag cloud component which we did not change details at 5 2 2 the method MetriculatorPluginActivator showMetriculator View has the most lines of code namely 40 This is because we create an anonymous class that implements an interface with plenty of methods Since there is no adapter for that interface we have to implement all the methods although we just need one The test package has 1410 lines of code All other packages views resources checkers metriculator model together count 3559 lines of code Ueli Kunz Julius Weder 36 December 22 2011 metriculator CDT metric plug in Efferent Coupling The highest efferent coupling on package level is 15 for the package ch hsr ifs metriculator checkers This package has many references to the model packages Number of Parameters The highest value is 4 which is good enough Number of Attributes Number of Members per Type The class Metriculator View has the highest value of 22 B 3 Dependecy Analysis We used Structurel01 to analyse the architecture of metriculator As visible in Fig ure B 1 Structurel01 detected eleven cyclomatic dependencies between the model and the metriculator package Nine out of this eleven dependencies are because the Ab tractMetricChecker class requires to use the plug in singleton The remaining two de
48. models and AbstractNode references as soon as Codan has finished its job 4 3 2 Scope Listeners The ScopeListener interface allows metric checkers to register themselves at Scope dAst Visitor instances to be notified if the scope changes Checkers for instance use this mechanism to report problems after function or type nodes have been analysed See Figure 4 10 to see the environment of the AbstractMetricChecker class 4 3 3 AST Visitors The ScopedA ST Visitor is responsible for visiting the AST and has a reference to the current scope node which is a AbstractNode instance The ScopedAST Visitor inherits from the AST Visitor of Codan and implements the visit and leave methods In the Scoped AS T Visitor all relevant ASTNodes of a file provided by the calling checker in stance are visited and added as nodes to the hybrid tree model Thus all logical nodes of the hybrid tree model are created with Scoped AS T Visitor classes ScopedAS T Visitor implementations like LSLOGCScopedAS T Visitor are responsible for analysing the ASTNodes relevant for the related metric ScopedA ST Visitor create the node values for the specific metric in the current scope There is always one Scope dAST Visitor associated with one AbstractMetricChecker Visitors can also notify reg istered ScopeListener instances if a scope is changing See chapter 4 3 2 for more details Figure 4 11 shows a class diagram of the ScopedA ST Visitor and related classes U
49. o E 3 2 Julius Weder This was the biggest project in which I have participated I learned a lot about working with the Eclipse Plug in Development Environment PDE as well as Eclipse CDT and creating a useful plug in I could improve my programming skills extremely not only in Java and C but also in applying all the software engineering aspects I have learned theoretical the last years It was also a great experience to manage this project in reference to project management continuous integration unit testing documentation and deployment At the beginning there were a lot things to set up to finally start with core work of this project but I can now say that it will be very helpful in future During the project I always felt highly motivated to improve the plug in as well as my skills Overall I am proud of the work we have done and it was surely a great experience which improved my professional skills and helps for future projects At this point I would also like to thank Ueli Kunz for that he is a very competent and helpful project partner It was during the hole project a good time as well as a intensive but productive teamwork Ueli Kunz Julius Weder 51 December 22 2011 metriculator CDT metric plug in List 0 1 0 2 0 3 2 1 2 2 2 3 2 4 4 1 4 2 4 3 4 4 4 5 4 6 4 T 4 8 4 9 4 10 4 11 4 12 B 1 B 2 C l C 2 C 3 C 4 C 5 C 6 CT of Figures Screenshot of
50. oe Rock RUE os a ue PES 6 2 2 1 Starting the Metrics Analysis lens 6 2 2 2 Change Metric Configuration lens 6 2 2 8 Visualise Metric as tag cloud 6 2 2 4 Analyse Metric Results e e 6 3 Metric Specification 7 3 1 Logical Source Lines of Code LSLOC oo 7 3 1 1 Lambda Expressions 22e 7 Ole DS ENU A pothole I a IA ke e LESE es 8 3 2 McCabe Cyclomatic Complexity CC 0 04 8 3 2 1 Example iu oue very e Roe Buk Poe ay X6 Rave Yeu 8 3 3 Number of Parameters per Function a 9 3 9 1 Exceptions 24 09S aha ps A ui 9 3 4 Number of Members per Type len 10 3 4 1 Explanation of Member Types llle 10 3 00 Bilerent Couplers 4 a ee eee Re Re EE AA a i 11 4 Implementation 12 41 Plugan Architecture e ul Aah Sele A eh ee ede Pee 12 Akis Plus Activator hata oh nds te Res tT a oo eor dut 13 4 2 Tree Structure 2 2 cu uo RE oe 3 koc E e RoR E Ry ee aa 13 4 2 1 TreeBuilder 4 2 2 2 2 ook us ERA LAS 14 422 Abstract Node zs i iu eae ade ee RUPES EORR RUE Res 14 ADS TTreePriter ng So A Ex Gus xc ee ee RA ved 17 4 3 MetiricsCheckers gt su VoL Ug Dan ote perit eR cm s ete hue 18 LES EM ETT PT 19 4 3 2 Scope Listeners lees 19 43 35 AST Visitors e Run a A eas 19 Ueli Kunz Julius Weder viii December 22 2011 metriculator CDT metric plug in 4 4 ProCessing m a ech ah ea BN ee go en ee ce ee ae al 20 5
51. ractMetric Y LogicTreeBuilder resetTreeBuilders showMetriculatorView JobObservable Attributes zi Operations ES amp Literals JOB ABOUT TO RUN JOB DONE Figure 4 3 Class diagram with classes related to MetriculatorPluginActivator 4 2 1 TreeBuilder As illustrated in Figure 4 4 metriculator uses three different tree models which corre spond to the three different views The hybrid tree model is used in the hybrid view the logical tree model is used in the logical view and the flat tree model is used in the filter view The TreeBuilder class is an abstract factory that is used by the checkers to create tree models AbstractMetricChecker uses a HybridTreeBuilder to add new nodes to the tree structure During the processing only the hybrid tree model is build because the hybrid tree model holds all the information that other tree models require Flat trees and logical trees are build on demand using the data of the hybrid tree model For instance when a user switches from the hybrid view to the logical or filter view FlatTreeBuilder and LogicalTreeBuilder use a PreOrderTreeVisitor to visit all nodes in the hybrid tree model and transform it to the new respective tree model The class factory method create TreeFromPath IPath is used by the Hybrid TreeBuilder to create a tree structure that represents a file system path To obtain a Logic TreeBuilder or FlatTreeBuilder instance clients use the buildF
52. rce code analysis is one type of source code investigation with the objective to get a picture of the written code itself One output of static source code analysis are source code metrics Many types of metrics exist since source code can be inspected under different aspects For instance the lines of code metric gives an impression about the size of the code Many metric analysis tools for almost any programming language already exist But at the time of this semester thesis there was no official C metric tool available that integrates well in to the Eclipse CDT C C Development Tooling platform CDT11 1 1 Motivation We intend to improve the quality of the Eclipse CDT platform as it will help other developers to create better software Code metrics provide a big picture of the quality of the source code There are plenty of well known and also standardised code measurement techniques around and implemented for other tools and languages We give our best to implement these techniques and standards for the CDT platform 1 2 Objectives Project organization Fixed one week iterations are used Redmine red11 is used for planning time tracking issue tracking and as information radiator for the super visor A project documentation is written Organization and results are reviewed weekly together with the supervisor Integration and Automation Sitting in front of a fresh Eclipse CDT installation a first semester student can install our me
53. red in advance to install the prereq uisites Before metriculator can be installed via update site users have to install CDT and the Zest framework separately The current composite update site does not work Issue 177 at http sinv 56013 edu hsr ch redmine issues 177 5 2 4 Merging of Function Declarations in Anonymous Namespaces This bugs only relates to the logical view If a function is declared in an anonymous namespace and the definition is outside of the anonymous namespace the definition does not replace the declaration Instead the declaration and the definition are deleted Issue 166 at http sinv 56013 edu hsr ch redmine issues 166 5 3 Future Work There are many interesting not yet implemented metrics that would increase the value of metriculator see chapter 1 2 1 for an incomplete list Performance can further be improved to allow analysing source code with over about 300 000 physical lines of source code Fixing that issue would make metriculator even Ueli Kunz Julius Weder 25 December 22 2011 metriculator CDT metric plug in more attractive to analyse large projects Other unresolved issues are listed in chapter 5 2 A reasonable application of metriculator in the future is to serve as refactoring assistant Based on a reported problem metriculator could suggest a refactoring or quick fix that solves that problem Ueli Kunz Julius Weder 26 December 22 2011 metriculator CDT metric plug in
54. ridId is the tree wide unique identifier of a node It is needed to identify the nodes and building the tree structures The hybrid d is build by concatenating the scope UniqueNames of all ancestor nodes and its own scopeUniqueName To understand what a scopeUniqueName is we first explain what a scopeName is The scopeName is the name of the represented underlying object For instance the file name or the signature of a function Logical nodes do not always have a name For in stance anonymous namespaces may exist at the same level Hence it would be a problem Ueli Kunz Julius Weder 15 December 22 2011 metriculator CDT metric plug in CompositeValue 3 Attributes Attributes aggregatedValue int lt children HashMap String AbstractNode gt nodeValue int metricValues HashMap AbstractMetric CompositeValue gt Operations Operations 3 copy CompositeVal getChildren ImmutableCollection lt AbstradNode gt getHybridId String getParent AbstractNode getScopeName String getScopeUniqueName String ILogicNode E Attributes Operations isAnonymous bo Figure 4 5 Class diagram with classes related to AbstractNode to identify objects by using only their scopeName because two namespaces at the same level with no name would have the same hybridld Therefore we introduced the scopeUniqueName The scopeUniqueName is build by ap pending the hash code
55. roject Wi E Efferent Coupling of Type amp Warning Property Pages Setti g Task Tags Template Default Va Customize Selected Dot m Restore Defaults Apply O ox jJ coma Figure C 3 metriculator checker preferences 3 Under Metric Problems you can enable or disable metrics you want to run see Figure C 3 4 Double click on a metric to open the metric preferences dialog 5 Specify the thresholds for the metrics see Figure C 4 6 Activate Deactivate problem reporting with problem markers see Figure C 4 If you deactivate the problem reporting the metriculator view will still highlight problematic cells but the problem will not be reported to the Eclipse Problems View That implies that no markers are created in source code editors Ueli Kunz Julius Weder 41 December 22 2011 metriculator CDT metric plug in r n 2 Customize Problem iJ McCabe Cyclomatic Complexity Edit problem preferences scope and launch options Preferences Scope Launching Y This problem is enabled Severity Warning Message Pattern Cyclomatic Complexity is above specified maximum of 0 Y Create problem markers If not active metriculator will still highlight problematic cells in the metriculator view Maximum cyclomatic complexity per function 15 Coe JC See Figure C 4 Customise Problem C 4 Working with the Views Use the
56. roject plan aa e 48 E 2 Final version of the project plan aooaa a 48 E 3 Time spent per member working on the project 50 E 4 Number of days it took to fix bugs after they were reported X axis is sorted ascending by date 22e 50 Ueli Kunz Julius Weder 53 December 22 2011 metriculator CDT metric plug in List of Tables A 1 Installed Jenkins plug ins 0 2020000000022 ae 31 Ueli Kunz Julius Weder 54 December 22 2011 metriculator CDT metric plug in F Nomenclature AST Abstract Syntax Tree An abstract representation of a program or source code usually focusing on domain specific information VCS Version Control System A software that helps managing multiple versions of files OSGi Specification for Java runtime service and modularisation platform PDE The Eclipse Plug in Development Environment provides utilities to create main tain test and build Eclipse artefacts p2 Stands fro provisioning platform and is the engine used to install plug ins and manage dependencies in Eclipse Ueli Kunz Julius Weder 55 December 22 2011 metriculator CDT metric plug in Bibliography cla11 cod11 cpp11 ecl11a ecl11b emm11 HF10 his11 jen11 jia11 jpr11 Ilv11 1s111 CDT11 Eclipse cdt project homepage http eclipse org cdt 2011 Project homepage of the clang project http clang 11vm org
57. rom TreeBuilder method 4 2 2 AbstractNode As illustrated in Figure 4 5 the AbstractNode class is a simplified composite structure to hold all relevant data the view displays Each node has one parent and any number of children An AbstractNode represents a file system object or an ASTNode The required information extracted from the ASTNode are hold in a NodelInfo instance We extract the values from the ASTNode instead of holding a reference to it so that the ASTNode Ueli Kunz Julius Weder 14 December 22 2011 setChangedAndNotifyJobAboutToRun setChangedAndNotifyJobDone metriculator CDT metric plug in 2 2 Attributes root AbstractNode Operations addChild AbstradNode createTreeFromPath IPath AbstractNode 2 FlatTreeBuilder Attributes Attributes Operations Zi Operations buildFrom TreeBuilder A buildFrom TreeBuilder Lo PreOrderTreeVisior J Attributes zi Operations visitNode AbstradNode eaveNode AbstradNode visit AbstractNode Figure 4 4 Class diagram with classes related to TreeBuilder instance can be removed by the garbage collector See chapter 5 1 for more details about related performance issues 4 2 2 1 AbstractNode Identifiers Each AbstractNode has three types of string values that name it hybridId scopeName scopeUniqueName All of them are illustrated in Figure 4 6 The hyb
58. s programming languages met11b But at the time of this project there was no metric tool available that integrates into the Eclipse C C Development Tooling platform CDT CDT11 Our goal was to create the first static source code analysis tool for CDT Procedure and Technologies metriculator is a plug in for CDT that itself is a plug in of the Eclipse framework CDT is a well known platform to develop C C software Beside CDT metriculator uses a framework called Codan cod11 to analyse C source code Codan is extensible so that third party tools such as metriculator can hook in and add their own code analysis features Metriculator analyses a user defined set of files The analysis results are applied to func tions classes namespaces files folders and projects The analysis results are viewable using different views that all rely on the same underlying data Results Metriculator supports the following five software metrics each of which is described in detail in section 3 Logical Source Lines of Code numbers the size of the software by counting the lines of code See 3 1 for detailed description Ueli Kunz Julius Weder iii December 22 2011 metriculator CDT metric plug in Cyclomatic Complexity also called McCabe is an indicator for the complexity of the software For instance many if statements increase the complexity and lead to unreadable and hard to maintain code See 3 2 for detailed description Num
59. t The following use cases apply to all implemented metrics 2 2 1 Starting the Metrics Analysis A user chooses some C projects he wants to analyse and runs the Codan command in Eclipse All calculated metrics will be displayed in the metriculator view 2 2 2 Change Metric Configuration A user can change the configuration of the Lines of Code metric by specifying an individ ual range of acceptable thresholds The view marks the values outside of the thresholds 2 2 3 Visualise Metric as tag cloud The user chooses a metric in the metriculator view Metriculator generates a tag cloud which visualises the values of the selected metric 2 2 4 Analyse Metric Results The user has run the analysis and the metriculator view is shown He can now switch between the different views sort columns or generate a tag cloud Ueli Kunz Julius Weder 6 December 22 2011 metriculator CDT metric plug in 3 Metric Specification This chapter specifies the details of the implemented metrics Most metric implementa tions follow standards specified in sub chapters with some extensions where required by C Each metric specification describes what the metric means which scopes it applies to and how it is measured 3 1 Logical Source Lines of Code LSLOC Source Lines Of Code SLOC or Lines Of Code Isl11 is a unit used to measure the size of software programs There exist different types of Lines Of Code LOC metric definitions
60. tests since they start Eclipse Integrates Git with Jenkins Integrates EMMA code coverage reports to Jenkins Displays a picture of Chuck Nor ris followed by enlightening state ments Table A 1 Installed Jenkins plug ins sudo aptitude install openjdk wget sudo apt key add sudo 6 jre etc apt sources list d jenkins list sudo aptitude update sudo aptitude install jenkins q O http pkg jenkins ci org debian jenkins ci org key sh c echo deb http pkg jenkins ci org debian binary gt Listing A 4 Jenkins set up commands Ueli Kunz Julius Weder December 22 2011 metriculator CDT metric plug in A 6 Build and Deployment Automation The CDT project supports ant and Maven as build automation platform We decided to give Maven in contribution with Tycho a try because it seemed a lot easier to maintain and has already been used in recent projects at HSR A 6 1 Set up amp Configuration for Windows Since we have never used Maven before we started using Maven on the local development environment It was not mandatory to do something locally but it simplified to get in touch with Maven for the first time We first set up the initial configuration files for our projects We followed the following steps to set up Maven on our Windows development machines e Download Maven from mavlla and extract archive to target e Add target bin to the PATH variable e Open
61. tic source code sections Optionally the detected problems will be displayed as markers with detailed problem information within the source code editors Each metric can be activated or deactivated and as well have variable threshold values that suite the needs of the specific domain There are five metrics already implemented and it is possible to extend metriculator with additional metrics without touching the existing source code Implemented met rics e Number of Logical Source Lines of Code LSLOC e Cyclomatic Complexity McCabe Number of Parameters per Function Number of Members per Type Efferent Coupling per Type Ueli Kunz Julius Weder i December 22 2011 metriculator CDT metric plug in Management Summary This chapter summarises the goals and outcomes of this thesis And gives a preview on what might be possible in the future Initial Situation Various aspects of source code can be expressed in numbers Such an aspect is called a software metric For instance a well known software metric measures the number of lines of code Software metrics allow users to view the analysed source code from different perspectives They may give a high level overview of the size and quality of the analysed source code But they can also help developers to identify problematic source code sections Static analysis tools investigate source code without running the application itself Many static analysis tools exist for variou
62. tion of the hybrid view Possible filters are files namespaces types and functions Ueli Kunz Julius Weder 43 December 22 2011 metriculator CDT metric plug in Figure C 8 Tag cloud visualisation of the nodes for a specific metric The higher the metric value the bigger the font size of the name in the cloud This tag cloud was generated for the metric LSLOC i metriculator 21 Problems 3 0 errors 3 warnings 0 others Description Resource Path Location Type 4 amp Warnings 3 items Cyclomatic Complexity is above specified maximum of 15 Example2 cpp Example src subsrc line 10 Code Analysis Problem Number of parameters is above specified maximum of4 Example2 cpp Example src subsrc line 40 Code Analysis Problem Figure C 9 Problems View 8 namespace nl g Qf Lines of code is above specified maximum of 15 4 Figure C 10 Problem marker in the source code editor with a message explaining the problem Ueli Kunz Julius Weder 44 December 22 2011 metriculator CDT metric plug in D Developer Manual Metriculator allows you to add new C metrics easily There are already some metrics implemented which might help you to implement additional metrics The following steps describe how you can add new metrics fast an simple This manual assumes that you are working with Eclipse and have installed the Plug in Development Environment PDE as well
63. tric plug in using an update site as long as metriculator is not integrated into the main CDT plug in An update site is created to allow the installation of metriculator using the Eclipse install wizard Quality The plug in code is covered with automated test cases Automated UI tests are not mandatory Delivered Assets At the end the project will be handed to the supervisor with two CDs and two paper versions of the documentation The CDs contain this project report a video demonstrating the usage of metriculator the source code and the deployable plug in Ueli Kunz Julius Weder 1 December 22 2011 metriculator CDT metric plug in Implemented metrics The metrics listed below are going to implemented with highest priority e Number of Members per Class e Lines of Code per File e Lines of Code per Class e Lines of Code per Function e Cyclomatic Complexity McCabe per Function e Number of Nested Levels e Number of Template Parameters e Number of Parameters per Function 1 2 1 Advanced Objectives If the basic objectives get finished before the end of the project we may start imple menting the following metrics Number of References of Type class struct union E Afferent Coupling Feature Envy Lack Of Cohesion per Function Number of Overloads Depth of Inheritance Tree Instability Abstractness Normalized Distance from Main Sequence Number of Locals in Scope Number of Lines of Code vs Lines of Comment
64. uncl int i 1 2 1 O nl classl memberFunc bool b 1 2 1 6 func20 ECC 18 0 O func3 int long bool int int long short 1 0 EE Figure 0 2 Screenshot of the metriculator filter view listing all functions Future Work There are many interesting not yet implemented metrics that would increase the value of metriculator see chapter 1 2 1 for an incomplete list Performance can further be improved to allow analysing source code with over about 300000 physical lines of source code Fixing that issue would make metriculator even more attractive to analyse large projects Other unresolved issues are listed in chap ter 5 2 A reasonable application of metriculator in the future is to serve as refactoring assistant Based on a reported problem metriculator could suggest a refactoring or quick fix that solves that problem Ueli Kunz Julius Weder v December 22 2011 metriculator CDT metric plug in Figure 0 3 Generated tag cloud that visualises the LSLOC value of all functions Ueli Kunz Julius Weder vi December 22 2011 metriculator CDT metric plug in Contents 1 Introduction 1 Tels Motivation 2 06 tund a a ea Bo PO ok WAG edu ont 1 1 2 Opjectives A nu hal eh A a Week i ae Sa 1 1 2 1 Advanced Objectives 2 ee 2 1 3 Project Duration en 22050 Y eds eh eo et ERR EU AE BET A 3 2 Requirements 4 2 1 metriculator view s s recs x mo kk o RS Roh Ro f Rode EUROS RR dO 4 2 2 Use Cases Brief Format
65. unction The Number of Parameters per Function metric is a software metric that counts the amount of parameters in a function or member function All parameters are counted This metric is an indicator whether a function is painful to call or maybe degrade per formance 3 3 1 Exceptions 1 If a function definition has a forward declaration the parameters of the forward declaration are not counted Following Listing 3 4 illustrates an example int doIt int bool does not count because of the definition below int doAnother int bool counts 2 parameters int doIt int i bool b counts 2 parameters return b i 0 d Listing 3 4 Code snippet for Number of Parameters per Function function declarations and definitions 2 If a member function declaration has an associated definition the parameters are not counted in the logical view but they are counted in the hybrid view expect they are in the same file then the hybrid view also does not count the function declaration Following Listing 3 5 illustrates an example class Example public void doIt int bool does not count because of the definition below void doAnother int bool counts 2 parameters F void Example doIt int i bool b counts 2 parameters do something Ueli Kunz Julius Weder 9 December 22 2011 metriculator CDT metric plug in Listing 3 5 Code snippet for Number of Parameters per Function m
66. unit of interest relies on Metriculator implements Efferent Coupling on type level That means that the more foreign types a type depends on the higher its efferent coupling value is A dependency exists if a type is directly referenced Following Listing 3 7 illustrates an example class Example2 class Example3 class Examplei 4 public void doSomething Example2 e2 counts 1 for type Example1 Example3 ex3 counts 1 to type Example 1 3 class Example2 public void doSomething Examplei ei counts 1 for type Example2 3 class Example3 3 Listing 3 7 Code snippet for Efferent Coupling Ueli Kunz Julius Weder 11 December 22 2011 metriculator CDT metric plug in 4 Implementation This chapter describes the main classes and concepts of the metriculator plug in First we give an overview of the package level architecture Then we dive into the details of the model package by commenting static and dynamic aspects At the end some performance related actions as well as known issues will be presented in detail 4 1 Plug in Architecture As visible in Figure 4 1 metriculator is based on the Codan framework Codan uses checkers to analyse source code Each checker is specialised in one problem for instance unreachable code Through extension points Codan allows third party tools to register other checkers Metriculator defines one checker per metric As soon as a user invokes the Cod
67. unz Julius Weder 42 December 22 2011 metriculator CDT metric plug in Problems ii netriculator 3 gt 3 Source Cloud amp k DOO e ig 78B McCabe LsLoc NbParams a 1 Example 16 4 g src 16 4 g subsrc 16 a c Example2 cpp 16 sOn o func20 func3 int long bool int int long short a 6 Examplel cpp On o funcl int i 9 classi e nl classl memberFunc bool b 4 Main cpp e main 8 3 a 6 1 1 1 1 1 1 1 1 Figure C 5 Hybrid view This is the default view It shows the projects and its content as a mix of physical and logical nodes in a tree structure 1 Problems i i D 3 Source Cloud amp E DOG e isi Scope McCabe LSLOC NbParams 6 mainQ 1 3 0 On 16 27 9 e funcl int i 1 2 1 4 classi 1 4 1 e nl classl memberFunc bool b 1 2 1 o func20 6 18 0 e func3 int long bool int int long short 1 0 7 Figure C 6 Logical view Shows the logical representation of the source code in a tree structure The logical elements like namespaces or classes are decoupled from the physical location and merged together I Problems 4 metriculator 23 3 Source Cloud amp fg D ere Scope McCabe LSLOC NbParams 0 main 1 3 0 O funcl int i 1 2 1 O nl classl memberFunc bool b 1 2 1 9 func20 18 0 9 func3 int long bool int int long short 1 0 E Figure C 7 Filter view Shows a filtered representa
68. us gt Other flags A 5 Continuous Integration Server We installed Jenkins 1 434 jen11 as continuous integration server Jenkins provides a solid platform for various plug ins that enhance its basic continuous integration features Our Jenkins instance requires the plug ins listed in table A 1 With Jenkins in place we can always monitor the health of our latest commits For example if some unit tests fail Jenkins will report that on the project homepage A 5 1 Set up amp Configuration To set up Jenkins we followed the set up instructions on jen11 Basically we used the following commands Ueli Kunz Julius Weder 30 December 22 2011 metriculator CDT metric plug in Name Version Description Maven 2 Project Static Analysis Utilities Static Analysis Collector FindBugs plug in Hudson Xvnc Jenkins GIT Jenkins Emma ChuckNorris 1 430 1 30 1 17 4 29 1 1 12 1 25 0 4 Allows to trigger Maven goals on build events Provides utilities for the static code analysis plug ins This plug in is an add on for the plug ins Checkstyle Dry Find Bugs PMD Tasks and Warnings the plug in collects the different analysis results and shows the re sults in a combined trend graph This plug in collects the FindBugs analysis results of the project mod ules and visualizes the found warn ings Allows projects to run Xvnc during a build Xvnc is required to run CDT plug in

- HSR - Institutional Repository

Contents

Download Pdf Manuals

Related Search

Related Contents