Software product quality control Dr. Stefan Wagner Dr. Florian Deißenböck Technische Universität München

Transcription

1 Tool-supported Software product quality control Dr. Stefan Wagner Dr. Florian Deißenböck Technische Universität München Google Developer Day Munich November 9, 2010

2 Continuous Quality Control Quality Model Quality Dashboards F. Deißenböck, Continuous Quality Control of Long-Lived Software Systems, 2009

3 Challenges Customizability, Extensibility Allow tailoring to project-specific needs for precision and relevance Aggregation and Visualization Minimize Time for result comprehension Trend Analysis Capture development over time and make improvements tangible Diversity of analyzed artifacts Allow analysis of relevant artifacts beyond code (Models, DB, ) Autonomous operation, Performance, Scalability Allow regular, automated execution for timely result data

4 ConQAT Graphical DSL to specify analysis configuration Processors perform analysis operations Edges describe data flow between processors Blocks make configuration fragments reusable

5 ConQAT ConQAT Engine Executes analysis configurations cq.edit arch.edit Clone Inspector Clone Detective for VS.NET Performs Type Checking Performs Caching (Files, ConQAT Engine Tokens, Ast Fragments, ) Runs in Batch mode Clone Detection Arch. Analysis Html Presentation ConQAT Bundles (=Plugins) Implement actual analyses as libraries of processors and blocks Can be developed and deployed independently of ConQAT

6 Architecture Conformance Analysis Goal: Discover Architecture violations Gaps in Arch. Doc. Motivation Program Comprehension Impact-Analysis Approach Documentation as development artifact Automated Conformance analysis F. Deißenböck et al., Flexible Architecture Conformance Assessment with ConQAT, 2010

7 Feilkas et al. Loss of Architectural Knowledge during System Evolution, ICPC 2009

8 Tailorable Clone Detection Java, C#, C/C++, VB, Cobol, PL/I, Words or Lines Shapers Ungapped Clones Gapped Clones Filtering, Blacklisting Metric Computation XML Report Trends Dashboard (+Visualizations)

9 Jürgens et al. CloneDetective A Workbench for Clone Detection Research, ICSE 2009

10 Clone Detection Storage load visualize tokenize & normalize extract clones find duplicates

11 Index-Based Clone Detection Google MapReduce Google BigTable Google MapReduce Source Code create/update Clone Index retrieve clones B. Hummel et al., Index-Based Code Clone Detection: Incremental, Distributed, Scalable, 2010

12 Distributed Dection Google Infrastructure: MapReduce for distributing computing tasks Bigtable for storing index 73.2 MLOC of Open Source code (C/C++/Java)* Procedure: Determine running time on 10, 20, 50, 80, 100 machines *All figures about source code in this talk refer to open source code and not to Google s code base.

13

14 Ultra-large Scale Clone Detection Code base 120 million C/C++ files 2,915,947,163 Lines of code Open source code from the index of Google Code Search Relevant use case: copyright infringement analysis Reporting all clones is useless Only the index creation was evaluated Index creation on 1000 machines took less than 7 hours real scalability

15 Visualization

16 ConQAT in Industry Development support: output integrated into project dashboards. Monitoring of SW developed by subcontractors. Quality control of software developed at multiple globally distributed sites. Clone Detection on Matlab/Simulink. Part of quality assurance criteria at quality gates. Used during one-shot quality assessments. Applied for continuous control of C++ & PL/SQL Code

17 Summary ConQAT is about integration of diverse analyses, allowing tailoring to project specific questions for continuous quality assessment ConQAT has been released under Apache License Commercial support for ConQAT: Contact: