Homework&2& Specula.ve&Analysis& Research&Projects& 10/5/14& On&bug&localiza.on& Due&Tuesday&(fake&Monday),&Oct&14,&& 9&AM&on&moodle& &

Size: px

Start display at page:

Download "Homework&2& Specula.ve&Analysis& Research&Projects& 10/5/14& On&bug&localiza.on& Due&Tuesday&(fake&Monday),&Oct&14,&& 9&AM&on&moodle& &"

Kathlyn Stewart
5 years ago
Views:

1 10/5/14& Homework&2& Specula.ve&Analysis& On&bug&localiza.on& Due&uesday&(fake&onday),&Oct&14,&& 9&A&on&moodle& & Research&Projects& Has&everyone&submiNed&& project&ideas&/&paper&selec.ons?& ednesday,&we&will&have&research&idea&presenta.ons& I&will&present&several&as&well& Students&will&have&un.l&riday&10/10&to&form&groups& 1&

2 Implement a new feature? Incorporate another developer s changes? decision making ix a bug? decision making Upgrade a library? Refactor for code reuse? Run tests? 2 / 25 2 / 25 Implement a new feature? ix a bug? Incorporate another developer s changes? decision making Developers often make decisions based on experience and intuition. Upgrade a library? Can we predict the future to help make decisions? Refactor for code reuse? Run tests? 2 / 25 Speculative analysis: predict the future and analyze it Speculative analysis: predict the future and analyze it speculate 4 / 25 4 / 25

3 Speculative analysis: predict the future and analyze it Speculative analysis: predict the future and analyze it speculate speculate 4 / 25 4 / 25 Speculative analysis: predict the future and analyze it Speculative analysis: predict the future and analyze it speculate speculate analyze execute test suite analyze execute test suite 4 / 25 inform developer # of resulting test failures 4 / 25 Speculative analysis: research questions Are there domains forspeculate which speculative analysis is possible? Quick ix Scout Can speculative analysis be made computationallyanalyze feasible? Collaborators: Kıvanç uşlu, Reid Holmes, ichael D. Ernst, and David Notkin inform developer Can speculative analysis help, and not overwhelm, developers? 4 / 25 5 / 25

Eclipse provides Quick ixes to resolve compilation errors.

6 / 25 6 / 25 e can speculatively apply each fix to find out how

6 / 25 Sometimes, local fixes cannot resolve an error.

4 Eclipse provides Quick ixes to resolve compilation errors. But Eclipse can t tell which fix is best. 6 / 25 6 / 25 e can speculatively apply each fix to find out how many errors remain. 6 / 25 Sometimes, local fixes cannot resolve an error. 6 / 25 Complex error dependencies... Speculation can discover remote fixes that resolve errors. 6 / / 25

Complex error dependencies Complex error dependencies...... http://quick-fix-scout.googlecode.

com 7 / 25 Speculative analysis for Quick ix quick fix quick fix speculate quick fix Exploring the future past version present version delta debugging continuous testing automated debugging future

5 Complex error dependencies Complex error dependencies / / 25 Speculative analysis for Quick ix quick fix quick fix speculate quick fix Exploring the future past version present version delta debugging continuous testing automated debugging future version quick fix quick fix quick fix analyze compile inform developer # of resulting compilation errors 8 / 25 9 / 25 Exploring the future Exploring the future past version present version future version past version present version future version regression testing mining software repositories continuous testing delta debugging automated debugging regression testing mining software repositories continuous testing delta debugging automated debugging Continuous development compilation [Childers et al. 2003; Eclipse 2011] execution [Henderson and eiser 1985; Karinthi and eiser 1987] testing [Sa and Ernst 2003, 2004] version control integration [Guimarães and Rito-Silva 2010] 9 / 25 9 / 25

6 Exploring the future past version mining software repositories regression testing present version delta debugging continuous testing automated debugging future version speculative analysis Proactive detection of collaboration conflicts Continuous development compilation [Childers et al. 2003; Eclipse 2011] execution [Henderson and eiser 1985; Karinthi and eiser 1987] testing [Sa and Ernst 2003, 2004] version control integration [Guimarães and Rito-Silva 2010] Speculative analysis is predictive. Collaborators: Reid Holmes, ichael D. Ernst, and David Notkin 9 / / 25 Version-control terminology Proactive conflict detection applies to both centralized and distributed version control. distributed (hg, git) centralized (cvs, svn) local commit: commit save incorporate: pull and push update and commit 11 / 25

7 h h h h h h h h h

8 h h h h he information was all there, but the developers didn t know it. hat could well-informed developers do? hat could well-informed developers do? avoid conflicts h h avoid conflicts become aware of conflicts earlier 13 / / 25 Introducing Crystal: a proactive conflict detector DEO Introducing Crystal: a proactive conflict detector DEO 14 / / 25

9 Speculative analysis in collaborative development Reducing false positives in conflict prediction local commit speculate incorporate from elinda incorporate from master incorporate to master analyze merge compile test Collaborative awareness Palantír [Sarma et al. 2003] ASDash [Biehl et al. 2007] Syde [Hattori and Lanza 2010] CollabVS [Dewan and Hegde 2007] Safe-commit [loka et al. 2009] Sourceree [Streeting 2010] inform developer collaborative relationships 15 / / 25 Reducing false positives in conflict prediction Utility of conflict detection Collaborative awareness Palantír [Sarma et al. 2003] ASDash [Biehl et al. 2007] Syde [Hattori and Lanza 2010] CollabVS [Dewan and Hegde 2007] Safe-commit [loka et al. 2009] Sourceree [Streeting 2010] Crystal analyzes concrete artifacts, eliminating false positives and false negatives. Can textual conflicts be prevented? Do build and test collaborative conflicts exist? 16 / / 25 histories of 9 open-source projects: size: 26K 1.4SLoC developers: 298 versions: 140,000 Perl5, Rails, Git, jquery, Voldemort, angos, Gallery3, Samba, Insoshi h h histories of 9 open-source projects: size: 26K 1.4SLoC developers: 298 versions: 140,000 Perl5, Rails, Git, jquery, Voldemort, angos, Gallery3, Samba, Insoshi

10 h h How frequent are textual conflicts? h h How frequent are textual conflicts? 16% of the merges have textual conflicts. h h } How frequent are textual conflicts? 16% of the merges have textual conflicts. How long do textual conflicts persist? h h } How frequent are textual conflicts? 16% of the merges have textual conflicts. How long do textual conflicts persist? Conflicts live a mean of 9.8 and median of 1.6 days. he worst case was over a year. h h } How frequent are textual conflicts? 16% of the merges have textual conflicts. How long do textual conflicts persist? Conflicts live a mean of 9.8 and median of 1.6 days. he worst case was over a year. How long do textually-safe merges persist? h h } How frequent are textual conflicts? 16% of the merges have textual conflicts. How long do textual conflicts persist? Conflicts live a mean of 9.8 and median of 1.6 days. he worst case was over a year. How long do textually-safe merges persist? extually-safe merges live a mean of 11.0 and median of 1.9 days.

11 Can textual conflicts be prevented? Can textual conflicts be prevented? here do textual conflicts come from? here do textual conflicts come from? 93% of textual conflicts developed from safe merges. 7% 93% 19 / / 25 Can textual conflicts be prevented? Do build and test collaborative conflicts exist? here do textual conflicts come from? 93% of textual conflicts developed from safe merges. 7% 93% he information Crystal computes can help prevent conflicts. program conflicts safe textual build test merges Git 17% <1% 4% 79% Perl5 8% 4% 28% 61% Voldemort 17% 10% 3% 69% Does merged code fail to build or fail tests? One in three conflicts are build or test conflicts. 19 / / 25 icrosoft Beacon A centralized version control-based tool. icrosoft product groups are using Beacon to help identify conflicts earlier in the development process. Contributions of speculative analysis past version mining software repositories regression testing present version delta debugging continuous testing automated debugging future version speculative analysis Next steps: easure Crystal s e ect on conflict frequency and persistence Evaluate qualitative e ects on user experience Identify what helps and what does not Improving developer awareness when making decisions compute precise, accurate information convert a pull mechanism to a push one Additional collaborators: Kıvanç uşlu, Christian Bird, homas Zimmermann 21 / / 25

13 Automating decision making: self-adaptation specification running system 24 / 25

14 Automating decision making: self-adaptation Automating decision making: self-adaptation specification generate adaptations potential systems specification generate adaptations potential systems running system running system observe analysis 24 / / 25 Automating decision making: self-adaptation uture research: automation specification generate adaptations potential systems specification generate adaptations potential systems running system observe running system observe employ adaptation decide analysis 1 Automating decision making: removing the developer 2 Using new automation to enrich speculative analysis employ adaptation decide analysis 3 Bridging requirement specification and behavioral model inference 24 / / 25 Jacob. Biehl, ary Czerwinski, Greg Smith, and George G. Robertson. ASDash: A visual dashboard for fostering awareness in software teams. In CHI, pages ,SanJose,CA,USA,Apr.2007.ISBN doi: / Bruce Childers, Jack. Davidson, and ary Lou So a. Continuous compilation: A new approach to aggressive and adaptive code transformation. In IPDPS, Prasun Dewan and Rajesh Hegde. Semi-synchronous conflict detection and resolution in asynchronous software development. In ECSC, pages ,Limerick,Ireland,2007. Eclipse. he Eclipse foundation ário Luís Guimarães and António Rito-Silva. owards real-time integration. In CHASE, pages 56 63, Cape own, South Africa, ay Lile Hattori and ichele Lanza. Syde: A tool for collaborative software development. In ICSE ool Demo, pages , Cape own, South Africa, ay ISBN doi: / Peter Henderson and ark eiser. Continuous execution: he VisiProg environment. In ICSE, pages 68 74, London, England, UK, Aug R. R. Karinthi and. eiser. Incremental re-execution of programs. In SII, pages 38 44, St. Paul, N, USA, June ISBN doi: / David Sa and ichael D. Ernst. Reducing wasted development time via continuous testing. In ISSRE, pages ,Denver, CO, USA, Nov ISBN David Sa and ichael D. Ernst. An experimental evaluation of continuous testing during development. In ISSA, pages76 85, Boston, A, USA, July doi: / Anita Sarma, Zahra Noroozi, and André van der Hoek. Palantír: Raising awareness among configuration management workspaces. In ICSE, pages , Portland, OR, ay ISBN X. Steve Streeting. Sourcetree Jan loka, Barbara Ryder, rank ip, and Xiaoxia Ren. Safe-commit analysis to facilitate team software development. In ICSE, pages , Vancouver, BC, Canada, ay ISBN doi: /ICSE

Mining the History of Synchronous Changes to Refine Code Ownership

Mining the History of Synchronous Changes to Refine Code Ownership Lile Hattori and Michele Lanza REVEAL@ Faculty of Informatics - University of Lugano, Switzerland Abstract When software repositories