Buildtechs
0 likes815 views
This document summarizes the results of a large-scale empirical study on the relationship between build systems and build maintenance activity. The study analyzed over 800,000 open source projects to compare how different build technologies (e.g. Make, Autotools, Maven) affect build churn, source code coupling, and authorship over time. The key findings are that framework-based build systems tend to have higher build churn, tighter source code coupling, and decreasing authorship as projects migrate to more advanced technologies.
![The Build “Tax”
“...nothing can be said
to be certain, except
death and taxes”
- Benjamin Franklin
Up to 27% of source
changes require build
changes, too!
An Empirical Study of
Build Maintenance Effort
S. McIntosh, B. Adams, T. H. D.
Nguyen,Y. Kamei, A. E. Hassan
[ICSE 2011]
5](https://image.slidesharecdn.com/buildtechs-130312105613-phpapp02/85/Buildtechs-6-320.jpg)
![Mining a Large-
Scale Corpus
Amassing and indexing a
large sample of version
control systems
Towards the census of public
source code history
A. Mockus [MSR 2009]
843,976 source
code repositories!
13](https://image.slidesharecdn.com/buildtechs-130312105613-phpapp02/85/Buildtechs-14-320.jpg)
![Software Forges
Mining a Large-
Scale Corpus
repo.or.cz
Amassing and indexing a
large sample of version
control systems
Towards the census of public
source code history
A. Mockus [MSR 2009]
843,976 source
code repositories!
13](https://image.slidesharecdn.com/buildtechs-130312105613-phpapp02/85/Buildtechs-15-320.jpg)
![Software Forges
Mining a Large-
Scale Corpus
repo.or.cz
Amassing and indexing a
large sample of version
control systems
Towards the census of public Software Ecosystems
source code history
A. Mockus [MSR 2009]
843,976 source
code repositories!
13](https://image.slidesharecdn.com/buildtechs-130312105613-phpapp02/85/Buildtechs-16-320.jpg)
![Software Forges
Mining a Large-
Scale Corpus
repo.or.cz
Amassing and indexing a
large sample of version
control systems
Towards the census of public Software Ecosystems
source code history
A. Mockus [MSR 2009]
843,976 source
Large Projects
code repositories!
13](https://image.slidesharecdn.com/buildtechs-130312105613-phpapp02/85/Buildtechs-17-320.jpg)
Buildtechs
- 1. A Large-Scale Empirical Study of the Relationship between Build Technology and Build Maintenance Shane Meiyappan Ahmed E. Bram Audris McIntosh Nagappan Hassan Adams Mockus
- 2. What is a build system? Source code 2
- 3. What is a build system? Source Deliverable code 2
- 4. Build systems describe how sources are translated into deliverables .c .o .a .cc .o .exe .deb .tex .dvi .pdf 3
- 5. 1) Listen for 2) Build changes changes Continuous Integration: Enabled by the build system 4) Publish 3) Execute results tests 4
- 6. The Build “Tax” “...nothing can be said to be certain, except death and taxes” - Benjamin Franklin Up to 27% of source changes require build changes, too! An Empirical Study of Build Maintenance Effort S. McIntosh, B. Adams, T. H. D. Nguyen,Y. Kamei, A. E. Hassan [ICSE 2011] 5
- 7. Teams even migrate to different build technologies 6
- 8. ...But does build technology really influence build maintenance activity? 7
- 9. Low- Abstraction- Framework- Level Based Driven File-level Low-level scripts Assumes standard dependencies are generated from project layout and explicitly declared project metadata dependencies Make Rake Autotools Dependency Management Automatically manages 3rd- party API dependencies 8
- 10. Low- Abstraction- Framework- Level Based Driven File-level Low-level scripts Assumes standard dependencies are generated from project layout and explicitly declared project metadata dependencies Make Rake Autotools Dependency Management Automatically manages 3rd- party API dependencies 9
- 11. Low- Abstraction- Framework- Level Based Driven File-level Low-level scripts Assumes standard dependencies are generated from project layout and explicitly declared project metadata dependencies Make Rake Autotools Dependency Management Automatically manages 3rd- party API dependencies 10
- 12. Low- Abstraction- Framework- Level Based Driven File-level Low-level scripts Assumes standard dependencies are generated from project layout and explicitly declared project metadata dependencies Make Rake Autotools Dependency Management Automatically manages 3rd- party API dependencies 11
- 13. Low- Abstraction- Framework- Level Based Driven File-level Low-level scripts Assumes standard dependencies are generated from project layout and explicitly declared project metadata dependencies Make Rake Autotools Dependency Management Automatically manages 3rd- party API dependencies 12
- 14. Mining a Large- Scale Corpus Amassing and indexing a large sample of version control systems Towards the census of public source code history A. Mockus [MSR 2009] 843,976 source code repositories! 13
- 15. Software Forges Mining a Large- Scale Corpus repo.or.cz Amassing and indexing a large sample of version control systems Towards the census of public source code history A. Mockus [MSR 2009] 843,976 source code repositories! 13
- 16. Software Forges Mining a Large- Scale Corpus repo.or.cz Amassing and indexing a large sample of version control systems Towards the census of public Software Ecosystems source code history A. Mockus [MSR 2009] 843,976 source code repositories! 13
- 17. Software Forges Mining a Large- Scale Corpus repo.or.cz Amassing and indexing a large sample of version control systems Towards the census of public Software Ecosystems source code history A. Mockus [MSR 2009] 843,976 source Large Projects code repositories! 13
- 18. Software Forges Framework builds tend to churn most repo.or.cz Low−level Abstraction Frame Dependency 200 Build churn rate 100 0 Ant Jam Makefile Rake SCons Autotools CMake Maven Bundler Ivy 14
- 19. Source code coupling is Software Ecosystems tightest with framework and abstraction 1.2 1.0 0.8 0.6 0.4 0.2 −0.2 Low-level Framework Low-level Abstraction Low-level Abstraction Apache − Ant vs. Maven Debian − Make vs. Autotools GNU − Make vs. Autotools 15
- 20. Migration Large Projects helps! KDE 1.00 Authorship Autotools 0.75 decreases Build author ratio ●● ●● ●● ●●● ●●● ●● 0.50 ●● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●●●● ● ●● ●● ● ● ● ●●●●●●● ● ● ●● ●●●●● 0.25 ●●● ● ● ● ● ●●●●● ●● ●●● ●● ●●● 0.00 2000 2004 2008 2012 16
- 21. Migration Large Projects helps! KDE 1.00 Authorship Autotools 0.75 decreases Build author ratio ● ● ●● ●● ● ●● ● ●● ●●●● ●● ● ● ● 0.50 ● ● ● ● ●● ●● ●● ● ●● ● ●●●● ● ●● ● ● ●● ● ● ●●●●●●● ● ● ● ● ● ● ● ●● ● 0.25 ● ●● ● ● ● 0.00 ●●●●● ●●●●● ●●●●● 2000 2004 2008 2012 Teams dedicate build experts to help cope with the build tax 16