Download presentation
Presentation is loading. Please wait.
Published byBasil Atkinson Modified over 9 years ago
1
WERST – Infrastructure Group Summary Notes July 2004 http://www.sce.carleton.ca/squall/WERST2004/
2
2 Infrastructure Issues Subject program –representative –access to programs Preparing subjects –dealing with missing information –confidentially constraints Infrastructure construction –cost versus validity Evaluating coverage criteria –data collection strategies –tools –analysis techniques Source for programs –writing your own –from industry –open source –from other researchers Reusing tools from other researchers Is infrastructure construction publishable? –General purpose –Where to publish
3
3 July 2004 Sub-Topics 1.Making experimental artifacts available 2.Tools to support experimentation 3.Tools to support experimental conduct 4.People 5.Experimental design descriptions Focused on controlled experiments, not case studies
4
4 July 2004 (1) Goals of Exp. Artifact Repository Support experimental replication Making experiments affordable Ensuring that artifacts are representative –What is our population like? Mechanism: share and reuse artifacts –Increase comparability –Decrease effort of experiments
5
5 July 2004 Types of Experimental Software Artifacts Program source files Compiled and executable modules Formal specifications Requirements Design documents Test inputs and oracles Faulty versions Data from previous experiments with the artifacts Execution environment (makefiles, shell scripts, etc.) Meta data (size, application, …)
6
6 July 2004 Experimental Software Artifacts Two Crosscutting Issues Different versions of the artifacts Different variants –Slightly different versions of the requirements –Different designs –Different implementation languages
7
7 July 2004 Repository Framework Quality of the contents –Completeness, correctness, … –Should the contents of the repository be verified to be of some level of quality? Where do the contents come from? –Various researchers –Repository “owners” –Created automatically
8
8 July 2004 Two Models for Repositories 5 star hotel – requires a proactive repository owner –Quality of artifacts have been verified by owner –Collections of artifacts “complete” –Experiments can be replicated –Artifact files should be stored with repository YMCA / youth hostel – requires a reactive administrator –Artifacts have not been verified –Collections may be incomplete, incorrect, … –Artifact files can be stored offsite Quality / Cost Quality / Quantity
9
9 July 2004 State of the Art Current distribution of artifacts Personal direct contacts –Unreliable & slow NASA’s SEL –Very difficult to find useful information NIST collection –No current support EXPSIR – Rothermel, Elbaum, Do, … –Collection of artifacts (Siemens) that are in use SERR – Alexander, Bieman, France –Under production SEEWeb – Offutt, Hayes –Repository available, not populated
10
10 July 2004 Repository Future Needs Must have community buy-in Must have a satisfactory quality –Must be widely available and accessible –Must have extremely high usability –Must be evolvable and extensible –Support must be available (repository and artifacts) –Must be well populated Must have some level of funding Community buy-in Funding Quality
11
11 July 2004 (2) Types of Experimental Tools Analysis tools –Aristotle, CBAT, SOOT, SUIF, Eclipse Dynamic information collection –InsECT, Mothra, MuJava, Proteum, ATAC, Frankl-DF?, Pure Coverage, DIE, BCEL, JIAPI, DynaInst, Atac, MuJava Test generation tools –Mothra/Godzilla, TSL tools Test drivers –JUnit Fault generators –Mutation Differencing tools
12
12 July 2004 Issues and Needs Support for the tools Usability of the tool and API Bugs in tools Unsupported language features –? operator –inheritance
13
13 July 2004 Tools Conclusions Current experimental tools are “all over the map” It can be okay to scope the problem to –Specific artifacts (in repositories) –Limited features (language, etc) It can be a very good idea to pre-compute data from some tools (ASTs, CFGs, etc) Standardizing of analysis results? Repositories and tools need to be connected
14
14 July 2004 What Holds Back Infrastructure Distribution? Lack of credit & support for building infrastructure –Funding –Publishing tools is hard –Nobody got tenure or a PhD for building experimental tools Value of experimental infrastructure is not recognized –Hard to publish replicated studies Difficulty of creating sufficiently reusable tools –Documentation –Usability –Faults in tools
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.