Record Details

Core DUNE - Testing Results (GitLab CI)

Harvard Dataverse (Africa Rice Center, Bioversity International, CCAFS, CIAT, IFPRI, IRRI and WorldFish)

View Archive Info
 
 
Field Value
 
Title Core DUNE - Testing Results (GitLab CI)
 
Identifier https://doi.org/10.7910/DVN/DR0BGD
 
Creator Prado Lima, Jackson
 
Publisher Harvard Dataverse
 
Description Context

The Distributed and Unified Numerics Environment (DUNE) system is a modular tool for solving partial differential equations with grid-based methods. It supports the easy implementation of methods like Finite Elements, Finite Volumes, and also Finite Differences). Using C++ techniques DUNE allows one to use very different implementation of the same concept (i.e. grid, solver) under a common interface with a very low overhead. It is available at here and hosted on GitLab.


This is a dataset from the DUNE system that contains records from GitLab CI build history. The data was mined in January, 2021, and the 16,971 logs identified are available in the zip file named "logs.zip".

Content

This dataset includes records from the period between 2016/07/07 and 2021/01/23. A total of 2186 builds were included in the analysis. We discarded build logs with some problem, identified by GitLab CI, such as a problem to extract information (non-valid build log), and for that, the test cases did not execute. We identified a total of 3094 failures, and 1010 builds in which at least one test failed. Moreover, we found 293 unique test cases identified from build logs and a range of test cases executed in the builds between 1 and 134. Besides that, this dataset is also organized by system configuration (a.k.a product variant). In total, we identified 29variants that failed at least once.


See DUNE_System.pdf file for a summary of the data, variants identified, test case volatility, failures by cycle, test duration, etc.

File Format

The data in the CSV files are separated by semicolons.


The data-filtering.csv files contain detailed information about each variant/system, and features-engineered.csv contains the data parsed to be used by a learning approach. The file data-variants.csv contains information across the variants. See the tool GitLabCI-Torrent for further information.
 
Subject Computer and Information Science
 
Contributor Prado Lima, Jackson