LogChunks: 797 Travis CI Build Logs with Failure Reason Labels

797 Travis CI build logs collected from 80 GitHub repositories across 29 programming languages. The dataset was created by Carolin Brandt of Delft University of Technology and includes manual annotations for the parts of logs describing build failures. These failure chunks are annotated with search keywords and categorized by their structural representation within the log.

Use Cases

• Train models to automatically identify failure reasons in CI logs based on manually labeled log chunks.
• Develop search and retrieval systems for build failures using the annotated keyword tags.
• Analyze patterns in log structure across different programming languages based on the provided categorization.
• Benchmark log parsing and summarization techniques against a manually annotated corpus.

Strengths

• Includes 797 real-world build logs from Travis CI.
• Logs are sourced from 80 distinct GitHub repositories representing 29 different programming languages.
• Features manual annotation of failure-relevant log chunks, keywords, and structural categories.

Limitations

• Row count and file size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Delft University of Technology

Collection Method

Logs were collected from Travis CI using a custom tool, and failure reasons were manually labeled.