Script for Parsing jtreg GitHub Actions Results
I was reviewing jtreg test failures in some GitHub actions last week. Since I was only interested in the failures, I decided to write a script to extract the failure details from the log files I had downloaded from GitHub. This looked like another task for the VS Code agent so I wrote the algorithm for extracting the failure details into individual text files, which would be easier to review. Below is the prompt I used.
write a python script that processes all the text files in a user-specified folder. For each text file:
- split it into sections using "--------------------------------------------------" as a separator
- generate a filename for the section from the first line after the separator. To do so, use the text after "TEST: " and replace /, #, and . with underscores and append the ".txt" extension.
- if the section contains the string "test jdk:" then write the content of the entire section to a file with the generated name.
I thought I would need to do at least some debugging of the script but I did not! The script worked flawlessly. I didn’t even need to execute it myself because the flow of using the agent included running the script on my raw folder and then generating a README file! See the script and the README at Add scripts generated by Claude Sonnet 4 VS Code agent ยท swesonga/scratchpad@f5e8057. The agent was using the Claude Sonnet 4 model.