---
description:
globs:
alwaysApply: true
---
Logging in this project should be detailed, scoped, and extensive. Logs are the primary way to debug test runs, so they should make execution history obvious.
There should be an explicit log scope for command line output. Without -v or -vv, all other logging should be silenced from the command line.
Otherwise, each module and function should have its own log scope. Error, warn, info, and debug levels should be used appropriately.
Each invocation of persistproc should write a complete log with debug verbosity to data_dir/persistproc.run.<timestamp>.log. This may require synchronizing across threads or using multiprocessing and a queue, since we're managing multiple processes potentially using threads, but we don't necessarily need to go that far until we know we need it.
The outcome is that after every failed test run, we'll have incredible detail to see exactly what happened and find an appropriate fix that doesn't involve setting arbitrary timeouts.
Prefer scoped log objects over the global logger.
Logging is already configured in logging_utils.py.
LOG FORMAT
Debug logs should be machine-readable first, human-readable second. It should be easy to write test cases against these logs because the wording in English will not shift around.
