Tenzir Test now keeps multiline record blocks intact when sorting test output, making nondeterministic event order easier to stabilize. The shell, Python fixture, and diff runners also apply pre-compare transforms consistently before comparing baselines.
🔧 Changes
Format-aware pre-compare sort
The pre-compare: sort transform now keeps multiline record blocks intact instead of sorting individual lines. Sorting a TQL test's default output no longer scrambles the fields of each event. A block runs from a { alone on a line in the first column through the matching } in the first column.
This makes it practical to stabilize non-deterministic event order without putting test scaffolding into the pipeline under test:
---
pre-compare: sort
---Tests that end in a trailing sort <field> operator solely to make output deterministic can drop that operator and declare pre-compare: sort in frontmatter or test.yaml instead.
NDJSON stays one event per line and sorts as before. Output without first-column braces also sorts line by line, unchanged. Two shapes are still not grouped and remain unsuitable for pre-compare: sort: indented structured output such as a pretty-printed JSON array, and multiline diagnostics, whose lines sort independently of their header.
The shell, Python fixture, and diff runners now also apply pre-compare transforms before comparing against the baseline; previously they accepted the option and ignored it. Baselines written with --update still record raw, untransformed output.
