pytest-multi-results-action

JavaScript GitHub Action that surfaces pytest JUnit XML results in the workflow summary. Built for multi-suite pipelines (e.g., chroot, QEMU, cloud), with configurable columns and focused details.

Fork of the excellent pytest-results-action by @pmeier, extended for multi-suite aggregation and metadata.

Features

• Multi-suite aggregation across multiple XML files or patterns
• Compact summary and per-suite results table with configurable result columns
• Optional details section focused on selected result types (e.g., failed, error)
• Metadata support via pytest-metadata and filename fallback

Quick start

Ensure pytest emits JUnit XML, then point the action at the files:

- name: Download test artifacts
  uses: actions/download-artifact@v4
  with:
    pattern: "*test*"
    path: test-artifacts
    merge-multiple: true

- name: Test Report
  if: always()
  uses: gardenlinux/pytest-multi-results-action@main
  with:
    files: |
      test-artifacts/**/results.xml
      test-artifacts/**/*.test.xml
    title: "Test Results"
    summary: true
    details: true
    result-types: "passed,skipped,failed,error"
    details-result-types: "failed,error"
    fail-on-empty: false

Inputs

Name	Description	Required	Default
files	JUnit XML file patterns (file, dir, or glob). Multi-suite aggregation supported	Yes	-
title	Section title	No	Test results
summary	Include top-level test summary	No	true
result-types	CSV result categories shown in summary/table	No	passed,skipped,xfailed,failed,xpassed,error
details	Show details section	No	true
details-result-types	CSV result categories shown in details	No	failed,error
metadata-fields	CSV metadata fields to show as columns (order preserved). Defaults to Suite	No	""
metadata-field-mapping	JSON map of metadata field → display name	No	{}
fail-on-empty	Fail if no XML found	No	true

Allowed result categories: passed, skipped, xfailed, failed, xpassed, error.

Metadata via pytest-metadata

This action can display extra columns (e.g., Artifact, Type, Namespace) by reading metadata embedded in the JUnit XML. The simplest way to inject these is with the pytest-metadata plugin.

Install and emit metadata:

pip install pytest-metadata

pytest \
  --metadata Artifact Artifact1 \
  --metadata Type Type1 \
  --metadata Namespace Namespace1 \
  --junit-xml=results.xml

Then configure which fields to show in the table and optionally map display names:

with:
  files: test-artifacts/**/*.xml
  metadata-fields: "Namespace,Type,Artifact"
  metadata-field-mapping: '{"Artifact": "Test Artifact", "Type": "Test Type", "Namespace": "Namespace"}'

Notes:

• The action reads pytest-metadata properties from the XML and makes them available as columns.
• If some fields are missing, it falls back to deriving basic info from filenames when possible.

Examples

Basic

- name: Test Report
  if: always()
  uses: gardenlinux/pytest-multi-results-action@main
  with:
    files: test-artifacts/**/*.xml
    summary: true
    details: true
    result-types: "failed,error"
    details-result-types: "failed,error"

With metadata columns

- name: Test Report (with metadata)
  if: always()
  uses: gardenlinux/pytest-multi-results-action@main
  with:
    files: test-artifacts/**/*.xml
    title: "Test Results"
    summary: true
    details: true
    result-types: "passed,skipped,failed,error"
    details-result-types: "failed,error"
    metadata-fields: "Namespace,Type,Artifact"
    metadata-field-mapping: '{"Artifact": "Test Artifact", "Type": "Test Type", "Namespace": "Namespace"}'
    fail-on-empty: false

Output

Posted to the workflow summary:

1. Test Summary — aggregated counts across suites (for selected result-types).
2. Test Results — per-suite table (metadata columns + Total + selected result-types + Duration). Failed/Error counts link to details.
3. Test Details — per-suite sections for selected details-result-types that have items. Each type capped to keep the report concise.

Examples

• Basic
• With metadata columns

Notes

• Keep summaries under GitHub's size limit (1 MB). Prefer showing only failed/error in details; upload full logs/artifacts as needed.
• If metadata is absent, the action falls back to extracting info from filenames.

Development

npm run format
npm run test
npm run test-metadata
npm run build

Credits

Forked from pytest-results-action by @pmeier.