ci(translations): hard-block translation regressions in CI (#39443)

Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: Đỗ Trọng Hải <41283691+hainenber@users.noreply.github.com>

.github/workflows/superset-translations-comment.yml

@@ -0,0 +1,87 @@
+name: Translation Regression Comment
+
+on:
+  workflow_run:
+    workflows: ["Translations"]
+    types: [completed]
+
+# This workflow posts a PR comment when the Translations workflow detects a
+# regression. It uses the workflow_run trigger so that it always runs in the
+# base-branch context and can safely be granted write permissions, even for
+# PRs from forks.
+#
+# IMPORTANT: This workflow must NEVER check out code from the PR branch.
+# All data comes from the artifact uploaded by the Translations workflow.
+
+permissions:
+  pull-requests: write
+  actions: read
+
+jobs:
+  post-comment:
+    runs-on: ubuntu-24.04
+    # Only act when the Translations workflow failed (which means a regression
+    # was detected — the workflow exits 1 on regression).
+    if: github.event.workflow_run.conclusion == 'failure'
+
+    steps:
+      - name: Download regression artifact
+        id: download
+        continue-on-error: true
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
+        with:
+          name: translation-regression
+          run-id: ${{ github.event.workflow_run.id }}
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          path: /tmp/translation-regression
+
+      - name: Post or update PR comment
+        if: steps.download.outcome == 'success'
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const fs = require('fs');
+
+            const prNumberFile = '/tmp/translation-regression/pr-number.txt';
+            const reportFile = '/tmp/translation-regression/regression-report.md';
+
+            if (!fs.existsSync(prNumberFile) || !fs.existsSync(reportFile)) {
+              console.log('Artifact files not found, skipping comment.');
+              return;
+            }
+
+            const prNumber = parseInt(fs.readFileSync(prNumberFile, 'utf8').trim(), 10);
+            if (!prNumber) {
+              console.log('Could not parse PR number, skipping comment.');
+              return;
+            }
+
+            const report = fs.readFileSync(reportFile, 'utf8');
+            const marker = '<!-- translation-regression-bot -->';
+            const body = `${marker}\n${report}`;
+
+            const { data: comments } = await github.rest.issues.listComments({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: prNumber,
+            });
+
+            const existing = comments.find(c => c.body && c.body.includes(marker));
+
+            if (existing) {
+              await github.rest.issues.updateComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                comment_id: existing.id,
+                body,
+              });
+              console.log(`Updated existing comment ${existing.id} on PR #${prNumber}`);
+            } else {
+              await github.rest.issues.createComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                issue_number: prNumber,
+                body,
+              });
+              console.log(`Created new comment on PR #${prNumber}`);
+            }

.github/workflows/superset-translations.yml

@@ -20,6 +20,9 @@ concurrency:
 jobs:
   frontend-check-translations:
     runs-on: ubuntu-24.04
+    permissions:
+      contents: read
+      pull-requests: read
     steps:
       - name: "Checkout ${{ github.ref }} ( ${{ github.sha }} )"
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6
@@ -51,12 +54,16 @@ jobs:
 
   babel-extract:
     runs-on: ubuntu-24.04
+    permissions:
+      contents: read
+      pull-requests: read
     steps:
       - name: "Checkout ${{ github.ref }} ( ${{ github.sha }} )"
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6
         with:
           persist-credentials: false
           submodules: recursive
+
       - name: Check for file changes
         id: check
         uses: ./.github/actions/change-detector/
@@ -64,12 +71,85 @@ jobs:
           token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Setup Python
-        if: steps.check.outputs.python
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
         uses: ./.github/actions/setup-backend/
 
-      - name: Install msgcat
-        run: sudo apt update && sudo apt install gettext
+      - name: Install gettext tools
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        run: sudo apt-get update && sudo apt-get install -y gettext
 
-      - name: Test babel extraction
-        if: steps.check.outputs.python
+      # Fetch the base ref so we can compare PR-introduced regressions
+      # against a fair baseline (also runs babel_update against the base
+      # source) — this isolates the PR's contribution from any pre-existing
+      # drift on the base branch.
+      - name: Fetch base ref and create comparison worktree
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        run: |
+          # For PRs use the base branch; for direct pushes compare against the previous commit.
+          BASE_REF="${{ github.event.pull_request.base.ref }}"
+          if [ -n "$BASE_REF" ]; then
+            git fetch --depth=1 origin "$BASE_REF"
+          else
+            git fetch --depth=2 origin "${{ github.ref }}"
+          fi
+          git worktree add /tmp/base-worktree FETCH_HEAD
+
+      # Run babel_update against BASE source + BASE translations. Any drift
+      # already present on the base branch (source strings that have changed
+      # without .po updates) shows up here as fuzzies — and will also show
+      # up in the PR run, so it cancels out in the comparison.
+      - name: Baseline — run babel_update against BASE source
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        working-directory: /tmp/base-worktree
         run: ./scripts/translations/babel_update.sh
+
+      - name: Record baseline translation counts
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        run: |
+          python scripts/translations/check_translation_regression.py \
+            --count \
+            --translations-dir /tmp/base-worktree/superset/translations \
+            > /tmp/before.json
+
+      # Reset the PR worktree's translations to the pristine BASE state so
+      # both babel_update runs start from the same .po files. The only
+      # difference between the runs is the source code.
+      - name: Reset PR worktree translations to pristine BASE
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        run: git checkout FETCH_HEAD -- superset/translations/
+
+      - name: Run babel_update against PR source
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        run: ./scripts/translations/babel_update.sh
+
+      - name: Check for translation regression
+        id: regression
+        if: steps.check.outputs.python == 'true' || steps.check.outputs.frontend == 'true'
+        continue-on-error: true
+        run: |
+          python scripts/translations/check_translation_regression.py \
+            --compare /tmp/before.json \
+            --report /tmp/regression-report.md
+
+      # Save the PR number so the comment workflow can post the report without
+      # needing write permissions on this pull_request-triggered job.
+      - name: Save PR number for comment workflow
+        if: >-
+          github.event_name == 'pull_request' &&
+          steps.regression.outcome == 'failure'
+        run: echo "${{ github.event.pull_request.number }}" > /tmp/pr-number.txt
+
+      - name: Upload regression artifact
+        if: >-
+          github.event_name == 'pull_request' &&
+          steps.regression.outcome == 'failure'
+        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7
+        with:
+          name: translation-regression
+          path: |
+            /tmp/regression-report.md
+            /tmp/pr-number.txt
+
+      - name: Fail if regression detected
+        if: steps.regression.outcome == 'failure'
+        run: exit 1