Track direct commits to doc-en master, not just merged PRs

About 15% of doc-en changes are pushed directly to master without
a pull request. These were invisible to the tracking workflow.

Add a second pass that fetches all commits on master from the last
7 days, filters out those already covered by a merged PR, and creates
sync issues for the remaining direct commits.

Author: lacatoire

.github/scripts/track-en-changes.sh

@@ -1,16 +1,18 @@
 #!/usr/bin/env bash
 #
-# Track merged PRs in php/doc-en and create issues in doc-fr
+# Track changes in php/doc-en and create issues in doc-fr
 # for files that need translation updates.
 #
 # How it works:
 #   - Fetches PRs merged in doc-en during the last 7 days
-#   - For each PR, checks if a matching issue already exists in doc-fr
+#   - Fetches commits pushed directly to master (not via a PR)
+#   - For each change, checks if a matching issue already exists in doc-fr
 #   - If not, creates one listing the FR files to update
 #
 # Why 7 days: the action runs daily, so 7 days gives us 6 days of margin.
 # Even if the action fails for a whole week, nothing is missed.
-# Duplicates are prevented by searching for the PR number in existing issue titles.
+# Duplicates are prevented by searching for the PR number or commit SHA
+# in existing issue titles.
 #
 set -euo pipefail
 
@@ -128,4 +130,124 @@ while read -r PR; do
 done < /tmp/prs.jsonl
 
 echo ""
-echo "Done. Created: $CREATED, Skipped (already exist): $SKIPPED"
+echo "=== Checking direct commits on master ==="
+
+# Fetch commits on master from the last 7 days (paginate if needed)
+PAGE=1
+ALL_SHAS=""
+while :; do
+  BATCH=$(gh api "repos/php/doc-en/commits?sha=master&since=$SINCE&per_page=100&page=$PAGE" \
+    --jq '.[].sha' 2>/dev/null || true)
+  COUNT=$(echo "$BATCH" | grep -c . 2>/dev/null || echo "0")
+  if [ -n "$BATCH" ]; then
+    ALL_SHAS="$ALL_SHAS"$'\n'"$BATCH"
+  fi
+  echo "  Page $PAGE: $COUNT commit(s)"
+  if [ "$COUNT" -lt 100 ]; then
+    break
+  fi
+  PAGE=$((PAGE + 1))
+done
+
+echo "$ALL_SHAS" | sed '/^$/d' > /tmp/commits.txt
+TOTAL_COMMITS=$(wc -l < /tmp/commits.txt)
+echo "Total: $TOTAL_COMMITS commit(s) on master in the last 7 days"
+
+DIRECT=0
+
+while read -r SHA; do
+  [ -z "$SHA" ] && continue
+
+  # Check if this commit is associated with a PR
+  PR_COUNT=$(gh api "repos/php/doc-en/commits/$SHA/pulls" --jq 'length' 2>/dev/null || echo "0")
+  if [ "$PR_COUNT" != "0" ]; then
+    continue
+  fi
+
+  DIRECT=$((DIRECT + 1))
+
+  # Get commit info
+  COMMIT_MSG=$(gh api "repos/php/doc-en/commits/$SHA" \
+    --jq '.commit.message | split("\n")[0]' 2>/dev/null)
+  COMMIT_DATE=$(gh api "repos/php/doc-en/commits/$SHA" \
+    --jq '.commit.author.date' 2>/dev/null | cut -dT -f1)
+  COMMIT_AUTHOR=$(gh api "repos/php/doc-en/commits/$SHA" \
+    --jq '.author.login // .commit.author.name' 2>/dev/null)
+  SHORT_SHA=${SHA:0:7}
+
+  echo ""
+  echo "Direct commit $SHORT_SHA by $COMMIT_AUTHOR: $COMMIT_MSG"
+
+  # Skip commits marked with [skip-revcheck]
+  if echo "$COMMIT_MSG" | grep -qi '\[skip-revcheck\]'; then
+    echo "  -> [skip-revcheck], skipping."
+    SKIPPED=$((SKIPPED + 1))
+    continue
+  fi
+
+  # Deduplication: search for the commit SHA in existing issues
+  EXISTING=$(gh issue list --repo "$GH_REPO" --search "\"$SHORT_SHA\"" \
+    --state all --json number --jq 'length')
+  if [ "$EXISTING" -gt 0 ]; then
+    echo "  -> Issue already exists, skipping."
+    SKIPPED=$((SKIPPED + 1))
+    continue
+  fi
+
+  # Get files changed in this commit
+  FILES=$(gh api "repos/php/doc-en/commits/$SHA" \
+    --jq '.files[].filename' 2>/dev/null || true)
+
+  if [ -z "$FILES" ]; then
+    echo "  -> No files found, skipping."
+    continue
+  fi
+
+  # Categorize files
+  UPDATE_LIST=""
+  NEW_LIST=""
+
+  while IFS= read -r FILE; do
+    if [[ "$FILE" == */versions.xml ]]; then
+      continue
+    fi
+    if [ -f "$FILE" ]; then
+      UPDATE_LIST="${UPDATE_LIST}- \`${FILE}\`"$'\n'
+    elif [[ "$FILE" == *.xml ]]; then
+      NEW_LIST="${NEW_LIST}- \`${FILE}\`"$'\n'
+    fi
+  done <<< "$FILES"
+
+  # Skip if no FR-relevant files
+  if [ -z "$UPDATE_LIST" ] && [ -z "$NEW_LIST" ]; then
+    echo "  -> No FR-relevant files, skipping."
+    continue
+  fi
+
+  # Build issue body
+  BODY="Commit: [\`$SHORT_SHA\`](https://github.com/php/doc-en/commit/$SHA) by @$COMMIT_AUTHOR ($COMMIT_DATE)"$'\n'
+
+  if [ -n "$UPDATE_LIST" ]; then
+    BODY+=$'\n'"**Fichiers FR à mettre à jour**"$'\n'
+    BODY+="$UPDATE_LIST"
+  fi
+
+  if [ -n "$NEW_LIST" ]; then
+    BODY+=$'\n'"**Nouveaux fichiers EN (pas encore traduits)**"$'\n'
+    BODY+="$NEW_LIST"
+  fi
+
+  # Create the issue
+  ISSUE_TITLE="[Sync EN] $COMMIT_MSG"
+  echo "$BODY" | gh issue create \
+    --repo "$GH_REPO" \
+    --title "$ISSUE_TITLE" \
+    --label "sync-en" \
+    --body-file -
+
+  echo "  -> Issue created."
+  CREATED=$((CREATED + 1))
+done < /tmp/commits.txt
+
+echo ""
+echo "Done. Created: $CREATED, Skipped: $SKIPPED, Direct commits found: $DIRECT"