improve handling of changelog processing for backports

.github/update-release-branch.py

@@ -1,5 +1,6 @@
 import argparse
 import datetime
+import re
 from github import Github
 import json
 import os
@@ -174,6 +175,60 @@ def get_today_string():
   today = datetime.datetime.today()
   return '{:%d %b %Y}'.format(today)
 
+def process_changelog_for_backports(source_branch_major_version, target_branch_major_version):
+
+  # changelog entries can use the following format to indicate
+  # that they only apply to newer versions
+  some_versions_only_regex = re.compile(r'\[v(\d+)\+ only\]')
+
+  output = ''
+
+  with open('CHANGELOG.md', 'r') as f:
+
+    # until we find the first section, just duplicate all lines
+    while True:
+      line = f.readline()
+      if not line:
+        raise Exception('Could not find any change sections in CHANGELOG.md') # EOF
+
+      output += line
+      if line.startswith('## '):
+        line = line.replace(f'## {source_branch_major_version}', f'## {target_branch_major_version}')
+        # we have found the first section, so now handle things differently
+        break
+
+    # found_content tracks whether we hit two headings in a row
+    found_content = False
+    output += '\n'
+    while True:
+      line = f.readline()
+      if not line:
+        break # EOF
+      line = line.rstrip('\n')
+
+      # filter out changenote entries that apply only to newer versions
+      match = some_versions_only_regex.search(line)
+      if match:
+        if int(target_branch_major_version) < int(match.group(1)):
+          continue
+
+      if line.startswith('## '):
+        line = line.replace(f'## {source_branch_major_version}', f'## {target_branch_major_version}')
+        if found_content == False:
+          # we have found two headings in a row, so we need to add the placeholder message.
+          output += 'No user facing changes.\n'
+        found_content = False
+        output += f'\n{line}\n\n'
+      else:
+        if line.strip() != '':
+          found_content = True
+          # we use the original line here, rather than the stripped version
+          # so that we preserve indentation
+          output += line + '\n'
+
+    with open('CHANGELOG.md', 'w') as f:
+      f.write(output)
+
 def update_changelog(version):
   if (os.path.exists('CHANGELOG.md')):
     content = ''
@@ -324,13 +379,7 @@ def main():
 
     # Migrate the changelog notes from vLatest version numbers to vOlder version numbers
     print(f'Migrating changelog notes from v{source_branch_major_version} to v{target_branch_major_version}')
-    subprocess.check_output(['sed', '-i', f's/^## {source_branch_major_version}\./## {target_branch_major_version}./g', 'CHANGELOG.md'])
-
-    # Remove changelog notes from all versions that do not apply to the vOlder branch
-    print(f'Removing changelog notes that do not apply to v{target_branch_major_version}')
-    for v in range(int(source_branch_major_version), int(target_branch_major_version), -1):
-      print(f'Removing changelog notes that are tagged [v{v}+ only\]')
-      subprocess.check_output(['sed', '-i', f'/^- \[v{v}+ only\]/d', 'CHANGELOG.md'])
+    process_changelog_for_backports(source_branch_major_version, target_branch_major_version)
 
     # Amend the commit generated by `npm version` to update the CHANGELOG
     run_git('add', 'CHANGELOG.md')