[CI] Fix update-website.yml (emscripten-core#25796)

- Ignore changes that only include AUTHORS.html
- Use EMSCRIPTEN_BOT_TOKEN when checking out the `-site` repo since we
need that in order to push to it
- Move `-site` repo checkout to within the emscripten checkout.

Author: sbc100

.github/workflows/update-website.yml

@@ -10,7 +10,7 @@ jobs:
     name: Update website
     runs-on: ubuntu-latest
     env:
-      GH_TOKEN: ${{ secrets.EMSCRIPTEN_BOT_TOKEN }}
+      GITHUB_TOKEN: ${{ secrets.EMSCRIPTEN_BOT_TOKEN }}
     steps:
       - name: Checkout repo
         uses: actions/checkout@v4
@@ -23,16 +23,18 @@ jobs:
           repository: kripken/emscripten-site
           ref: gh-pages
           path: site/emscripten-site
+          token: ${{ secrets.EMSCRIPTEN_BOT_TOKEN }}
       - name: pip install
         run: |
           which python3
           python3 --version
           python3 -m pip install -r requirements-dev.txt
       - name: Update docs
         run: |
-          git config user.name emscripten-bot
-          git config user.email [email protected]
-          ./bootstrap
+          set -o errexit
+          set -o xtrace
+          git config --global user.name emscripten-bot
+          git config --global user.email [email protected]
           if ./tools/maint/update_docs.py; then
              echo "rebaseline_tests returned zero, expectations up-to-date"
              # Exit early and don't create a PR
@@ -46,6 +48,8 @@ jobs:
           fi
           # Create a PR against the emscripten-site repo
           cd site/emscripten-site
-          git push origin update
-          gh pr create --fill --base gh-pages --reviewer sbc100,kripken
-          gh pr merge --squash --auto
+          git push -f origin update
+          gh pr create --fill --head update --base gh-pages --reviewer sbc100,kripken
+          # TODO: add --auto here once we have the review requirment in place
+          # for PR to the -site repo
+          gh pr merge --squash # --auto

.gitignore

@@ -30,6 +30,9 @@ coverage.xml
 # Test output
 /out/
 
+# When updating the website we check it out here.
+/site/emscripten-site/
+
 # Windows ps1 launchers (created by ./tools/maint/create_entry_points.py)
 *.ps1
 # ...except the templates.

tools/maint/update_docs.py

@@ -15,11 +15,24 @@
 root_dir = os.path.dirname(os.path.dirname(script_dir))
 site_dir = os.path.join(root_dir, 'site')
 
+message_template = '''\
+Automatic update of the emscripten website
+
+This is an automated change generated by `tools/maint/update_docs.py` in the emscripten repo.
+
+The change was generated at git revision https://github.com/emscripten-core/emscripten/commit/%s
+'''
+
 
 def is_git_clean(dirname):
   return subprocess.check_output(['git', 'status', '-uno', '--porcelain'], text=True, cwd=dirname).strip() == ''
 
 
+def get_changed_files(dirname):
+  files_changed = subprocess.check_output(['git', 'status', '-uno', '--porcelain'], text=True, cwd=dirname).splitlines()
+  return  [line[3:].strip() for line in files_changed]
+
+
 def main(args):
   if args:
     site_out = args[0]
@@ -42,17 +55,21 @@ def main(args):
   # Build and install the docs
   subprocess.check_call(['make', 'install', f'EMSCRIPTEN_SITE={site_out}'], cwd=site_dir)
 
-  if is_git_clean(site_out):
+  files_changed = get_changed_files(site_dir)
+  # This AUTHORS.html file happens to always contains the current date, so we don't want
+  # to consider updates that contain only this one file
+  if 'docs/contributing/AUTHORS.html' in files_changed:
+    files_changed.remove('docs/contributing/AUTHORS.html')
+
+  if not files_changed:
     print('docs are up-to-date; no changes found')
     return 0
 
   # Create a new branch and commit the changes.
   subprocess.check_call(['git', 'checkout', '-b', 'update'], cwd=site_out)
   subprocess.check_call(['git', 'add', '.'], cwd=site_out)
-
   hash = subprocess.check_output(['git', 'rev-parse', '--short', 'HEAD'], text=True, cwd=root_dir).strip()
-  message = 'Update emscripten website\n\n'
-  message += f'These docs were generated based on git revision {hash}'
+  message = message_template % hash
   subprocess.run(['git', 'commit', '-F', '-'], input=message, text=True, cwd=site_out)
   return 2