env: maintain tracked page identifier cache between builds

When a cache is kept for tracking previously published pages and their
identifiers, if a processing event results in pages not needing to be
updated, unchanged pages can cause their page identifiers to be removed
from the active list. This can result in a following pass to cleanup
legacy-flagged pages for pages that are still considered unchanged.

The cache implementation should be maintaining the assumed cached
identifier for pages if they remain changed for a run. Updating the
implementation to always populate the cached page identifiers before
replacing them with new active page identifiers (if any).

Signed-off-by: James Knight <git@jdknight.me>

Author: jdknight

sphinxcontrib/confluencebuilder/env.py

@@ -210,6 +210,9 @@ def save_cache(self):
         new_dochashs = dict(self._cached_dochash)
         new_dochashs.update(self._active_dochash)
 
+        new_pids = dict(self._cached_pids)
+        new_pids.update(self._active_pids)
+
         try:
             with self._cache_cfg_file.open('w', encoding='utf-8') as f:
                 json.dump(new_cfg, f)
@@ -224,6 +227,6 @@ def save_cache(self):
 
         try:
             with self._cache_publish_file.open('w', encoding='utf-8') as f:
-                json.dump(self._active_pids, f)
+                json.dump(new_pids, f)
         except OSError as e:
             self.builder.warn('failed to save cache (pids): ' + e)