Improve custom Zubbi directives

The new Sphinx version broke our Zubbi directive implementation. The
reason is that the Sphinx log output changed and thus our custom
directives cannot parse the output properly anymore.

We use the Sphinx log to write (and later extract) the custom data from
our directives. This never felt like an appropriate solution but most
propably it was the easiest to do and we didn't find a better solution.

This change improves the Zubbi directives by storing the data in Sphinx'
domaindata dictionary which seems to be more appropriate for this use
case. This is also what other directives (including the "native" Sphinx
directives) do for storing and extracting data).

zubbi/doc.py

@@ -20,42 +20,57 @@
 import sys
 import tempfile
 
-from docutils.parsers.rst import Directive
 from readme_renderer import markdown
 from sphinx.application import Sphinx
 from sphinx.util import logging as sphinx_logging
 from sphinx.util.console import nocolor
-from sphinx.util.docutils import docutils_namespace, patch_docutils
+from sphinx.util.docutils import docutils_namespace, patch_docutils, SphinxDirective
 
 
 LOGGER = logging.getLogger(__name__)
 
 
-class ZubbiDirective(Directive):
+class ZubbiDirective(SphinxDirective):
     has_content = True
     required_arguments = 0
     optional_arguments = 1
     final_argument_whitespace = True
     logger = sphinx_logging.getLogger("ZubbiDirective")
 
-    def run(self):
 
+class SupportedOS(ZubbiDirective):
+    directive_name = "supported_os"
+
+    def run(self):
         if len(self.arguments) > 0:
             values = self.arguments[0].split(",")
-            for value in values:
-                # Log the value with directive_name for later extraction
-                self.logger.info("%s: %s", self.directive_name, value)
+            # Store the platforms in Sphinx' domain data, so we can extract
+            # them later on during the rendering process.
+            # NOTE (felix): The "correct" solution might be to define a own
+            # domain and let the domain create the initial domaindata, e.g.
+            # https://opendev.org/zuul/zuul-sphinx/src/branch/master/zuul_sphinx/zuul.py#L714
+            # However, as a simple solution, this should be sufficient.
+            zubbi_domain_data = self.env.domaindata.setdefault("zubbi", {})
+            zubbi_domain_data["platforms"] = [v.strip().lower() for v in values]
         # We don't want to render anything, so we return an empty list of nodes
         return []
 
 
-class SupportedOS(ZubbiDirective):
-    directive_name = "supported_os"
-
-
 class Reusable(ZubbiDirective):
     directive_name = "reusable"
 
+    def run(self):
+        reusable = False
+        if len(self.arguments) > 0:
+            reusable = self.arguments[0].strip().lower() in ["true", "yes"]
+        # Store the platforms in Sphinx' domain data, so we can extract
+        # them later on during the rendering process.
+        zubbi_domain_data = self.env.domaindata.setdefault("zubbi", {})
+        zubbi_domain_data["reusable"] = reusable
+
+        # We don't want to render anything, so we return an empty list of nodes
+        return []
+
 
 class SphinxBuildError(RuntimeError):
     pass
@@ -103,16 +118,11 @@ def render_sphinx(content):
             if app.statuscode:
                 raise SphinxBuildError
 
-        # Extract the platforms from the logger output
-        platforms = []
-        reusable = False
-        status_log.seek(0)
-        for line in status_log.readlines():
-            prefix, _, value = line.partition(":")
-            if prefix == SupportedOS.directive_name:
-                platforms.append(value.strip().lower())
-            elif prefix == Reusable.directive_name:
-                reusable = value.strip().lower() in ["true", "yes"]
+            # Extract the data from our custom directives from the domain data
+            zubbi_domain_data = app.env.domaindata.get("zubbi", {})
+            platforms = zubbi_domain_data.get("platforms", [])
+            reusable = zubbi_domain_data.get("reusable", False)
+
         with build_path.open() as build:
             html_parts = json.load(build)