snapshot of https://github.com/zulip/zulip.git @ 48c69911c7dd4e03b20552eb3bd77ad8151c1357 (2026-09-15)

zerver/apps.py

@@ -0,0 +1,89 @@
+import logging
+from typing import Any
+
+import django_stubs_ext
+from django.apps import AppConfig
+from django.conf import settings
+from django.core.cache import cache
+from django.core.checks import register
+from django.db import connection
+from django.db.models.signals import post_migrate, pre_migrate
+from typing_extensions import override
+
+from zerver.checks import (
+    check_auth_settings,
+    check_external_host_setting,
+    check_jitsi_server_url,
+    check_required_settings,
+    check_uploads_settings,
+)
+
+
+def flush_cache(sender: AppConfig | None, **kwargs: Any) -> None:
+    logging.info("Clearing memcached cache after migrations")
+    cache.clear()
+
+
+def create_zulip_schema(sender: AppConfig, **kwargs: Any) -> None:  # nocoverage
+    # We do this here, and not in a zerver migration, because we
+    # cannot be sure that zerver migrations run _first_, of all of the
+    # Django apps.  Updating `SET search_path` is just for the current
+    # connection; the user's default search_path is set in the
+    # zerver/0001 migrations.
+    with connection.cursor() as cursor:
+        cursor.execute("CREATE SCHEMA IF NOT EXISTS zulip")
+        cursor.execute("SET search_path = zulip,public")
+
+
+def skip_smokescreen_proxy_for_boto3() -> None:
+    # https://github.com/boto/botocore/issues/2644: the request boto3
+    # makes to fetch EC2 instance credentials from the IMDS endpoint
+    # always reads the proxy configuration from the environment,
+    # ignoring any per-client setting. Our application processes route
+    # outgoing HTTP through Smokescreen, so that metadata request would
+    # be proxied; and Smokescreen would refuse the internal
+    # IMDS address (169.254.169.254). Disable proxying for boto3 so the
+    # metadata request (and other boto3 traffic) goes direct.
+    #
+    # This must run at process startup so that it takes effect in every
+    # process, including queue workers.
+    import botocore.utils
+
+    botocore.utils.should_bypass_proxies = lambda url: True
+
+
+class ZerverConfig(AppConfig):
+    name: str = "zerver"
+
+    @override
+    def ready(self) -> None:
+        register(check_required_settings)
+        register(check_external_host_setting)
+        register(check_auth_settings)
+        register(check_uploads_settings)
+        register(check_jitsi_server_url)
+
+        if settings.SENTRY_DSN:  # nocoverage
+            from zproject.config import get_config
+            from zproject.sentry import setup_sentry
+
+            setup_sentry(settings.SENTRY_DSN, get_config("machine", "deploy_type", "development"))
+
+        from django.contrib.auth.forms import SetPasswordMixin
+
+        django_stubs_ext.monkeypatch(extra_classes=[SetPasswordMixin])
+
+        if settings.S3_SKIP_PROXY is True and settings.LOCAL_UPLOADS_DIR is None:  # nocoverage
+            skip_smokescreen_proxy_for_boto3()
+
+        # We import zerver.signals here for the side effect of
+        # registering the user_logged_in signal receiver.  This import
+        # needs to be here (rather than e.g. at top-of-file) to avoid
+        # running that code too early in Django's setup process, but
+        # in any case, this is an intentionally unused import.
+        import zerver.signals  # noqa: F401
+
+        if settings.POST_MIGRATION_CACHE_FLUSHING:
+            post_migrate.connect(flush_cache, sender=self)
+
+        pre_migrate.connect(create_zulip_schema)