fix: use longer timeout for initial public-inbox clone (CM-1318)

Signed-off-by: Uroš Marolt <uros@marolt.me>

Author: themarolt

services/apps/mailing_list_integration/src/crowdmail/services/mirror/mirror_service.py

@@ -18,7 +18,7 @@
 )
 from crowdmail.logger import logger
 from crowdmail.services.parse.noteren import get_email_from_git, git_run_command
-from crowdmail.settings import LORE_MIRROR_DIR
+from crowdmail.settings import CLONE_TIMEOUT_SEC, LORE_MIRROR_DIR
 
 RETRYABLE_MIRROR_ERRORS = (NetworkError, RateLimitError, RemoteServerError)
 
@@ -108,7 +108,9 @@ async def ensure_mirror(list_name: str, source_url: str, mirror_dir: str = LORE_
         clone_url = source_url.rstrip("/") + "/"
         logger.info(f"Cloning lore list '{list_name}' from {clone_url} into {list_dir}")
         os.makedirs(os.path.dirname(list_dir), exist_ok=True)
-        await _run_shell_command(["public-inbox-clone", "-q", clone_url, list_dir])
+        await _run_shell_command(
+            ["public-inbox-clone", "-q", clone_url, list_dir], timeout=CLONE_TIMEOUT_SEC
+        )
         # public-inbox-clone drops a manifest scoped to clone time; remove it
         # and re-fetch from inside list_dir so the manifest is regenerated
         # correctly there (needed to detect new epoch shards, e.g. git/1.git,

services/apps/mailing_list_integration/src/crowdmail/settings.py

@@ -46,3 +46,7 @@ def load_env_var(key: str, required=True, default=None):
 # messages instead of buffering an entire list's history in memory before one
 # flush at the end — large lore archives can have 100k+ messages.
 ACTIVITY_FLUSH_BATCH_SIZE = int(load_env_var("ACTIVITY_FLUSH_BATCH_SIZE", default="500"))
+
+# Initial public-inbox-clone of a large archive (e.g. linux-kernel) can legitimately
+# take much longer than the default command timeout; incremental fetches stay short.
+CLONE_TIMEOUT_SEC = int(load_env_var("CLONE_TIMEOUT_SEC", default="3600"))