Fix SIGHUP reload tight-loop in watch mode (#702)

* Initial plan

* Fix _reload_requested tight-loop: reset flag before reload to capture concurrent SIGHUPs

Co-authored-by: seanthegeek <44679+seanthegeek@users.noreply.github.com>
Agent-Logs-Url: https://github.com/domainaware/parsedmarc/sessions/879d0bb1-9037-41f7-bc89-f59611956d2e

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: seanthegeek <44679+seanthegeek@users.noreply.github.com>

parsedmarc/cli.py

@@ -207,8 +207,9 @@ def _parse_config_file(config_file, opts):
                     index_prefix_domain_map = yaml.safe_load(f)
             except OSError as exc:
                 raise ConfigurationError(
-                    "Failed to read index_prefix_domain_map file "
-                    "'{0}': {1}".format(map_path, exc)
+                    "Failed to read index_prefix_domain_map file '{0}': {1}".format(
+                        map_path, exc
+                    )
                 ) from exc
             except yaml.YAMLError as exc:
                 raise ConfigurationError(
@@ -254,7 +255,7 @@ def _parse_config_file(config_file, opts):
             except Exception as ns_error:
                 raise ConfigurationError(
                     "DNS pre-flight check failed: {}".format(ns_error)
-                )
+                ) from ns_error
             if not dummy_hostname:
                 raise ConfigurationError(
                     "DNS pre-flight check failed: no PTR record for {} from {}".format(
@@ -271,8 +272,6 @@ def _parse_config_file(config_file, opts):
             opts.debug = bool(general_config.getboolean("debug"))
         if "verbose" in general_config:
             opts.verbose = bool(general_config.getboolean("verbose"))
-        if "silent" in general_config:
-            opts.silent = bool(general_config.getboolean("silent"))
         if "warnings" in general_config:
             opts.warnings = bool(general_config.getboolean("warnings"))
         if "fail_on_output_error" in general_config:
@@ -906,31 +905,39 @@ def _init_output_clients(opts):
             )
 
     if opts.s3_bucket:
-        clients["s3_client"] = s3.S3Client(
-            bucket_name=opts.s3_bucket,
-            bucket_path=opts.s3_path,
-            region_name=opts.s3_region_name,
-            endpoint_url=opts.s3_endpoint_url,
-            access_key_id=opts.s3_access_key_id,
-            secret_access_key=opts.s3_secret_access_key,
-        )
+        try:
+            clients["s3_client"] = s3.S3Client(
+                bucket_name=opts.s3_bucket,
+                bucket_path=opts.s3_path,
+                region_name=opts.s3_region_name,
+                endpoint_url=opts.s3_endpoint_url,
+                access_key_id=opts.s3_access_key_id,
+                secret_access_key=opts.s3_secret_access_key,
+            )
+        except Exception:
+            logger.warning("Failed to initialize S3 client; skipping", exc_info=True)
 
     if opts.syslog_server:
-        clients["syslog_client"] = syslog.SyslogClient(
-            server_name=opts.syslog_server,
-            server_port=int(opts.syslog_port),
-            protocol=opts.syslog_protocol or "udp",
-            cafile_path=opts.syslog_cafile_path,
-            certfile_path=opts.syslog_certfile_path,
-            keyfile_path=opts.syslog_keyfile_path,
-            timeout=opts.syslog_timeout if opts.syslog_timeout is not None else 5.0,
-            retry_attempts=opts.syslog_retry_attempts
-            if opts.syslog_retry_attempts is not None
-            else 3,
-            retry_delay=opts.syslog_retry_delay
-            if opts.syslog_retry_delay is not None
-            else 5,
-        )
+        try:
+            clients["syslog_client"] = syslog.SyslogClient(
+                server_name=opts.syslog_server,
+                server_port=int(opts.syslog_port),
+                protocol=opts.syslog_protocol or "udp",
+                cafile_path=opts.syslog_cafile_path,
+                certfile_path=opts.syslog_certfile_path,
+                keyfile_path=opts.syslog_keyfile_path,
+                timeout=opts.syslog_timeout if opts.syslog_timeout is not None else 5.0,
+                retry_attempts=opts.syslog_retry_attempts
+                if opts.syslog_retry_attempts is not None
+                else 3,
+                retry_delay=opts.syslog_retry_delay
+                if opts.syslog_retry_delay is not None
+                else 5,
+            )
+        except Exception:
+            logger.warning(
+                "Failed to initialize syslog client; skipping", exc_info=True
+            )
 
     if opts.hec:
         if opts.hec_token is None or opts.hec_index is None:
@@ -940,42 +947,60 @@ def _init_output_clients(opts):
         verify = True
         if opts.hec_skip_certificate_verification:
             verify = False
-        clients["hec_client"] = splunk.HECClient(
-            opts.hec, opts.hec_token, opts.hec_index, verify=verify
-        )
+        try:
+            clients["hec_client"] = splunk.HECClient(
+                opts.hec, opts.hec_token, opts.hec_index, verify=verify
+            )
+        except Exception:
+            logger.warning(
+                "Failed to initialize Splunk HEC client; skipping", exc_info=True
+            )
 
     if opts.kafka_hosts:
         ssl_context = None
-        if opts.kafka_skip_certificate_verification:
-            logger.debug("Skipping Kafka certificate verification")
+        if opts.kafka_ssl:
             ssl_context = create_default_context()
-            ssl_context.check_hostname = False
-            ssl_context.verify_mode = CERT_NONE
-        clients["kafka_client"] = kafkaclient.KafkaClient(
-            opts.kafka_hosts,
-            username=opts.kafka_username,
-            password=opts.kafka_password,
-            ssl_context=ssl_context,
-        )
+            if opts.kafka_skip_certificate_verification:
+                logger.debug("Skipping Kafka certificate verification")
+                ssl_context.check_hostname = False
+                ssl_context.verify_mode = CERT_NONE
+        try:
+            clients["kafka_client"] = kafkaclient.KafkaClient(
+                opts.kafka_hosts,
+                username=opts.kafka_username,
+                password=opts.kafka_password,
+                ssl=opts.kafka_ssl,
+                ssl_context=ssl_context,
+            )
+        except Exception:
+            logger.warning("Failed to initialize Kafka client; skipping", exc_info=True)
 
     if opts.gelf_host:
-        clients["gelf_client"] = gelf.GelfClient(
-            host=opts.gelf_host,
-            port=int(opts.gelf_port),
-            mode=opts.gelf_mode,
-        )
+        try:
+            clients["gelf_client"] = gelf.GelfClient(
+                host=opts.gelf_host,
+                port=int(opts.gelf_port),
+                mode=opts.gelf_mode,
+            )
+        except Exception:
+            logger.warning("Failed to initialize GELF client; skipping", exc_info=True)
 
     if (
         opts.webhook_aggregate_url
         or opts.webhook_forensic_url
         or opts.webhook_smtp_tls_url
     ):
-        clients["webhook_client"] = webhook.WebhookClient(
-            aggregate_url=opts.webhook_aggregate_url,
-            forensic_url=opts.webhook_forensic_url,
-            smtp_tls_url=opts.webhook_smtp_tls_url,
-            timeout=opts.webhook_timeout,
-        )
+        try:
+            clients["webhook_client"] = webhook.WebhookClient(
+                aggregate_url=opts.webhook_aggregate_url,
+                forensic_url=opts.webhook_forensic_url,
+                smtp_tls_url=opts.webhook_smtp_tls_url,
+                timeout=opts.webhook_timeout,
+            )
+        except Exception:
+            logger.warning(
+                "Failed to initialize webhook client; skipping", exc_info=True
+            )
 
     return clients
 
@@ -1970,7 +1995,6 @@ def _main():
         logger.info("Watching for email - Quit with ctrl-c")
 
         while True:
-            _reload_requested = False
             try:
                 watch_inbox(
                     mailbox_connection=mailbox_connection,
@@ -2003,7 +2027,10 @@ def _main():
             if not _reload_requested:
                 break
 
-            # Reload configuration
+            # Reload configuration — clear the flag first so that any new
+            # SIGHUP arriving while we reload will be captured for the next
+            # iteration rather than being silently dropped.
+            _reload_requested = False
             logger.info("Reloading configuration...")
             try:
                 # Build a fresh opts starting from CLI-only defaults so that
@@ -2047,6 +2074,31 @@ def _main():
                 if opts.debug:
                     logger.setLevel(logging.DEBUG)
 
+                # Refresh FileHandler if log_file changed
+                old_log_file = getattr(opts, "active_log_file", None)
+                new_log_file = opts.log_file
+                if old_log_file != new_log_file:
+                    # Remove old FileHandlers
+                    for h in list(logger.handlers):
+                        if isinstance(h, logging.FileHandler):
+                            h.close()
+                            logger.removeHandler(h)
+                    # Add new FileHandler if configured
+                    if new_log_file:
+                        try:
+                            fh = logging.FileHandler(new_log_file, "a")
+                            file_formatter = logging.Formatter(
+                                "%(asctime)s - %(levelname)s"
+                                " - [%(filename)s:%(lineno)d] - %(message)s"
+                            )
+                            fh.setFormatter(file_formatter)
+                            logger.addHandler(fh)
+                        except Exception as log_error:
+                            logger.warning(
+                                "Unable to write to log file: {}".format(log_error)
+                            )
+                    opts.active_log_file = new_log_file
+
                 logger.info("Configuration reloaded successfully")
             except Exception:
                 logger.exception(

parsedmarc/mail/gmail.py

@@ -178,10 +178,12 @@ class GmailConnection(MailboxConnection):
     def watch(self, check_callback, check_timeout, should_reload=None):
         """Checks the mailbox for new messages every n seconds"""
         while True:
-            sleep(check_timeout)
-            check_callback(self)
             if should_reload and should_reload():
                 return
+            sleep(check_timeout)
+            if should_reload and should_reload():
+                return
+            check_callback(self)
 
     @lru_cache(maxsize=10)
     def _find_label_id_for_label(self, label_name: str) -> str:

parsedmarc/mail/graph.py

@@ -281,10 +281,10 @@ class MSGraphConnection(MailboxConnection):
     def watch(self, check_callback, check_timeout, should_reload=None):
         """Checks the mailbox for new messages every n seconds"""
         while True:
-            sleep(check_timeout)
-            check_callback(self)
             if should_reload and should_reload():
                 return
+            sleep(check_timeout)
+            check_callback(self)
 
     @lru_cache(maxsize=10)
     def _find_folder_id_from_folder_path(self, folder_name: str) -> str: