pyobs-core 2.0.0.dev33__tar.gz → 2.0.0.dev35__tar.gz

{pyobs_core-2.0.0.dev33 → pyobs_core-2.0.0.dev35}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyobs-core
-Version: 2.0.0.dev33
+Version: 2.0.0.dev35
 Summary: robotic telescope software
 Author-email: Tim-Oliver Husser <thusser@uni-goettingen.de>
 License-Expression: MIT

{pyobs_core-2.0.0.dev33 → pyobs_core-2.0.0.dev35}/pyobs/comm/comm.py

@@ -663,7 +663,27 @@ class Comm:
                 # handle it
                 ret = handler(event, from_client)
                 if asyncio.iscoroutine(ret):
-                    asyncio.create_task(ret)
+                    task = asyncio.create_task(ret)
+                    task.add_done_callback(functools.partial(self._log_handler_exception, handler=handler, event=event))
+
+    @staticmethod
+    def _log_handler_exception(task: asyncio.Task[Any], handler: Callable[..., Any], event: Event) -> None:
+        """Log an event handler's exception with context, instead of leaving it to asyncio's own
+        "Task exception was never retrieved" warning at garbage-collection time, which gives no
+        indication of which handler or event actually failed -- every event handler is dispatched
+        as a fire-and-forget task (see _send_event_to_module), so this is the only place any of
+        them get their exceptions surfaced at all.
+        """
+        if task.cancelled():
+            return
+        exc = task.exception()
+        if exc is not None:
+            log.error(
+                "Unhandled exception in event handler %s for %s.",
+                getattr(handler, "__qualname__", handler),
+                event.__class__.__name__,
+                exc_info=exc,
+            )
 
 
 __all__ = ["Comm"]

{pyobs_core-2.0.0.dev33 → pyobs_core-2.0.0.dev35}/pyobs/comm/xmpp/xmppcomm.py

@@ -844,7 +844,12 @@ class XmppComm(Comm):
             Return value from method.
         """
 
-        # try multiple times
+        # try multiple times -- unlike _get_capabilities/_subscribe_with_retry this still gives up
+        # after a fixed budget instead of retrying indefinitely, since some callers run from open()
+        # and need to know a send failed rather than hang forever (see #664/#666). But the wait
+        # between attempts is jittered for the same reason as those: many modules' _safe_send calls
+        # can end up retrying around the same moment (e.g. a fleet-wide restart), and a fixed wait
+        # keeps them all retrying in lockstep instead of spreading the load out.
         iq = None
         for i in range(self._safe_send_attempts):
             try:
@@ -857,11 +862,11 @@ class XmppComm(Comm):
             except slixmpp.exceptions.IqTimeout as timeout:
                 # timeout occurred, try again after some wait
                 iq = timeout.iq
-                await asyncio.sleep(self._safe_send_wait)
+                await asyncio.sleep(_retry_delay(i + 1, cap=self._safe_send_wait * 4, base=self._safe_send_wait))
 
             except TimeoutError:
                 # our own timeout fired instead of slixmpp's, try again after some wait
-                await asyncio.sleep(self._safe_send_wait)
+                await asyncio.sleep(_retry_delay(i + 1, cap=self._safe_send_wait * 4, base=self._safe_send_wait))
 
         # never should reach this
         raise slixmpp.exceptions.IqTimeout(iq)

{pyobs_core-2.0.0.dev33 → pyobs_core-2.0.0.dev35}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "pyobs-core"
-version = "2.0.0.dev33"
+version = "2.0.0.dev35"
 description = "robotic telescope software"
 authors = [{ name = "Tim-Oliver Husser", email = "thusser@uni-goettingen.de" }]
 requires-python = ">=3.11"