parsl 2025.9.22__tar.gz → 2025.10.6__tar.gz

{parsl-2025.9.22/parsl.egg-info → parsl-2025.10.6}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: parsl
-Version: 2025.9.22
+Version: 2025.10.6
 Summary: Simple data dependent workflows in Python
 Home-page: https://github.com/Parsl/parsl
-Download-URL: https://github.com/Parsl/parsl/archive/2025.09.22.tar.gz
+Download-URL: https://github.com/Parsl/parsl/archive/2025.10.06.tar.gz
 Author: The Parsl Team
 Author-email: parsl@googlegroups.com
 License: Apache 2.0
@@ -11,11 +11,10 @@ Keywords: Workflows,Scientific computing
 Classifier: Development Status :: 5 - Production/Stable
 Classifier: Intended Audience :: Developers
 Classifier: License :: OSI Approved :: Apache Software License
-Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Python: >=3.9.0
+Requires-Python: >=3.10.0
 Provides-Extra: monitoring
 Provides-Extra: visualization
 Provides-Extra: aws

{parsl-2025.9.22 → parsl-2025.10.6}/README.rst

@@ -1,6 +1,6 @@
 Parsl - Parallel Scripting Library
 ==================================
-|licence| |docs| |NSF-1550588| |NSF-1550476| |NSF-1550562| |NSF-1550528| |NumFOCUS| |CZI-EOSS| 
+|licence| |docs| |NSF-1550588| |NSF-1550476| |NSF-1550562| |NSF-1550528| |NumFOCUS| |CZI-EOSS| |paper| 
 
 Parsl extends parallelism in Python beyond a single computer.
 
@@ -67,6 +67,9 @@ then explore the `parallel computing patterns <https://parsl.readthedocs.io/en/s
 .. |NumFOCUS| image:: https://img.shields.io/badge/powered%20by-NumFOCUS-orange.svg?style=flat&colorA=E1523D&colorB=007D8A
     :target: https://numfocus.org
     :alt: Powered by NumFOCUS
+.. |paper| image:: https://img.shields.io/badge/Software_DOI-10.1145/3307681.3325400-blue
+    :target: https://doi.org/10.1145/3307681.3325400
+    :alt: Conference proceedings paper
 
    
 Quickstart
@@ -120,7 +123,7 @@ For Developers
 Requirements
 ============
 
-Parsl is supported in Python 3.9+. Requirements can be found `here <requirements.txt>`_. Requirements for running tests can be found `here <test-requirements.txt>`_.
+Parsl is supported in Python 3.10+. Requirements can be found `here <requirements.txt>`_. Requirements for running tests can be found `here <test-requirements.txt>`_.
 
 Code of Conduct
 ===============

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/benchmark/perf.py

@@ -2,32 +2,45 @@ import argparse
 import concurrent.futures
 import importlib
 import time
+from typing import Any, Dict
 
 import parsl
+from parsl.dataflow.dflow import DataFlowKernel
 
 min_iterations = 2
 
 
 # TODO: factor with conftest.py where this is copy/pasted from?
-def load_dfk_from_config(filename):
+def load_dfk_from_config(filename: str) -> DataFlowKernel:
     spec = importlib.util.spec_from_file_location('', filename)
+
+    if spec is None:
+        raise RuntimeError("Could not import configuration")
+
+    module = importlib.util.module_from_spec(spec)
+
+    if spec.loader is None:
+        raise RuntimeError("Could not load configuration")
+
+    spec.loader.exec_module(module)
+
     module = importlib.util.module_from_spec(spec)
     spec.loader.exec_module(module)
 
     if hasattr(module, 'config'):
-        parsl.load(module.config)
+        return parsl.load(module.config)
     elif hasattr(module, 'fresh_config'):
-        parsl.load(module.fresh_config())
+        return parsl.load(module.fresh_config())
     else:
         raise RuntimeError("Config module does not define config or fresh_config")
 
 
 @parsl.python_app
-def app(extra_payload, parsl_resource_specification={}):
+def app(extra_payload: Any, parsl_resource_specification: Dict = {}) -> int:
     return 7
 
 
-def performance(*, resources: dict, target_t: float, args_extra_size: int):
+def performance(*, resources: dict, target_t: float, args_extra_size: int) -> None:
     n = 10
 
     delta_t: float
@@ -82,6 +95,7 @@ Example usage: python -m parsl.benchmark.perf --config parsl/tests/configs/workq
     parser.add_argument("--resources", metavar="EXPR", help="parsl_resource_specification dictionary")
     parser.add_argument("--time", metavar="SECONDS", help="target number of seconds for an iteration", default=120, type=float)
     parser.add_argument("--argsize", metavar="BYTES", help="extra bytes to add into app invocation arguments", default=0, type=int)
+    parser.add_argument("--version", action="version", version=f"parsl-perf from Parsl {parsl.__version__}")
 
     args = parser.parse_args()
 
@@ -90,10 +104,9 @@ Example usage: python -m parsl.benchmark.perf --config parsl/tests/configs/workq
     else:
         resources = {}
 
-    load_dfk_from_config(args.config)
-    performance(resources=resources, target_t=args.time, args_extra_size=args.argsize)
-    print("Cleaning up DFK")
-    parsl.dfk().cleanup()
+    with load_dfk_from_config(args.config):
+        performance(resources=resources, target_t=args.time, args_extra_size=args.argsize)
+        print("Tests complete - leaving DFK block")
     print("The end")
 
 

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/dataflow/dflow.py

@@ -175,7 +175,7 @@ class DataFlowKernel:
         else:
             checkpoint_files = []
 
-        self.memoizer = Memoizer(self, memoize=config.app_cache, checkpoint_files=checkpoint_files)
+        self.memoizer = Memoizer(memoize=config.app_cache, checkpoint_files=checkpoint_files)
         self.checkpointed_tasks = 0
         self._checkpoint_timer = None
         self.checkpoint_mode = config.checkpoint_mode
@@ -375,6 +375,7 @@ class DataFlowKernel:
                 logger.info("Task {} failed due to dependency failure so skipping retries".format(task_id))
                 task_record['time_returned'] = datetime.datetime.now()
                 self._send_task_log_info(task_record)
+                self.memoizer.update_memo(task_record)
                 with task_record['app_fu']._update_lock:
                     task_record['app_fu'].set_exception(e)
 
@@ -400,6 +401,7 @@ class DataFlowKernel:
                 self.update_task_state(task_record, States.failed)
                 task_record['time_returned'] = datetime.datetime.now()
                 self._send_task_log_info(task_record)
+                self.memoizer.update_memo(task_record)
                 with task_record['app_fu']._update_lock:
                     task_record['app_fu'].set_exception(e)
 
@@ -446,6 +448,7 @@ class DataFlowKernel:
                         self.update_task_state(task_record, States.failed)
                         task_record['time_returned'] = datetime.datetime.now()
                         self._send_task_log_info(task_record)
+                        self.memoizer.update_memo(task_record)
                         with task_record['app_fu']._update_lock:
                             task_record['app_fu'].set_exception(
                                 TypeError(f"join_app body must return a Future or list of Futures, got {joinable} of type {type(joinable)}"))
@@ -521,6 +524,7 @@ class DataFlowKernel:
 
                 self.update_task_state(task_record, States.failed)
                 task_record['time_returned'] = datetime.datetime.now()
+                self.memoizer.update_memo(task_record)
                 with task_record['app_fu']._update_lock:
                     task_record['app_fu'].set_exception(e)
 
@@ -561,8 +565,6 @@ class DataFlowKernel:
         if not task_record['app_fu'] == future:
             logger.error("Internal consistency error: callback future is not the app_fu in task structure, for task {}".format(task_id))
 
-        self.memoizer.update_memo(task_record, future)
-
         # Cover all checkpointing cases here:
         # Do we need to checkpoint now, or queue for later,
         # or do nothing?
@@ -591,6 +593,7 @@ class DataFlowKernel:
         logger.info(f"Task {task_record['id']} completed ({old_state.name} -> {new_state.name})")
         task_record['time_returned'] = datetime.datetime.now()
 
+        self.memoizer.update_memo(task_record)
         with task_record['app_fu']._update_lock:
             task_record['app_fu'].set_result(result)
 
@@ -1202,7 +1205,6 @@ class DataFlowKernel:
 
         self.log_task_states()
 
-        # Checkpointing takes priority over the rest of the tasks
         # checkpoint if any valid checkpoint method is specified
         if self.checkpoint_mode is not None:
             self.checkpoint()

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/dataflow/memoization.py

@@ -4,20 +4,16 @@ import hashlib
 import logging
 import os
 import pickle
+import types
+from concurrent.futures import Future
 from functools import lru_cache, singledispatch
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence
+from typing import Any, Dict, List, Optional, Sequence
 
 import typeguard
 
 from parsl.dataflow.errors import BadCheckpoint
 from parsl.dataflow.taskrecord import TaskRecord
 
-if TYPE_CHECKING:
-    from parsl import DataFlowKernel  # import loop at runtime - needed for typechecking - TODO turn into "if typing:"
-
-import types
-from concurrent.futures import Future
-
 logger = logging.getLogger(__name__)
 
 
@@ -150,17 +146,13 @@ class Memoizer:
 
     """
 
-    def __init__(self, dfk: DataFlowKernel, *, memoize: bool = True, checkpoint_files: Sequence[str]):
+    def __init__(self, *, memoize: bool = True, checkpoint_files: Sequence[str]):
         """Initialize the memoizer.
 
-        Args:
-            - dfk (DFK obj): The DFK object
-
         KWargs:
             - memoize (Bool): enable memoization or not.
             - checkpoint (Dict): A checkpoint loaded as a dict.
         """
-        self.dfk = dfk
         self.memoize = memoize
 
         checkpoint = self.load_checkpoints(checkpoint_files)
@@ -242,16 +234,14 @@ class Memoizer:
         assert isinstance(result, Future) or result is None
         return result
 
-    def update_memo(self, task: TaskRecord, r: Future[Any]) -> None:
+    def update_memo(self, task: TaskRecord) -> None:
         """Updates the memoization lookup table with the result from a task.
+        This doesn't move any values around but associates the memoization
+        hashsum with the completed (by success or failure) AppFuture.
 
         Args:
-             - task (dict) : A task dict from dfk.tasks
-             - r (Result future): Result future
+             - task (TaskRecord) : A task record from dfk.tasks
         """
-        # TODO: could use typeguard
-        assert isinstance(r, Future)
-
         task_id = task['id']
 
         if not self.memoize or not task['memoize'] or 'hashsum' not in task:
@@ -265,7 +255,7 @@ class Memoizer:
             logger.info(f"Replacing app cache entry {task['hashsum']} with result from task {task_id}")
         else:
             logger.debug(f"Storing app cache entry {task['hashsum']} with result from task {task_id}")
-        self.memo_lookup_table[task['hashsum']] = r
+        self.memo_lookup_table[task['hashsum']] = task['app_fu']
 
     def _load_checkpoints(self, checkpointDirs: Sequence[str]) -> Dict[str, Future[Any]]:
         """Load a checkpoint file into a lookup table.

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/executor.py

@@ -160,6 +160,12 @@ GENERAL_HTEX_PARAM_DOCS = """provider : :class:`~parsl.providers.base.ExecutionP
 """  # Documentation for params used by both HTEx and MPIEx
 
 
+class HTEXFuture(Future):
+    def __init__(self, task_id) -> None:
+        super().__init__()
+        self.parsl_executor_task_id = task_id
+
+
 class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageInformation):
     __doc__ = f"""Executor designed for cluster-scale
 
@@ -237,7 +243,7 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
     @typeguard.typechecked
     def __init__(self,
                  label: str = 'HighThroughputExecutor',
-                 provider: ExecutionProvider = LocalProvider(),
+                 provider: Optional[ExecutionProvider] = None,
                  launch_cmd: Optional[str] = None,
                  interchange_launch_cmd: Optional[Sequence[str]] = None,
                  address: Optional[str] = None,
@@ -267,7 +273,9 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
 
         logger.debug("Initializing HighThroughputExecutor")
 
-        BlockProviderExecutor.__init__(self, provider=provider, block_error_handler=block_error_handler)
+        BlockProviderExecutor.__init__(self,
+                                       provider=provider if provider else LocalProvider(),
+                                       block_error_handler=block_error_handler)
         self.label = label
         self.worker_debug = worker_debug
         self.storage_access = storage_access
@@ -501,10 +509,7 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
             else:
 
                 for serialized_msg in msgs:
-                    try:
-                        msg = pickle.loads(serialized_msg)
-                    except pickle.UnpicklingError:
-                        raise BadMessage("Message received could not be unpickled")
+                    msg = pickle.loads(serialized_msg)
 
                     if msg['type'] == 'result':
                         try:
@@ -671,7 +676,7 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
                 logger.debug("Sending hold to manager: {}".format(manager['manager']))
                 self._hold_manager(manager['manager'])
 
-    def submit(self, func, resource_specification, *args, **kwargs):
+    def submit(self, func: Callable, resource_specification: dict, *args, **kwargs) -> HTEXFuture:
         """Submits work to the outgoing_q.
 
         The outgoing_q is an external process listens on this
@@ -692,32 +697,83 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
 
         self.validate_resource_spec(resource_specification)
 
-        if self.bad_state_is_set:
-            raise self.executor_exception
-
-        self._task_counter += 1
-        task_id = self._task_counter
-
         # handle people sending blobs gracefully
         if logger.getEffectiveLevel() <= logging.DEBUG:
             args_to_print = tuple([ar if len(ar := repr(arg)) < 100 else (ar[:100] + '...') for arg in args])
             logger.debug("Pushing function {} to queue with args {}".format(func, args_to_print))
 
-        fut = Future()
-        fut.parsl_executor_task_id = task_id
-        self.tasks[task_id] = fut
-
         try:
             fn_buf = pack_apply_message(func, args, kwargs, buffer_threshold=1 << 20)
         except TypeError:
             raise SerializationError(func.__name__)
 
-        msg = {"task_id": task_id, "resource_spec": resource_specification, "buffer": fn_buf}
+        context = {}
+        if resource_specification:
+            context["resource_spec"] = resource_specification
+
+        return self.submit_payload(context, fn_buf)
+
+    def submit_payload(self, context: dict, buffer: bytes) -> HTEXFuture:
+        """
+        Submit specially crafted payloads.
+
+        For use-cases where the ``HighThroughputExecutor`` consumer needs the payload
+        handled by the worker in a special way.  For example, if the function is
+        serialized differently than Parsl's default approach, or if the task must
+        be setup more precisely than Parsl's default ``execute_task`` allows.
+
+        An example interaction:
+
+        .. code-block: python
+
+            >>> htex: HighThroughputExecutor  # setup prior to this example
+            >>> ctxt = {
+            ...   "task_executor": {
+            ...     "f": "full.import.path.of.custom_execute_task",
+            ...     "a": ("additional", "arguments"),
+            ...     "k": {"some": "keyword", "args": "here"}
+            ...   }
+            ... }
+            >>> fn_buf = custom_serialize(task_func, *task_args, **task_kwargs)
+            >>> fut = htex.submit_payload(ctxt, fn_buf)
+
+        The custom ``custom_execute_task`` would be dynamically imported, and
+        invoked as:
+
+        .. code-block: python
+
+            args = ("additional", "arguments")
+            kwargs = {"some": "keyword", "args": "here"}
+            result = custom_execute_task(fn_buf, *args, **kwargs)
+
+        Parameters
+        ----------
+        context:
+            A task-specific context associated with the function buffer.  Parsl
+            currently implements the keys ``task_executor`` and ``resource_spec``
+
+        buffer:
+            A serialized function, that will be deserialized and executed by
+            ``execute_task`` (or custom function, if ``task_executor`` is specified)
+
+        Returns
+        -------
+        An HTEXFuture (a normal Future, with the attribute ``.parsl_executor_task_id``
+        set).  The future will be set to done when the associated function buffer has
+        been invoked and completed.
+        """
+        if self.bad_state_is_set:
+            raise self.executor_exception
+
+        self._task_counter += 1
+        task_id = self._task_counter
+
+        fut = HTEXFuture(task_id)
+        self.tasks[task_id] = fut
 
-        # Post task to the outgoing queue
+        msg = {"task_id": task_id, "context": context, "buffer": buffer}
         self.outgoing_q.put(msg)
 
-        # Return the future
         return fut
 
     @property

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/interchange.py

@@ -332,7 +332,7 @@ class Interchange:
             msg = self.task_incoming.recv_pyobj()
 
             # Process priority, higher number = lower priority
-            resource_spec = msg.get('resource_spec', {})
+            resource_spec = msg['context'].get('resource_spec', {})
             priority = resource_spec.get('priority', float('inf'))
             queue_entry = (-priority, -self.task_counter, msg)
 
@@ -360,9 +360,10 @@ class Interchange:
             mtype = meta['type']
         except Exception as e:
             logger.warning(
-                f'Failed to read manager message ([{type(e).__name__}] {e})'
+                'Failed to read manager message; ignoring message'
+                f' (Exception: [{type(e).__name__}] {e})'
             )
-            logger.debug('Message:\n   %r\n', msg_parts, exc_info=e)
+            logger.debug('Raw message bytes:\n   %r\n', msg_parts, exc_info=e)
             return
 
         logger.debug(

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/mpi_executor.py

@@ -16,7 +16,6 @@ from parsl.executors.status_handling import BlockProviderExecutor
 from parsl.jobs.states import JobStatus
 from parsl.launchers import SimpleLauncher
 from parsl.monitoring.radios.base import RadioConfig
-from parsl.providers import LocalProvider
 from parsl.providers.base import ExecutionProvider
 
 
@@ -47,7 +46,7 @@ class MPIExecutor(HighThroughputExecutor):
     @typeguard.typechecked
     def __init__(self,
                  label: str = 'MPIExecutor',
-                 provider: ExecutionProvider = LocalProvider(),
+                 provider: Optional[ExecutionProvider] = None,
                  launch_cmd: Optional[str] = None,
                  interchange_launch_cmd: Optional[str] = None,
                  address: Optional[str] = None,

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/mpi_resource_management.py

@@ -166,11 +166,10 @@ class MPITaskScheduler(TaskScheduler):
 
     def put_task(self, task_package: dict):
         """Schedule task if resources are available otherwise backlog the task"""
-        resource_spec = task_package.get("resource_spec", {})
+        resource_spec = task_package.get("context", {}).get("resource_spec", {})
 
-        nodes_needed = resource_spec.get("num_nodes")
-        tid = task_package["task_id"]
-        if nodes_needed:
+        if nodes_needed := resource_spec.get("num_nodes"):
+            tid = task_package["task_id"]
             try:
                 allocated_nodes = self._get_nodes(nodes_needed)
             except MPINodesUnavailable:

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/process_worker_pool.py

@@ -1,6 +1,7 @@
 #!/usr/bin/env python3
 
 import argparse
+import importlib
 import logging
 import math
 import multiprocessing
@@ -17,7 +18,7 @@ from importlib.metadata import distributions
 from multiprocessing.context import SpawnProcess
 from multiprocessing.managers import DictProxy
 from multiprocessing.sharedctypes import Synchronized
-from typing import Dict, List, Optional, Sequence
+from typing import Callable, Dict, List, Optional, Sequence
 
 import psutil
 import zmq
@@ -373,6 +374,8 @@ class Manager:
             if socks.get(ix_sock) == zmq.POLLIN:
                 pkl_msg = ix_sock.recv()
                 tasks = pickle.loads(pkl_msg)
+                del pkl_msg
+
                 last_interchange_contact = time.time()
 
                 if tasks == HEARTBEAT_CODE:
@@ -454,6 +457,7 @@ class Manager:
                                               'exception': serialize(RemoteExceptionWrapper(*sys.exc_info()))}
                             pkl_package = pickle.dumps(result_package)
                             self.pending_result_queue.put(pkl_package)
+                            del pkl_package
                     except KeyError:
                         logger.info("Worker {} was not busy when it died".format(worker_id))
 
@@ -770,17 +774,33 @@ def worker(
             ready_worker_count.value -= 1
         worker_enqueued = False
 
-        _init_mpi_env(mpi_launcher=mpi_launcher, resource_spec=req["resource_spec"])
+        ctxt = req["context"]
+        res_spec = ctxt.get("resource_spec", {})
+
+        _init_mpi_env(mpi_launcher=mpi_launcher, resource_spec=res_spec)
+
+        exec_func: Callable = execute_task
+        exec_args = ()
+        exec_kwargs = {}
 
         try:
-            result = execute_task(req['buffer'])
+            if task_executor := ctxt.get("task_executor", None):
+                mod_name, _, fn_name = task_executor["f"].rpartition(".")
+                exec_mod = importlib.import_module(mod_name)
+                exec_func = getattr(exec_mod, fn_name)
+
+                exec_args = task_executor.get("a", ())
+                exec_kwargs = task_executor.get("k", {})
+
+            result = exec_func(req['buffer'], *exec_args, **exec_kwargs)
             serialized_result = serialize(result, buffer_threshold=1000000)
         except Exception as e:
             logger.info('Caught an exception: {}'.format(e))
             result_package = {'type': 'result', 'task_id': tid, 'exception': serialize(RemoteExceptionWrapper(*sys.exc_info()))}
         else:
             result_package = {'type': 'result', 'task_id': tid, 'result': serialized_result}
-            # logger.debug("Result: {}".format(result))
+            del serialized_result
+        del req
 
         logger.info("Completed executor task {}".format(tid))
         try:
@@ -792,6 +812,7 @@ def worker(
                                         })
 
         result_queue.put(pkl_package)
+        del pkl_package, result_package
         tasks_in_progress.pop(worker_id)
         logger.info("All processing finished for executor task {}".format(tid))
 

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/high_throughput/zmq_pipes.py

@@ -136,30 +136,18 @@ class TasksOutgoing:
         self.port = self.zmq_socket.bind_to_random_port(tcp_url(ip_address),
                                                         min_port=port_range[0],
                                                         max_port=port_range[1])
-        self.poller = zmq.Poller()
-        self.poller.register(self.zmq_socket, zmq.POLLOUT)
 
     def put(self, message):
         """ This function needs to be fast at the same time aware of the possibility of
         ZMQ pipes overflowing.
 
-        The timeout increases slowly if contention is detected on ZMQ pipes.
         We could set copy=False and get slightly better latency but this results
         in ZMQ sockets reaching a broken state once there are ~10k tasks in flight.
         This issue can be magnified if each the serialized buffer itself is larger.
         """
-        timeout_ms = 1
-        while True:
-            socks = dict(self.poller.poll(timeout=timeout_ms))
-            if self.zmq_socket in socks and socks[self.zmq_socket] == zmq.POLLOUT:
-                # The copy option adds latency but reduces the risk of ZMQ overflow
-                logger.debug("Sending TasksOutgoing message")
-                self.zmq_socket.send_pyobj(message, copy=True)
-                logger.debug("Sent TasksOutgoing message")
-                return
-            else:
-                timeout_ms *= 2
-                logger.debug("Not sending due to non-ready zmq pipe, timeout: {} ms".format(timeout_ms))
+        logger.debug("Sending TasksOutgoing message")
+        self.zmq_socket.send_pyobj(message)
+        logger.debug("Sent TasksOutgoing message")
 
     def close(self):
         self.zmq_socket.close()
@@ -192,20 +180,15 @@ class ResultsIncoming:
         self.port = self.results_receiver.bind_to_random_port(tcp_url(ip_address),
                                                               min_port=port_range[0],
                                                               max_port=port_range[1])
-        self.poller = zmq.Poller()
-        self.poller.register(self.results_receiver, zmq.POLLIN)
 
     def get(self, timeout_ms=None):
         """Get a message from the queue, returning None if timeout expires
         without a message. timeout is measured in milliseconds.
         """
-        socks = dict(self.poller.poll(timeout=timeout_ms))
-        if self.results_receiver in socks and socks[self.results_receiver] == zmq.POLLIN:
-            m = self.results_receiver.recv_multipart()
-            logger.debug("Received ResultsIncoming message")
-            return m
-        else:
-            return None
+        if zmq.POLLIN == self.results_receiver.poll(timeout_ms, zmq.POLLIN):
+            logger.debug("Receiving ResultsIncoming multipart message")
+            return self.results_receiver.recv_multipart()
+        return None
 
     def close(self):
         self.results_receiver.close()

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/executors/taskvine/executor.py

@@ -107,13 +107,17 @@ class TaskVineExecutor(BlockProviderExecutor, putils.RepresentationMixin):
                  function_exec_mode: Union[Literal['regular'], Literal['serverless']] = 'regular',
                  manager_config: TaskVineManagerConfig = TaskVineManagerConfig(),
                  factory_config: TaskVineFactoryConfig = TaskVineFactoryConfig(),
-                 provider: Optional[ExecutionProvider] = LocalProvider(init_blocks=1),
+                 provider: Optional[ExecutionProvider] = None,
                  storage_access: Optional[List[Staging]] = None,
                  remote_monitoring_radio: Optional[RadioConfig] = None):
 
         # Set worker launch option for this executor
         if worker_launch_method == 'factory' or worker_launch_method == 'manual':
             provider = None
+        elif worker_launch_method == 'provider' and provider is None:
+            # provider method chosen, but no explicit provider supplied to __init__
+            # so default to LocalProvider
+            provider = LocalProvider(init_blocks=1)
 
         # Initialize the parent class with the execution provider and block error handling enabled.
         # If provider is None, then no worker is launched via the provider method.

{parsl-2025.9.22 → parsl-2025.10.6}/parsl/tests/configs/taskvine_ex.py

@@ -7,4 +7,4 @@ from parsl.executors.taskvine import TaskVineExecutor, TaskVineManagerConfig
 
 def fresh_config():
     return Config(executors=[TaskVineExecutor(manager_config=TaskVineManagerConfig(port=9000),
-                                              worker_launch_method='factory')])
+                                              worker_launch_method='provider')])