feat: Add group class for better control of multi dut tests

Author: lucasssvaz

docs/apis/pytest-embedded.rst

@@ -12,6 +12,11 @@
    :undoc-members:
    :show-inheritance:
 
+.. automodule:: pytest_embedded.group
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
 .. automodule:: pytest_embedded.dut_factory
    :members:
    :undoc-members:

docs/usages/expecting.rst

@@ -2,7 +2,7 @@
  Expecting Functions
 #####################
 
-In testing, most of the work involves expecting a certain string or pattern and then making assertions. This is supported by the functions :func:`~pytest_embedded.dut.Dut.expect`, :func:`~pytest_embedded.dut.Dut.expect_exact`, and :func:`~pytest_embedded.dut.Dut.expect_unity_test_output`.
+In testing, most of the work involves expecting a certain string or pattern and then making assertions. This is supported by the functions :func:`~pytest_embedded.dut.Dut.expect`, :func:`~pytest_embedded.dut.Dut.expect_exact`, :class:`~pytest_embedded.group.DutGroup` (multi-DUT synchronization), and :func:`~pytest_embedded.dut.Dut.expect_unity_test_output`.
 
 All of these functions accept the following keyword arguments:
 
@@ -186,6 +186,143 @@ As with the :func:`~pytest_embedded.dut.Dut.expect` function, the ``pattern`` ar
        for _ in range(2):
            dut.expect_exact(pattern_list)
 
+***************************
+ Multi-DUT Synchronization
+***************************
+
+When you use ``--count N`` (or equivalent), each board has its own serial stream and its own :class:`~pytest_embedded.dut.Dut` instance. Waiting for readiness on each device with separate ``expect`` calls works, but:
+
+-  Sequential calls use **per-call** timeouts, so two ``expect_exact(..., timeout=120)`` lines can behave like a much larger wall-clock budget than a single 120s deadline.
+-  The **slowest** device should not delay matching on others more than your chosen global timeout.
+
+:class:`~pytest_embedded.group.DutGroup`
+========================================
+
+``DutGroup`` is a transparent proxy: **every method** available on a single :class:`~pytest_embedded.dut.Dut` can be called on the group. The call runs on all members **in parallel** and returns a list of per-DUT results.
+
+.. code:: python
+
+   from pytest_embedded import DutGroup
+
+   def test_two_boards(dut):
+       group = DutGroup(dut[0], dut[1])
+       # or from a list:
+       group = DutGroup(*dut)
+
+It is also available as ``Dut.DutGroup`` for discoverability.
+
+expect / expect_exact
+---------------------
+
+``expect`` and ``expect_exact`` support both **broadcast** (one pattern for all DUTs) and **per-DUT** patterns (N patterns for N DUTs), all running in parallel:
+
+.. code:: python
+
+   # Broadcast -- same pattern to every DUT
+   group.expect_exact("[READY]", timeout=120)
+
+   # Per-DUT patterns -- one per DUT, in constructor order
+   group.expect_exact("[AP] ready", "[CLIENT] ready", timeout=120)
+
+   # Regex -- also supports broadcast and per-DUT forms
+   results = group.expect(r"IP=(\S+)", timeout=10)
+   ip0 = results[0].group(1).decode()
+   ip1 = results[1].group(1).decode()
+
+   # Same as :class:`~pytest_embedded.dut.Dut`: a single pattern may use the keyword form
+   group.expect_exact(pattern="[READY]", timeout=120)
+
+Other methods
+-------------
+
+Any other :class:`~pytest_embedded.dut.Dut` method called on the group is forwarded with the **same arguments** to every DUT in parallel:
+
+.. code:: python
+
+   group.write(ssid)
+
+For per-DUT arguments on non-expect methods, index into the group:
+
+.. code:: python
+
+   group[0].write(ap_config)
+   group[1].write(client_config)
+
+Container protocol
+------------------
+
+``DutGroup`` supports indexing, iteration, and length:
+
+.. code:: python
+
+   group[0]        # first DUT
+   group[-1]       # last DUT
+   len(group)      # number of DUTs
+   list(group)     # iterate over DUTs
+   group.duts      # underlying tuple (read-only)
+
+Non-callable attributes are returned as a list:
+
+.. code:: python
+
+   procs = group.pexpect_proc  # [proc_0, proc_1, ...]
+
+Names and clearer errors
+------------------------
+
+Pass optional **member** labels and an optional **group** label so logs and failures are easy to read:
+
+.. code:: python
+
+   group = DutGroup(*dut, names=("ap", "client"), group_name="wifi_ap")
+   # group.names -> ("ap", "client"); group.group_name -> "wifi_ap"
+
+If you omit ``names``, members default to ``dut-0``, ``dut-1``, … (aligned with per-DUT log file names when using ``--count``).
+
+When any parallel call fails on one DUT, pytest-embedded raises :exc:`pytest_embedded.group.DutGroupMemberError`. Its message and attributes identify the member (``member_name``, ``member_index``) and group (``group_name``), and the original error (for example :exc:`pexpect.TIMEOUT`) is chained as :attr:`__cause__`. A structured line is also written to the Python logger at ERROR (including the underlying exception context).
+
+Full example
+------------
+
+.. code:: python
+
+   from pytest_embedded import DutGroup
+
+   def test_wifi_ap(dut):
+       group = DutGroup(*dut)
+
+       # Phase 1: wait for both devices to be ready
+       group.expect_exact("[READY]", timeout=120)
+
+       # Phase 2: exchange SSID
+       group.expect_exact("Send SSID:", timeout=10)
+       group.write(ap_ssid)
+
+       # Phase 3: exchange password
+       group.expect_exact("Send Password:", timeout=10)
+       group.write(ap_password)
+
+       # Phase 4: verify connection
+       results = group.expect(r"IP=(\S+)", timeout=30)
+       for r in results:
+           assert r.group(1) != b""
+
+Phase synchronization
+=====================
+
+``DutGroup`` methods can be called **multiple times** in one test to synchronize phases. Each call blocks until every DUT has matched before continuing. After a successful match, those substrings are consumed from each DUT's buffer; emit new output for the next phase.
+
+.. code:: python
+
+   group = DutGroup(*dut)
+   group.expect_exact("Init OK", timeout=30)
+   group.expect_exact("Server started", timeout=10)
+   group.expect_exact("Connected", timeout=60)
+
+.. note::
+
+   If one DUT fails, pending work is cancelled where possible; expects that have already started may still run until they match or time out, because pexpect cannot always be interrupted from another thread. The failure is reported as :exc:`~pytest_embedded.group.DutGroupMemberError` with the underlying error as :attr:`~BaseException.__cause__`.
+
 ***********************************************************
  :func:`~pytest_embedded.dut.Dut.expect_unity_test_output`
 ***********************************************************

pytest-embedded/pytest_embedded/__init__.py

@@ -3,7 +3,8 @@
 from .app import App
 from .dut import Dut
 from .dut_factory import DutFactory
+from .group import DutGroup, DutGroupMemberError
 
-__all__ = ['App', 'Dut', 'DutFactory']
+__all__ = ['App', 'Dut', 'DutFactory', 'DutGroup', 'DutGroupMemberError']
 
 __version__ = '2.7.0'

pytest-embedded/pytest_embedded/dut.py

@@ -9,6 +9,7 @@
 import pexpect
 
 from .app import App
+from .group import DutGroup
 from .log import MessageQueue, PexpectProcess
 from .unity import UNITY_SUMMARY_LINE_REGEX, TestSuite
 from .utils import Meta, _InjectMixinCls, remove_asci_color_code, to_bytes, to_list
@@ -49,6 +50,8 @@ def __init__(
         # junit related
         self.testsuite = TestSuite(self.test_case_name)
 
+
+
     @property
     def logdir(self):
         return self._meta.logdir
@@ -232,3 +235,7 @@ def run_all_single_board_cases(
             requires enable service ``idf``
         """
         pass
+
+
+#: Alias for :class:`~pytest_embedded.group.DutGroup` for discoverability.
+Dut.DutGroup = DutGroup