binman: Run tests concurrently

At present the tests run one after the other using a single CPU. This is
not very efficient. Bring in the concurrencytest module and run the tests
concurrently, using one process for each CPU by default. A -P option
allows this to be overridden, which is necessary for code-coverage to
function correctly.

This requires fixing a few tests which are currently not fully
independent.

At some point we might consider doing this across all pytests in U-Boot.
There is a pytest version that supports specifying the number of processes
to use, but it did not work for me.

Signed-off-by: Simon Glass <sjg@chromium.org>

1 files changed, 144 insertions, 0 deletions

diff --git a/tools/concurrencytest/concurrencytest.py b/tools/concurrencytest/concurrencytest.py
new file mode 100644
index 0000000000..418d7eed21
--- /dev/null
+++ b/tools/concurrencytest/concurrencytest.py
@@ -0,0 +1,144 @@
+#!/usr/bin/env python
+# SPDX-License-Identifier: GPL-2.0+
+#
+# Modified by: Corey Goldberg, 2013
+#
+# Original code from:
+#   Bazaar (bzrlib.tests.__init__.py, v2.6, copied Jun 01 2013)
+#   Copyright (C) 2005-2011 Canonical Ltd
+
+"""Python testtools extension for running unittest suites concurrently.
+
+The `testtools` project provides a ConcurrentTestSuite class, but does
+not provide a `make_tests` implementation needed to use it.
+
+This allows you to parallelize a test run across a configurable number
+of worker processes. While this can speed up CPU-bound test runs, it is
+mainly useful for IO-bound tests that spend most of their time waiting for
+data to arrive from someplace else and can benefit from cocncurrency.
+
+Unix only.
+"""
+
+import os
+import sys
+import traceback
+import unittest
+from itertools import cycle
+from multiprocessing import cpu_count
+
+from subunit import ProtocolTestCase, TestProtocolClient
+from subunit.test_results import AutoTimingTestResultDecorator
+
+from testtools import ConcurrentTestSuite, iterate_tests
+
+
+_all__ = [
+    'ConcurrentTestSuite',
+    'fork_for_tests',
+    'partition_tests',
+]
+
+
+CPU_COUNT = cpu_count()
+
+
+def fork_for_tests(concurrency_num=CPU_COUNT):
+    """Implementation of `make_tests` used to construct `ConcurrentTestSuite`.
+
+    :param concurrency_num: number of processes to use.
+    """
+    def do_fork(suite):
+        """Take suite and start up multiple runners by forking (Unix only).
+
+        :param suite: TestSuite object.
+
+        :return: An iterable of TestCase-like objects which can each have
+        run(result) called on them to feed tests to result.
+        """
+        result = []
+        test_blocks = partition_tests(suite, concurrency_num)
+        # Clear the tests from the original suite so it doesn't keep them alive
+        suite._tests[:] = []
+        for process_tests in test_blocks:
+            process_suite = unittest.TestSuite(process_tests)
+            # Also clear each split list so new suite has only reference
+            process_tests[:] = []
+            c2pread, c2pwrite = os.pipe()
+            pid = os.fork()
+            if pid == 0:
+                try:
+                    stream = os.fdopen(c2pwrite, 'wb', 1)
+                    os.close(c2pread)
+                    # Leave stderr and stdout open so we can see test noise
+                    # Close stdin so that the child goes away if it decides to
+                    # read from stdin (otherwise its a roulette to see what
+                    # child actually gets keystrokes for pdb etc).
+                    sys.stdin.close()
+                    subunit_result = AutoTimingTestResultDecorator(
+                        TestProtocolClient(stream)
+                    )
+                    process_suite.run(subunit_result)
+                except:
+                    # Try and report traceback on stream, but exit with error
+                    # even if stream couldn't be created or something else
+                    # goes wrong.  The traceback is formatted to a string and
+                    # written in one go to avoid interleaving lines from
+                    # multiple failing children.
+                    try:
+                        stream.write(traceback.format_exc())
+                    finally:
+                        os._exit(1)
+                os._exit(0)
+            else:
+                os.close(c2pwrite)
+                stream = os.fdopen(c2pread, 'rb', 1)
+                test = ProtocolTestCase(stream)
+                result.append(test)
+        return result
+    return do_fork
+
+
+def partition_tests(suite, count):
+    """Partition suite into count lists of tests."""
+    # This just assigns tests in a round-robin fashion.  On one hand this
+    # splits up blocks of related tests that might run faster if they shared
+    # resources, but on the other it avoids assigning blocks of slow tests to
+    # just one partition.  So the slowest partition shouldn't be much slower
+    # than the fastest.
+    partitions = [list() for _ in range(count)]
+    tests = iterate_tests(suite)
+    for partition, test in zip(cycle(partitions), tests):
+        partition.append(test)
+    return partitions
+
+
+if __name__ == '__main__':
+    import time
+
+    class SampleTestCase(unittest.TestCase):
+        """Dummy tests that sleep for demo."""
+
+        def test_me_1(self):
+            time.sleep(0.5)
+
+        def test_me_2(self):
+            time.sleep(0.5)
+
+        def test_me_3(self):
+            time.sleep(0.5)
+
+        def test_me_4(self):
+            time.sleep(0.5)
+
+    # Load tests from SampleTestCase defined above
+    suite = unittest.TestLoader().loadTestsFromTestCase(SampleTestCase)
+    runner = unittest.TextTestRunner()
+
+    # Run tests sequentially
+    runner.run(suite)
+
+    # Run same tests across 4 processes
+    suite = unittest.TestLoader().loadTestsFromTestCase(SampleTestCase)
+    concurrent_suite = ConcurrentTestSuite(suite, fork_for_tests(4))
+    runner.run(concurrent_suite)