Fix occasional test failures when running multiple jobs

The way runtest.py passes the list of fixture directories is racy because it
sets it in os.environ['FIXTURE_DIRS'] and then spawns the subprocess, counting
on Python to start the subprocess before that list is overwritten when spawning
the next directory. At least on Windows, the environment is not copied in
subprocess.run so runtest.py may overwrite the list of fixture directories
with the list for test #2 while the subprocess module is still kicking off
test #1. I was able to easily reproduce this by running the command:
`python runtest.py -j 2 test\MSVC\VSWHERE.py test\AS\ASPPFLAGS.py`
a few times in a row. However, with this fix, that command repeatedly succeeds.

To validate ths fix, I also ran that command with "--xml a.xml" and
"--xml a.xml --nopipefiles" to validate that those other executors worked
correctly.

Author: grossag

CHANGES.txt

@@ -33,6 +33,8 @@ RELEASE  VERSION/DATE TO BE FILLED IN LATER
   From Adam Gross:
     - Fix minor bug affecting SCons.Node.FS.File.get_csig()'s usage of the MD5 chunksize.
       User-facing behavior does not change with this fix (GH Issue #3726).
+    - Fix occasional test failures caused by not being able to find a file or directory fixture
+      when running multiple tests with multiple jobs.
 
   From Joachim Kuebart:
     - Suppress missing SConscript deprecation warning if `must_exist=False`

runtest.py

@@ -293,8 +293,8 @@ def escape(s):
 if not catch_output:
     # Without any output suppressed, we let the subprocess
     # write its stuff freely to stdout/stderr.
-    def spawn_it(command_args):
-        cp = subprocess.run(command_args, shell=False)
+    def spawn_it(command_args, env):
+        cp = subprocess.run(command_args, shell=False, env=env)
         return cp.stdout, cp.stderr, cp.returncode
 else:
     # Else, we catch the output of both pipes...
@@ -310,15 +310,16 @@ def spawn_it(command_args):
         #   http://http://thraxil.org/users/anders/posts/2008/03/13/Subprocess-Hanging-PIPE-is-your-enemy/
         # and pass temp file objects to Popen() instead of the ubiquitous
         # subprocess.PIPE.
-        def spawn_it(command_args):
+        def spawn_it(command_args, env):
             # Create temporary files
             tmp_stdout = tempfile.TemporaryFile(mode='w+t')
             tmp_stderr = tempfile.TemporaryFile(mode='w+t')
             # Start subprocess...
             cp = subprocess.run(command_args,
                                 stdout=tmp_stdout,
                                 stderr=tmp_stderr,
-                                shell=False)
+                                shell=False,
+                                env=env)
 
             try:
                 # Rewind to start of files
@@ -348,11 +349,12 @@ def spawn_it(command_args):
         #   (but the subprocess isn't writing anything there).
         #   Hence a deadlock.
         # Be dragons here! Better don't use this!
-        def spawn_it(command_args):
+        def spawn_it(command_args, env):
             cp = subprocess.run(command_args,
                                 stdout=subprocess.PIPE,
                                 stderr=subprocess.PIPE,
-                                shell=False)
+                                shell=False,
+                                env=env)
             return cp.stdout, cp.stderr, cp.returncode
 
 
@@ -380,8 +382,8 @@ def execute(self):
 
 class SystemExecutor(RuntestBase):
     """ Test class for tests executed with spawn_it() """
-    def execute(self):
-        self.stderr, self.stdout, s = spawn_it(self.command_args)
+    def execute(self, env):
+        self.stderr, self.stdout, s = spawn_it(self.command_args, env)
         self.status = s
         if s < 0 or s > 2:
             sys.stdout.write("Unexpected exit status %d\n" % s)
@@ -400,15 +402,16 @@ class PopenExecutor(RuntestBase):
     # and the 'allow_pipe_files' option, please check out the
     # definition of spawn_it() above.
     if allow_pipe_files:
-        def execute(self):
+        def execute(self, env):
             # Create temporary files
             tmp_stdout = tempfile.TemporaryFile(mode='w+t')
             tmp_stderr = tempfile.TemporaryFile(mode='w+t')
             # Start subprocess...
             cp = subprocess.run(self.command_str.split(),
                                 stdout=tmp_stdout,
                                 stderr=tmp_stderr,
-                                shell=False)
+                                shell=False,
+                                env=env)
             self.status = cp.returncode
 
             try:
@@ -423,11 +426,12 @@ def execute(self):
                 tmp_stdout.close()
                 tmp_stderr.close()
     else:
-        def execute(self):
+        def execute(self, env):
             cp = subprocess.run(self.command_str.split(),
                                 stdout=subprocess.PIPE,
                                 stderr=subprocess.PIPE,
-                                shell=False)
+                                shell=False,
+                                env=env)
             self.status = cp.returncode
             self.stdout = cp.stdout
             self.stderr = cp.stderr
@@ -756,11 +760,16 @@ def run_test(t, io_lock=None, run_async=True):
     if head:
         fixture_dirs.append(head)
     fixture_dirs.append(os.path.join(scriptpath, 'test', 'fixture'))
-    os.environ['FIXTURE_DIRS'] = os.pathsep.join(fixture_dirs)
+
+    # Set the list of fixture dirs directly in the environment. Just putting
+    # it in os.environ and spawning the process is racy. Make it reliable by
+    # overriding the environment passed to execute().
+    env = dict(os.environ)
+    env['FIXTURE_DIRS'] = os.pathsep.join(fixture_dirs)
 
     test_start_time = time_func()
     if execute_tests:
-        t.execute()
+        t.execute(env)
 
     t.test_time = time_func() - test_start_time
     log_result(t, io_lock=io_lock)