Merge branch 'tests20200301'

* tests20200301:
  tests/salt: don't depend on initial tags set
  tests: make audio tests less racy

qubes/tests/integ/salt.py

@@ -206,8 +206,10 @@ class TC_00_Dom0(SaltTestMixin, qubes.tests.SystemTestCase):
         vm = self.app.add_new_vm('AppVM', label='red',
             name=vmname)
         self.loop.run_until_complete(vm.create_on_disk())
+        new_tags = list(sorted(list(vm.tags) + ['tag1', 'tag2']))
         vm.tags.add('tag1')
         vm.tags.add('tag3')
+        old_tags = list(sorted(vm.tags))
 
         with open(os.path.join(self.salt_testdir, 'test_state.sls'), 'w') as f:
             f.write(vmname + ':\n')
@@ -224,13 +226,14 @@ class TC_00_Dom0(SaltTestMixin, qubes.tests.SystemTestCase):
         state_out = list(cmd_output['local'].values())[0]
         del state_out['start_time']
         del state_out['duration']
+        self.maxDiff = None
         self.assertEqual(state_out, {
             'comment': '====== [\'tags\'] ======\n',
             'name': vmname,
             'result': True,
             'changes': {
                 'qvm.tags': {'qvm.tags': {
-                        'new': ['tag1', 'tag2'], 'old': ['tag1', 'tag3'],
+                        'new': new_tags, 'old': old_tags,
                     }
                 },
             },
@@ -241,7 +244,7 @@ class TC_00_Dom0(SaltTestMixin, qubes.tests.SystemTestCase):
 
         self.assertIn(vmname, self.app.domains)
         vm = self.app.domains[vmname]
-        self.assertEqual(set(vm.tags), {'tag1', 'tag2'})
+        self.assertEqual(set(vm.tags), set(new_tags))
 
     def test_020_qubes_pillar(self):
         vmname = self.make_vm_name('appvm')

qubes/tests/integ/vm_qrexec_gui.py

@@ -805,6 +805,20 @@ class TC_00_AppVMMixin(object):
         finally:
             self.app.clockvm = None
 
+    def wait_for_pulseaudio_startup(self, vm):
+        self.loop.run_until_complete(
+            self.wait_for_session(self.testvm1))
+        try:
+            self.loop.run_until_complete(vm.run_for_stdio(
+                "timeout 30s sh -c 'while ! pactl info; do sleep 1; done'"
+            ))
+        except subprocess.CalledProcessError as e:
+            self.fail('Timeout waiting for pulseaudio start in {}: {}{}'.format(
+                vm.name, e.stdout, e.stderr))
+        # and some more...
+        self.loop.run_until_complete(asyncio.sleep(1))
+
+
     @unittest.skipUnless(spawn.find_executable('parecord'),
                          "pulseaudio-utils not installed in dom0")
     def test_220_audio_playback(self):
@@ -817,10 +831,7 @@ class TC_00_AppVMMixin(object):
         except subprocess.CalledProcessError:
             self.skipTest('pulseaudio-utils not installed in VM')
 
-        self.loop.run_until_complete(
+        self.wait_for_pulseaudio_startup(self.testvm1)
-            self.wait_for_session(self.testvm1))
-        # and some more...
-        self.loop.run_until_complete(asyncio.sleep(1))
         # generate some "audio" data
         audio_in = b'\x20' * 44100
         self.loop.run_until_complete(
@@ -845,8 +856,13 @@ class TC_00_AppVMMixin(object):
             p.wait()
             # allow few bytes missing, don't use assertIn, to avoid printing
             # the whole data in error message
-            if audio_in[:-8] not in recorded_audio.file.read():
+            recorded_audio = recorded_audio.file.read()
-                self.fail('played sound not found in dom0')
+            if audio_in[:-8] not in recorded_audio:
+                found_bytes = recorded_audio.count(audio_in[0])
+                all_bytes = len(audio_in)
+                self.fail('played sound not found in dom0, '
+                          'missing {} bytes out of {}'.format(
+                              all_bytes-found_bytes, all_bytes))
 
     def _configure_audio_recording(self, vm):
         '''Connect VM's output-source to sink monitor instead of mic'''
@@ -883,10 +899,7 @@ class TC_00_AppVMMixin(object):
         except subprocess.CalledProcessError:
             self.skipTest('pulseaudio-utils not installed in VM')
 
-        self.loop.run_until_complete(
+        self.wait_for_pulseaudio_startup(self.testvm1)
-            self.wait_for_session(self.testvm1))
-        # and some more...
-        self.loop.run_until_complete(asyncio.sleep(1))
         # connect VM's recording source output monitor (instead of mic)
         self._configure_audio_recording(self.testvm1)
 
@@ -924,10 +937,7 @@ class TC_00_AppVMMixin(object):
         except subprocess.CalledProcessError:
             self.skipTest('pulseaudio-utils not installed in VM')
 
-        self.loop.run_until_complete(
+        self.wait_for_pulseaudio_startup(self.testvm1)
-            self.wait_for_session(self.testvm1))
-        # and some more...
-        self.loop.run_until_complete(asyncio.sleep(1))
         da = qubes.devices.DeviceAssignment(self.app.domains[0], 'mic')
         self.loop.run_until_complete(
             self.testvm1.devices['mic'].attach(da))
@@ -954,7 +964,11 @@ class TC_00_AppVMMixin(object):
             self.testvm1.run_for_stdio('cat audio_rec.raw'))
         # allow few bytes to be missing
         if audio_in[:-8] not in recorded_audio:
-            self.fail('VM not recorded expected data')
+            found_bytes = recorded_audio.count(audio_in[0])
+            all_bytes = len(audio_in)
+            self.fail('VM not recorded expected data, '
+                      'missing {} bytes out of {}'.format(
+                          all_bytes-found_bytes, all_bytes))
 
     def test_250_resize_private_img(self):
         """