Auto merge of #9607 - mbrubeck:chaos, r=jgraham

Add a "mach test-wpt --chaos" mode for reproducing intermittent failures using rr

This adds a new `--chaos` flag to the `test-wpt` and `test-css` commands.  This will run a test repeatedly until it fails, recording it with the rr debugger in chaos mode. Requires a recent master build of rr.

```
./mach test-wpt --chaos test.html
```
is shorthand for:
```
./mach test-wpt --repeat-until-unexpected \
                --debugger rr \
                --debugger-args "record --chaos" \
                --include test.html
```
This PR currently includes changes to wptrunner which were also submitted upstream as w3c/wptrunner#170.  We should not merge this until that upstream PR is merged. CC @jgraham

<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.svg" height="40" alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/9607)
<!-- Reviewable:end -->

python/servo/testing_commands.py

@@ -41,6 +41,8 @@ def create_parser_wpt():
     parser = wptcommandline.create_parser()
     parser.add_argument('--release', default=False, action="store_true",
                         help="Run with a release build of servo")
+    parser.add_argument('--chaos', default=False, action="store_true",
+                        help="Run under chaos mode in rr until a failure is captured")
     return parser
 
 
@@ -332,19 +334,24 @@ class MachCommands(CommandBase):
     @Command('test-wpt',
              description='Run the web platform tests',
              category='testing',
-             parser=wptcommandline.create_parser)
-    @CommandArgument('--release', default=False, action="store_true",
-                     help="Run with a release build of servo")
+             parser=create_parser_wpt)
     def test_wpt(self, **kwargs):
         self.ensure_bootstrapped()
         hosts_file_path = path.join(self.context.topdir, 'tests', 'wpt', 'hosts')
-
         os.environ["hosts_file_path"] = hosts_file_path
-        os.environ["RUST_BACKTRACE"] = "1"
-
-        kwargs["debug"] = not kwargs["release"]
-
         run_file = path.abspath(path.join(self.context.topdir, "tests", "wpt", "run_wpt.py"))
+        return self.wptrunner(run_file, **kwargs)
+
+    # Helper for test_css and test_wpt:
+    def wptrunner(self, run_file, **kwargs):
+        os.environ["RUST_BACKTRACE"] = "1"
+        kwargs["debug"] = not kwargs["release"]
+        if kwargs.pop("chaos"):
+            kwargs["debugger"] = "rr"
+            kwargs["debugger_args"] = "record --chaos"
+            kwargs["repeat_until_unexpected"] = True
+            # TODO: Delete rr traces from green test runs?
+
         run_globals = {"__file__": run_file}
         execfile(run_file, run_globals)
         return run_globals["run_tests"](**kwargs)
@@ -398,11 +405,8 @@ class MachCommands(CommandBase):
              parser=create_parser_wpt)
     def test_css(self, **kwargs):
         self.ensure_bootstrapped()
-
         run_file = path.abspath(path.join("tests", "wpt", "run_css.py"))
-        run_globals = {"__file__": run_file}
-        execfile(run_file, run_globals)
-        return run_globals["run_tests"](**kwargs)
+        return self.wptrunner(run_file, **kwargs)
 
     @Command('update-css',
              description='Update the web platform tests',

tests/wpt/harness/wptrunner/wptcommandline.py

@@ -72,6 +72,8 @@ def create_parser(product_choices=None):
                         help="Multiplier relative to standard test timeout to use")
     parser.add_argument("--repeat", action="store", type=int, default=1,
                         help="Number of times to run the tests")
+    parser.add_argument("--repeat-until-unexpected", action="store_true", default=None,
+                        help="Run tests in a loop until one returns an unexpected result")
 
     parser.add_argument("--no-capture-stdio", action="store_true", default=False,
                         help="Don't capture stdio and write to logging")

tests/wpt/harness/wptrunner/wptrunner.py

@@ -160,10 +160,15 @@ def run_tests(config, test_paths, product, **kwargs):
             browser_kwargs = get_browser_kwargs(ssl_env=ssl_env, **kwargs)
 
             repeat = kwargs["repeat"]
-            for repeat_count in xrange(repeat):
-                if repeat > 1:
-                    logger.info("Repetition %i / %i" % (repeat_count + 1, repeat))
+            repeat_count = 0
+            repeat_until_unexpected = kwargs["repeat_until_unexpected"]
 
+            while repeat_count < repeat or repeat_until_unexpected:
+                repeat_count += 1
+                if repeat_until_unexpected:
+                    logger.info("Repetition %i" % (repeat_count))
+                elif repeat > 1:
+                    logger.info("Repetition %i / %i" % (repeat_count, repeat))
 
                 unexpected_count = 0
                 logger.suite_start(test_loader.test_ids, run_info)
@@ -208,6 +213,8 @@ def run_tests(config, test_paths, product, **kwargs):
 
                 unexpected_total += unexpected_count
                 logger.info("Got %i unexpected results" % unexpected_count)
+                if repeat_until_unexpected and unexpected_total > 0:
+                    break
                 logger.suite_end()
 
     return unexpected_total == 0