Add first bindings unit test and enable testing.

third_party/WebKit/Tools/Scripts/run-bindings-tests

Index: third_party/WebKit/Tools/Scripts/run-bindings-tests
diff --git a/third_party/WebKit/Tools/Scripts/run-bindings-tests b/third_party/WebKit/Tools/Scripts/run-bindings-tests
index 6f4ab449292688ebd6f65b5811cec58206fa40e3..ea6a6adc5c1b17d87092f57de647cab4a9b1b003 100755
--- a/third_party/WebKit/Tools/Scripts/run-bindings-tests
+++ b/third_party/WebKit/Tools/Scripts/run-bindings-tests
@@ -26,7 +26,10 @@
 import sys
 
 from webkitpy.bindings.bindings_tests import run_bindings_tests
+from webkitpy.common import webkit_finder
+webkit_finder.add_typ_dir_to_sys_path()
 
+import typ
 
 def main(argv):
     """Runs Blink bindings IDL compiler on test IDL files and compares the
@@ -36,15 +39,38 @@ def main(argv):
     (this is automatically done as a presubmit script),
     and submit changes to the test results in the same patch.
     This makes it easier to track and review changes in generated code.
-
-    Options:
-       --reset-results: Overwrites reference files with the generated results.
-       --verbose: Show output on success and logging messages (not just failure)
     """
-    reset_results = '--reset-results' in argv
-    verbose = '--verbose' in argv
+    argument_parser = typ.ArgumentParser()

[peria, 2016/12/12 04:43:10]

optional:
I prefer to put a routine to parse options into a separate function.

[dglazkov, 2016/12/12 05:08:22]

SG, I'll think of a way to break it up better.

+    argument_parser.add_argument('--reset-results',
+        default=False,
+        action='store_true',
+        help='Overwrites reference files with the generated results.')
+    argument_parser.add_argument('--skip-unit-tests',
+        default=False,
+        action='store_true',
+        help='Skip running unit tests (only run reference tests).')
+    argument_parser.add_argument('--skip-reference-tests',
+        default=False,
+        action='store_true',
+        help='Skip running reference tests (only run unit tests).')
+
+    # First, run bindings unit tests.

[peria, 2016/12/12 04:43:10]

Why don't you use typ.main() as you did in
https://codereview.chromium.org/2554503004/?

I think it good to check if one or more tests exist,
but it is a bit too aggressive to run tests directly.

[dglazkov, 2016/12/12 05:08:22]

Ah, turns out, I can't. I need to effectively break apart typ.main, since I need
to inform typ that the embedding app has additional arguments (--reset-results),
and feed it its own argument parser. First, I had to go teach typ that this is
possible
(https://github.com/dpranke/typ/commit/c8845b2354afb6350244beb2d4430da3a2b203e3)
and Dirk graciously rolled it in to chromium over the weekend. Nesting yak
shaves! :)

I didn't realize this was necessary in
https://codereview.chromium.org/2554503004/, not until I actually tried using
--reset-results :D

Otherwise, supplying "--reset-results" as a command argument results in typ
throwing an exception, complaining of unknown argument.

+    runner = typ.Runner()
+    runner.parse_args(argument_parser, argv[1:])
+    args = runner.args
+    if argument_parser.exit_status is not None:
+        return argument_parser.exit_status
+    runner.args.top_level_dir = webkit_finder.get_bindings_scripts_dir()
+    if not args.skip_unit_tests:
+        return_code, _, _ = runner.run()
+        if return_code != 0:
+            return return_code
+
+    if args.skip_reference_tests:
+        return 0
 
-    return run_bindings_tests(reset_results, verbose)
+    # Now run the bindings end-to-end tests.
+    return run_bindings_tests(args.reset_results, args.verbose)
 
 
 if __name__ == '__main__':