Bug 1702689 - Fix browsertime_benchmark.js to collect all pagecycle results...

Bug 1702689 - Fix browsertime_benchmark.js to collect all pagecycle results r=perftest-reviewers,sparky

Fixed browsertime benchmark script to correctly run the tests, fix python code to collect all the pagecycles data and disable unity-webgl test running on windows7.

Differential Revision: https://phabricator.services.mozilla.com/D110683

taskcluster/ci/test/browsertime-desktop.yml

@@ -198,7 +198,8 @@ browsertime-benchmark:
                         by-app:
                             firefox:
                                 by-test-platform:
-                                    (linux|windows|macos)(?!.*shippable).*: []
+                                    windows7-32.*: []
+                                    (linux|windows10|macos)(?!.*shippable).*: []
                                     default: ["autoland"]
                             default: []
                     default:
@@ -219,6 +220,8 @@ browsertime-benchmark:
                     .*-ref-hw-2017/.*: 2100
                     default: 1500
             jetstream2: 8000
+            motionmark-htmlsuite: 1500
+            unity-webgl: 1500
             default: 900
     fetches:
         fetch:
@@ -256,7 +259,11 @@ browsertime-benchmark-wasm:
                     macos.*shippable.*: ["trunk", "mozilla-beta"]
                     windows10.*shippable.*: ["trunk", "mozilla-beta"]
                     default: []
-    max-run-time: 900
+    max-run-time:
+        by-subtest:
+            wasm-godot: 1500
+            wasm-godot-baseline: 1500
+            default: 900
     fetches:
         fetch:
             - wasm-misc

testing/raptor/browsertime/browsertime_benchmark.js

@@ -50,8 +50,14 @@ module.exports = async function(context, commands) {
       (await commands.js.run(`return performance.now();`)) - starttime >=
         page_timeout
     ) {
+      ret = false;
       context.log.error("Benchmark timed out. Aborting...");
     } else if (data) {
+      // Reset benchmark results
+      await commands.js.run(
+        "return window.sessionStorage.removeItem('benchmark_results');"
+      );
+
       context.log.info("Value of benchmark data: ", data);
       data = JSON.parse(data);
 

testing/raptor/raptor/results.py

@@ -560,15 +560,8 @@ class BrowsertimeResultsHandler(PerftestResultsHandler):
             custom_types = raw_result["extras"][0]
             if custom_types:
                 for custom_type in custom_types:
-                    # TODO (bug 1702689): Fix non-youtube-playback multi-cycle data so we
-                    # don't have to use this hack
-                    if any(["youtube" in k for k in custom_types[custom_type]]):
-                        for k, v in custom_types[custom_type].items():
-                            bt_result["measurements"].setdefault(k, []).append([v])
-                    else:
-                        bt_result["measurements"].update(
-                            {k: [v] for k, v in custom_types[custom_type].items()}
-                        )
+                    for k, v in custom_types[custom_type].items():
+                        bt_result["measurements"].setdefault(k, []).append(v)
             else:
                 # extracting values from browserScripts and statistics
                 for bt, raptor in conversion: