Bug 217966 – [GTK] inspector/console/queryHolders.html is failing

• Component: Web Inspector
• Status: REOPENED

[Diego Pino]

The test is very flaky since r267815. Inspecting the range r267800-r267815, r267807 introduced several changes in inspector code touching WebKitGTK code. I haven't bisected the range though (before r267815 the test sporadically timeout).

To reproduce the failure:

$ Tools/Scripts/run-webkit-tests --gtk --release --iterations 5 inspector/console/queryHolders.html

Diff: https://build.webkit.org/results/GTK-Linux-64-bit-Release-Tests/r268725%20(16529)/inspector/console/queryHolders-diff.txt

--- /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-expected.txt
+++ /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-actual.txt
@@ -75,7 +75,9 @@
 [RepeatedReferenceHolder, Root] (2)

 -- Running test case: CommandLineAPI.queryHolders.OnlyHeldByDebugger
-PASS: The result should have 0 items.
+FAIL: The result should have 0 items.
+    Expected: 0
+    Actual: 2

 -- Running test case: CommandLineAPI.queryHolders.NonObject
 PASS: Should produce an exception.

[Diego Pino]

I bisected the range [r267800-r267815] and I got r267814 as the first revision where I can reproduce the regression.

[Blaze Burg]

Carlos, I don't see anything obviously wrong which would cause this. Can you take a peek?

[Diego Pino]

(In reply to Brian Burg from comment #2)
> Carlos, I don't see anything obviously wrong which would cause this. Can you
> take a peek?

It seems this same issue was tracked in bug https://bugs.webkit.org/show_bug.cgi?id=217193

Recently the implementation of `item` introduced in r267814 was renamed to `at` in r268760, among other changes. Since then the test is no longer flaky, although it sporadically times out as it used to happen before.

https://results.webkit.org/?suite=layout-tests&test=inspector%2Fconsole%2FqueryHolders.html&platform=GTK&platform=WPE

[Diego Pino]

*** This bug has been marked as a duplicate of bug 217193 ***

[Diego Pino]

This test is no longer flaky in WebKitGTK as the results in the WebKitGTK post-commit bot report this test has been consistently failing in the last 4000 revisions.

The current diff is:

https://build.webkit.org/results/GTK-Linux-64-bit-Release-Tests/r278645%20%282223%29/inspector/console/queryHolders-diff.txt

--- /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-expected.txt
+++ /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-actual.txt
@@ -54,13 +54,13 @@
 [Root] (1)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseThen
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseCatch
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseFinally
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.EventListenerCallback
 [<div id="event-listener-target">, Root] (2)

[Diego Pino]

Now this test is failing with a different output:

--- /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-expected.txt
+++ /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/inspector/console/queryHolders-actual.txt
@@ -54,13 +54,13 @@
 [Root] (1)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseThen
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseCatch
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.PromiseFinally
-[Root, object] (2)
+[Object, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.EventListenerCallback
 [<div id="event-listener-target">, Root] (2)
@@ -75,7 +75,9 @@
 [RepeatedReferenceHolder, Root] (2)
 
 -- Running test case: CommandLineAPI.queryHolders.OnlyHeldByDebugger
-PASS: The result should have 0 items.
+FAIL: The result should have 0 items.
+    Expected: 0
+    Actual: 2
 
 -- Running test case: CommandLineAPI.queryHolders.NonObject
 PASS: Should produce an exception.


The reason for the [Root, object] and [Object, Root] difference is due to using a different label for Objects in GLib abd other WebKit ports.

https://webkit-search.igalia.com/webkit/source/LayoutTests/inspector/console/queryHolders.html#83

It can be easily solved by capitalizing the propertiy names before sorting.

As for the other part, the error is happening in this line:

https://webkit-search.igalia.com/webkit/source/LayoutTests/inspector/console/queryHolders.html#162

I have no idea why size is 2 for GLib ports but 0 in other ports.