Bug 95414 – REGRESSION(r127039): It broke inspector/timeline/timeline-load.html fails on many platforms

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Roger Fong]

inspector/timeline/timeline-load.html:

[Roger Fong]

Test fails on Apple-Windows port :( any idea why?

[Roger Fong]

diff: 

--- /home/buildbot/slave/win-release-tests/build/layout-test-results/inspector/timeline/timeline-load-expected.txt	2012-08-29 13:57:46.075447000 -0700
+++ /home/buildbot/slave/win-release-tests/build/layout-test-results/inspector/timeline/timeline-load-actual.txt	2012-08-29 13:57:46.074446900 -0700
@@ -9,6 +9,3 @@
 Running: testIncompatible
 passed
 
-Running: testMalformed
-passed
-

Last test case checks try-catch execution branch in TimelineModel.js: WebInspector.TimelineModelLoader.transferChunk

Try to add "window.console.log('something');" around this code to see if it is the source of the problem.

[Mikhail Pozdnyakov]

Have same problem on EFL port

[Csaba Osztrogonác]

It isn't a win specific bug, it is a regression caused by r127039
and affedted platforms are: Win, EFL, Qt Debug

here is the Qt debug diff too:
--- /home/webkitbuildbot/slaves/debug64bit/buildslave/qt-linux-64-debug/build/layout-test-results/inspector/timeline/timeline-load-expected.txt 
+++ /home/webkitbuildbot/slaves/debug64bit/buildslave/qt-linux-64-debug/build/layout-test-results/inspector/timeline/timeline-load-actual.txt 
@@ -6,9 +6,4 @@
 Running: testNormal
 passed
 
-Running: testIncompatible
-passed
 
-Running: testMalformed
-passed
-

[Roger Fong]

the transferChunk code does not get called. In fact, from what I can tell, the last test isn't being run at all...

[Roger Fong]

Also would like to note that it seems like the 3rd test in the test suite just isn't being run in the Windows case.

If I switch the orders of the test cases, (malformed first, incompatible second), malformed test passes fine, incompatible test does not get run.

[Roger Fong]

Upon further investigation I'm seeing strange behavior in InspectorTest.completeTest.

In the first test, the code path makes it through the completeTest call and moves on to test the second test. The second test does not make it through completeTest and the third test does not get called.

I can put the tests in any order and only the first 2 will ever be run.

I would try to debug further...
However, depending on where and what I log, the results change, which is madness.
(Logging integers vs string literals changes things unexpectedly).

I'm adding this test to the Windows skip list for now to get the bots greener.

[Roger Fong]

It was fixed here this morning apparently...
http://trac.webkit.org/changeset/127135
by removing the completeTest call...

Sigh, closing bug...