Bug 69415 – REGRESSION (r96346-r96347): http/tests/websocket/tests/hybi/close.html crashes on mac

• Status: NEW
• Component: Tools / Tests

[Yuta Kitamura]

http/tests/websocket/tests/hybi/close.html started crashing on SnowLeopard Intel Debug/Release (Tests) between r96346 and r96347 (inclusive).

http://trac.webkit.org/log/trunk?rev=96347&stop_rev=96346&limit=3

http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r96345%20(33519)/results.html passed
http://build.webkit.org/results/SnowLeopard%20Intel%20Release%20(Tests)/r96347%20(33520)/results.html failed

The crash log doesn't seem helpful. I don't have any idea about the cause at the moment.

Inviting authors and reviewers of these two revisions. Do you have any insights?

[Mark Hahnenberg]

Hmm this is strange.  The crash output references getCalDataVirtual, not visitChildrenVirtual which is what r96346 modified.  I was also working on getCallDataVirtual concurrently.  Are you sure the crashes started with these two revisions?  Is it only on Snow Leopard?

[Alexey Proskuryakov]

<http://build.webkit.org/old-results/SnowLeopard%20Intel%20Release%20(Tests)/r96347%20(33520)/http/tests/websocket/tests/hybi/close-crash-log.txt> is not a runtime crash. DumpRenderTree fails to start because it cannot find a symbol in JavaScriptCore. But that makes no sense, as all tests would crash in the same way then.

[Yuta Kitamura]

Results from the Debug tester seem consistent with the regression range of the Release tester:

http/tests/websocket/tests/hybi/close.html started failing on SnowLeopard Intel Debug (Tests) between r96342 and r96347 (inclusive).

http://trac.webkit.org/log/trunk?rev=96347&stop_rev=96342&limit=7

http://build.webkit.org/results/SnowLeopard%20Intel%20Debug%20(Tests)/r96341%20(2373)/results.html passed
http://build.webkit.org/results/SnowLeopard%20Intel%20Debug%20(Tests)/r96347%20(2374)/results.html failed

[Yuta Kitamura]

(In reply to comment #1)
> Is it only on Snow Leopard?

Yes, as far as I know.

Crash logs are quite confusing and I still don't understand the cause. I'm going to try to reproduce locally and do some digging...

[Alexey Proskuryakov]

As mentioned before, the crash log simply states that WebCore framework doesn't exist at the time of launching DumpRenderTree. This sounds like a tools issue.

[Dirk Pranke]

(In reply to comment #5)
> As mentioned before, the crash log simply states that WebCore framework doesn't exist at the time of launching DumpRenderTree. This sounds like a tools issue.

You mean JSC, not WebCore, right? Also, as your own comment #2 states, it seems like if that was the case, all of the tests would've failed, not just the one.

Is it possible there's just something wrong with that machine that will clear up if we reboot it? Are we seeing this on multiple machines?

[Alexey Proskuryakov]

I mean WebCore:

  Library not loaded: /Volumes/Big/slave/snowleopard-intel-debug/build/WebKitBuild/Debug/WebCore.framework/Versions/A/WebCore
  Referenced from: /Volumes/Big/slave/snowleopard-intel-debug-tests/build/WebKitBuild/Debug/./libWebCoreTestSupport.dylib
  Reason: image not found

It's unclear to me how this can be happening. Perhaps a new build is started while an old one still runs tests?

[Alexey Proskuryakov]

Or maybe a wrong CrashLog is uploaded?

[Dirk Pranke]

(In reply to comment #7)
> I mean WebCore:
> 
>   Library not loaded: /Volumes/Big/slave/snowleopard-intel-debug/build/WebKitBuild/Debug/WebCore.framework/Versions/A/WebCore
>   Referenced from: /Volumes/Big/slave/snowleopard-intel-debug-tests/build/WebKitBuild/Debug/./libWebCoreTestSupport.dylib
>   Reason: image not found
> 
> It's unclear to me how this can be happening. Perhaps a new build is started while an old one still runs tests?

Ah, I see, I was looking at the crash log in comment #2.

[Brent Fulgham]

This appears to still be crashing on the iOS simulator.

https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#tests=http%2Ftests%2Fwebsocket%2Ftests%2Fhybi%2Fclose.html