Bug 139143 - [iOS] run-webkit-tests records most DumpRenderTree.app crashes as time-outs
Summary: [iOS] run-webkit-tests records most DumpRenderTree.app crashes as time-outs
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: iPhone / iPad iOS 8.1
: P2 Normal
Assignee: Daniel Bates
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-12-01 12:45 PST by Daniel Bates
Modified: 2014-12-23 22:37 PST (History)
6 users (show)

See Also:


Attachments
Patch (10.05 KB, patch)
2014-12-01 12:47 PST, Daniel Bates
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Daniel Bates 2014-12-01 12:45:02 PST
iOS DumpRenderTree records most crashes as time outs because ReportCrash(8) delays the exit notification to LayoutTestRelay past the run-webkit-tests time-out time limit. We should use a similar approach as used in the patch for bug #37859 and write "#CRASHED" to the standard error stream when DumpRenderTree.app receives a signal for a fatal error (e.g. SIGSEGV). Then run-webkit-tests script will mark the test as crashing when it reads "#CRASHED" from the standard error stream.
Comment 1 Daniel Bates 2014-12-01 12:47:44 PST
Created attachment 242332 [details]
Patch
Comment 2 David Kilzer (:ddkilzer) 2014-12-02 10:02:11 PST
Comment on attachment 242332 [details]
Patch

r=me
Comment 3 Daniel Bates 2014-12-02 10:03:58 PST
Comment on attachment 242332 [details]
Patch

Clearing flags on attachment: 242332

Committed r176669: <http://trac.webkit.org/changeset/176669>
Comment 4 Daniel Bates 2014-12-02 10:04:02 PST
All reviewed patches have been landed.  Closing bug.
Comment 5 Alexey Proskuryakov 2014-12-23 22:37:21 PST
We probably need this fix on Mac too, and also in WebKitTestRunner on both platforms.

On Mac, ReportCrash usually finishes before run-webkit-tests decides that it was a timeout, but not on Yosemite. WebKit2 is less affected, because we do print #CRASHED for WebProcess crashes - but UI process has the same problem.