Bug 74039 - chromium win test bots can get stuck
Summary: chromium win test bots can get stuck
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Dirk Pranke
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-12-07 16:47 PST by Dirk Pranke
Modified: 2012-05-08 14:32 PDT (History)
6 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Dirk Pranke 2011-12-07 16:47:19 PST
It appears there are cases where the Chromium win test bots can get stuck or wedged still (see, e.g.: http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Vista/builds/9519/steps/webkit_tests/logs/stdio). This can be very disruptive to gardeners and others watching the bots.

What seems to be happening is that DRT crashes, but instead of exiting, a dialog gets displayed on the bot. Because the chromium port-specific implementation on the Port class relies on DRT to enforce the timeout and exit, the worker thread in NRWT hangs indefinitely instead of timing out. We should fix this :)

At one point we also had code to detect when these threads were wedged, but that was removed (in r90207). We should consider restoring that code, perhaps just temporarily, if fixing this properly turns out to be hard.
Comment 1 Tony Chang 2011-12-07 16:51:43 PST
What is in the dialog?  Where did it come from?
Comment 2 Dirk Pranke 2011-12-07 17:01:33 PST
The dialog said something like "DumpRenderTree.exe has stopped working". I'm not sure what the source of the dialog was (it wasn't the regular windows debugger dialog).
Comment 3 Tony Chang 2011-12-07 18:26:02 PST
I thought we disabled the visual studio debugger dialog on our windows bots.
Comment 4 Tony Chang 2011-12-07 18:27:08 PST
Looks like bev fixed it on that bot: http://code.google.com/p/chromium/issues/detail?id=106799
Comment 5 Dirk Pranke 2011-12-07 18:53:28 PST
yeah, it's still broken on that bot. Regardless of the state of the chromium bots, we still need to fix these tests so that the chromium run_test() call implements the timeout properly.
Comment 6 Dirk Pranke 2012-05-08 14:32:23 PDT
this was fixed with the change to using "DRT mode" in NRWT/test-shell in r115903.