On 06/29/2012 05:40 PM, Brady Eidson wrote:

On Jun 28, 2012, at 11:57 PM, Horky Chen <[email protected] <mailto:[email protected]>> wrote:

Hi,

On Mac OS, if one time-out larger than 30s would be used, --time-out-ms cannot work well.

According to the run-webkit-tests script, custom Time-Out can be assigned for each test case. But, unfortunately, below line in LayoutTestControllerMac.mm blocked the setting if it is larger than 30s (waitUntilDone & notifyDone are used):
    static const CFTimeInterval waitToDumpWatchdogInterval = 30.0;

I think this is just the default, WebKitTestRunner has a --timeout that should control this if given. If that's not the case than it seems like a bug for me. On the other hand, I don't think run-webkit-tests supports setting custom timeout for a particular test.

That is hard code, and no parameter can be accepted to adjust it. There are two time-out settings for one test case, is it possible to use common time-out setting?

The two timeouts have slightly different purpose. The interval of the watchdog timer in the driver (in WebKitTestRunner) can detect if the test is slow (the web process don't finish in the interval). In this case we can rely on the driver and there is no need to restart it (because it is still responsive, just the test takes too much time). The timeout of the test harness (run-webkit-tests) take into action when the driver become unresponsive, and in this case it is necessary to restart it.


Would you please help to double check about it?

Historically there has been absolutely no excuse to *expect* a single test take longer than 30 seconds.

Such a test needs to be redesigned or broken up in to smaller tests.

Thanks,
~Brady


Best Regards!
Horky
_______________________________________________
webkit-dev mailing list
[email protected] <mailto:[email protected]>
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev



_______________________________________________
webkit-dev mailing list
[email protected]
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

_______________________________________________
webkit-dev mailing list
[email protected]
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

Reply via email to