W3C navigation timing problem - agent 362 WPT Server r1936

[ash]

Hi,

I have problem with scripted scenario that is run with IE9.

combineSteps
navigate https://some.domain/someURI/aaa.aspx
setValue id=userid myID
clickAndWait id=sign_in_button
waitforComplete

I got following results:



The waterfall of course is broken because of such values.

For simpler tests results seem to be measured properly.[/quote]

[ash]

I investigated this problem deeper and it seems that it reveals for all my IE9 agents. In every test i get always the same values for domContentLoaded and loadEvent (same between tests).

The problem is present also for non-scripted tests. Opposite to what I wrote before.

---

Below is example of the affected waterfall chart, but for stripview waterfall charts are drawn proporly.

---

The wrong values (2147483647) in binary notations are 1111111111111111111111111111111

It may help to investigate what is wrong.

[pmeenan]

You may need to update the server-side page_data.inc code which filters out bogus timing data if it sees it (usually negative values that are converted to unsigned or something like that).

This change specifically: https://github.com/WPO-Foundation/webpag...08ec8bb7e0

[ash]

Has the source code been moved to github?

I usually upgrade my WPT server with svn update so probably this is why I have outdated version.

[pmeenan]

Yes. You should be able to rename the current directory, git clone to the old directory name, copy the config files over and move the results folder over.

Sorry about that - I couldn't find a cleaner way to do it. I did leave the old code on SVN to prevent accidentally wiping out someone's server in case they did an update so it is a little more confusing than it would normally have to be.

[ash]

Thanks Patrick for your support. It seems that my issue is solved now.