redirects confuse FF-measurements

Support
1

Hi,
I am using wpt 2.5 (incl. agent-patch for Chrome/FF-agents) wth IE 8, Chrome 16 and FF 7-agents.

I have the following problems with my FF 7-measurements:

Many of my pages do redirects in different ways. Some pages are loaded from the onload-event of a wait-page, for example. The FF-results of these pages are getting into a mess: e.g. a search result-page shows a document complete-time of about 4 and a fully loaded-time of about 5.5 seconds, tested with IE 8. Same page tested with FF 7 shows a document complete-time of 0.1 and a fully loaded-time of about 3.5 seconds. The first html-request for the wait-page is measured for document loaded-time, apparently (see attachement).

I will try to integrate some end condition-commands in wpt-scripts like requiredRequest (with requests for static resources on the main page) but don’ know whether that will change the measurements.

Do you have another idea to obtain comparable IE- and FF-measurements with pages incl. redirects?

2

Do you have a publicly-reachable page that you can share for testing? It could be a simple bug in the new agent code (firefox/chrome) where the IE agent uses the time of the last onLoad, it is possible that the wptdriver agent is using the first onload and not allowing it to be updated.

I’ll take a look at the code but a test case would help.

3

The script of my example:

logData 0 navigate http://www.otto.de execAndWait document.getElementById('TconHrefModul_02').click(); logData 1 exec document.getElementById('searchValue').value='shirt'; execAndWait document.getElementById('top_search_submit').click();

4

Some news about that subject?

5

Is it possible that changing the FF-version (e.g. from 7 to 3.6 or 8) will solve this problem?

6

Sorry, last week was crazy - I’ll take a look today. I don’t think changing the Firefox version will have any effect.

7

Found the issue and have a fix. You can grab an updated overlay.js for the firefox extension from here: http://webpagetest.googlecode.com/svn/trunk/agent/browser/firefox/extension/chrome/content/overlay.js (goes in templates\Firefox\extensions\wptdriver@webpagetest.org\chrome\content) or if you want to wait until later today I’ll also have a full binary update for the agent that includes the fix.

Thanks,

-Pat

8

ok, the updated agent package is available from here: http://www.webpagetest.org/work/update/wptupdate.zip

Just extract the contents to your agent directory or drop it in the work/update directory on your server (if you drop it in work/update you will need to extract wptupdate.ini from the zip to the same directory).

Here is a sample result: http://www.webpagetest.org/result/120206_4V_34NKV/

Thanks,

-Pat

9

Thanks a lot!!!
I will test the patch as soon as possible.

10

That works, great!!!

But:
Chrome-Jobs broke since I brought in the new patch. On agents with chrome-version 16 chrome-jobs work, on agents with chrome 17 the jobs aren’t working (maybe that error appeared independent of the patch).
Chrome opens with “about:blank” in adress line and nothing happens till timeout. I have installed the patch through automatic update-mechanism (deleted everything from work/update-folder and copied zip and ini to it). Some weeks ago I installed your 2.5 agent-patch to get Chrome16/IE8 to work on same agents.

btw:
If work/update/wptupdate.zip and work/update/wptupdate.ini are for agent-updates, what are work/update/update.zip and work/update/update.ini for?

11

Try grabbing the update again, I had to roll out a fix for Chrome 17 when it launched.

wptupdate.* is for the Chrome/Firefox support (wptdriver) and update.* is for the IE code (urlblast, etc). Eventually IE will migrate to use wptdriver as well and it will be brought together.

Thanks,

-Pat

12

Hi,
we’re using wpt version 2.5 with your agent-patch and we know observe another failure of the same FF-Job (see above):
Approximately every 60 job-runs doc ready- AND fully loaded-time are about 0.1 seconds and the result doesn’t contain any request-data at all. Screenshot shows our wait-page. The Measurement stopped after onload-event of first html-request apparently. Interestingly that happens either in first view (measurement of second view OK) or in second view (measurement of first view OK than).
Regards, Nils

Attachement shows a result with failed first view-measurement.

13

Hi,
is your agent-fix for firefox-measurements included in version 2.7.2 and/or in version 2.8?
We actually have the effect of measured wait-pages where redirects are involved with wpt-instances version 2.7.2 and 2.8.
I reproduced that on public instance: http://www.webpagetest.org/result/121122_Q3_FMN/
Regards Nils

14

Are there any news on this topic?
I also noticed that problem today with wpt 2.8 and “execAndWait” which does not wait for all browser activity to complete. Instead it stops waiting after the first page is loaded. It would be helpful if it would also wait for javascript calls which may redirect to further URLs.
Currently I don’t have a proper workaround for that.

Thanks in advance,

Kai