MyBB Internal: One or more warnings occured. Please contact your administrator for assistance.
WebPagetest Forums - FirstPaint value displayed zero in few test runs

WebPagetest Forums

Full Version: FirstPaint value displayed zero in few test runs
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
FirstPaint value in some of the test runs are displayed as zero. I run most of my tests 3 times and randomly i see a FirstPaint value as zero.

For example:

Test Run1, Test Run2 has a value and Test Run3 can be 0. In other test runs Test Run1 can be 0 Test Run2 and TestRun3 displays a FirstPaint value.

This is specific to FirstPaint metric. This happened consistently in Firefox, now i see this happening IE11 too. I am unable to identify the root cause. Did some one come across this issue and found a solution?
What agent version are you running? Try grabbing the latest and see if that fixes it:

I can't remember how far back it was when I made the change but I changed how video is captured and how render detection works to deal with issues with GPU-accelerated rendering.
Hello Patrick,

I am currently using the following versions:

update.ini: ver=378
wptupdate.ini: ver=181

I was using a slightly older version from mid May, 2014 before posting this thread. However, updating the agents doesn't help. I observed several tests results since the update and they still show zeros in the FirstPaint value. I am executing the tests in IE11.
If you enable video capture do you get the video frames in the filmstrip?
Yes, i enabled video recording in my tests and can see video frames in the filmstrip. I can also play videos.

I've attached the plot which shows the flakiness (Firstpaint value is zero on few test runs). Red line from the plot represents first paint first view value.
Here is an additional plot where i have first paint as 0 randomly. Each colour indicates a different URL and all the URL's are tested on the same machine and same browser(IE 11).
Just wanted to follow up if any one has insights about the flakiness ? The filmstrip works and the video works even when the first paint value is zero.
Can you publish one of the failed tests to the public instance (if it isn't sensitive)? I'd like to see if anything jumps out in the filmstrip for why it would be having issues.

Are they straight url tests or are they multi-step scripts?

Is there anything that shows up in the screen shots (like other software over the browser window)?
Hello Patrick,

I am sorry for responding late. My priorities changed and was not focussed on WPT for a while.

1) The tests i mentioned are running against dev environments which has features that are not public yet, i am afraid if i can make the results public

2) The test URLs are straight URLs not multi-step scripts

3) Yes, the screen shots work on the failed cases(where firstpaint is zero)

After my previous post i realised that StartRender(a custom WPT metric) is a better way to measure the time taken to show the first pixel on the screen when compared to FirstPaint(as reported by the browser). We decided to replace FirstPaint with StartRender as one of our performance KPI's. StartRender is not flaky and is a better representation wrt the film strip view
Reference URL's