Difference of frame analysis private vs. public


I’m running a private instance with the nodejs agent and a local chrome browser on a debian jessie server. I have added some code to browser_local_chrome.js to record the video of my Xvfb server. In general it works, so I get the video, the frames are generated etc. all seems fine.

There are two issues compared to the public instance:

a) The start render is far earlier on my instance. Not sure, what the reason is. My video show the status line (of loading the content) already since the beginning (more or less), the video in the public version, shows the status line not before the page is shown. Maybe there is a difference in the processing?!

b) My video is always 10 sec. It is not stopped when all the content is loaded and executed. This leads to some wrong results. Any idea, why my video is not stopped immediately after but after 10secs? (seems a timeout or so - related to the waitForCoalesce? or should it be cut later when the processing happens?).

Would appreciate any hint or tip.


Seems like the latest code from the repo fixed this.

Question a) is still open…

If you have a sample video (mp4) I can try running it through the python vode that does the visual progress and metric extraction: https://github.com/WPO-Foundation/visualmetrics

My guess is that it is getting tripped up by the status line since I’ve only used the code to process mobile videos. Should be easy enough to tweak to have it ignore a strip at the bottom of the viewport for the initial render detection.