speed.cloudflare.com

[UCyborg]

This site is unable to measure one's internet speed due to JavaScript error about some variable or something being undefined. The error shows for multiple JS files and its code is minified. Maybe only about:config tweak is needed? It works on Firefox 52.9. User agent doesn't seem to matter at first glance.

[moonbat]

It would help if you posted the actual error in code tags.

[Night Wing]

@ UCyborg

There is another site, which I use, and it will show you your internet speed. Have you tried Speedtest?

https://www.speedtest.net/

Once on the page, find your server to your ISP, then left click on the large "GO" button, wait about 10 seconds and then the site will automatically start to find your internet speed (download and upload).

[UCyborg]

The error:

Code: Select all

"Error fetching https://speed.cloudflare.com/__down?measId=5338652838411533&bytes=0: TypeError: i is undefined"

@Night Wing
I know about Speedest by Ookla, but I brought Cloudflare's site up because we're interested in having as most compatible browser as possible.

[Isengrim]

Off-topic:
I occasionally use fast.com which is run by Netflix.

Have you tried Speedtest?

I've found that some ISPs prioritize traffic to speedtest.net to make the speed results look faster than they are. Or at least they used to. I recommend running several tests at different sites to be sure.

[moonbat]

Can confirm the error on Pale Moon, the site requires WebGL, but that doesn't stop these 'undefined' errors. The page loads on Basilisk but the test doesn't run. Different set of errors -

Code: Select all

Error: WebGL: texImage2D: Conversion requires change inalpha-premultiplication.  
Error: WebGL: texImage2D: Alpha-premult and y-flip are deprecated for non-DOM-Element uploads.

[Night Wing]

Off-topic:
I occasionally use fast.com which is run by Netflix.

I've found that some ISPs prioritize traffic to speedtest.net to make the speed results look faster than they are. Or at least they used to. I recommend running several tests at different sites to be sure.

Off-topic:
Looks like a little comparison is in order.

I live in a very small city (population 1,300). This is what our local politicians call it, but I call it a small town and this city/town is in a rural county in southeast Texas. We do have broadband and my ISP is Comcast/Xfinity.

Using "Speedtest", my download internet speed is 315 Mbps. With "Fast", my download speed is 320 Mbps. Basically, winner and loser wise, it is a "draw". But if I'm going to nitpick, "Fast" is faster than "Speedtest" which makes your comment "moot".

[JustOff]

The console displays a lot of errors like:

Code: Select all

"Error fetching https://speed.cloudflare.com/__down?measId=728300315040175&bytes=0: TypeError: i is undefined"

due to unsuccessful execution of the following expression:

Code: Select all

var i = performance.getEntriesByName(h).slice(-1)[0]

// where "h" equals "https://speed.cloudflare.com/__down?measId=728300315040175&bytes=0"

The regression happened while removing telemetry stubs in Issue #929 (UXP), where the Performance interface was somehow broken.

[Isengrim]

Off-topic:
Looks like a little comparison is in order.
Using "Speedtest", my download internet speed is 315 Mbps. With "Fast", my download speed is 320 Mbps. Basically, winner and loser wise, it is a "draw". But if I'm going to nitpick, "Fast" is faster than "Speedtest" which makes your comment "moot".

Off-topic:
Not saying it was a definite widespread thing that is still going on, just saying that I had seen it before and it's just something to keep in mind. It's still a good idea to check multiple sites if you have doubts.

[Moonchild]

The regression happened while removing telemetry stubs in Issue #929 (UXP),

Can you pinpoint the exact commit where this regressed?

[badnick]

The page loads on Basilisk but the test doesn't run.

Same for me but also on Pale Moon the page loads but no test.

[JustOff]

Can you pinpoint the exact commit where this regressed?

Telemetry: Remove stubs and related code

[Moonchild]

Right that was that huge difficult to review commit.
Might have been a bit overzealous removing code -- although nothing immediately stands out as related to DOM performance.

[adesh]

That is my commit. I can look into it if no one else is checking this, although I doubt I'd be able to find something quickly and logically. Most of my effort will be trial and error, and I may not find anything.

[Moonchild]

I'd appreciate it if you try. There's clearly something amiss with it that slipped by (all of) us.
I've already found one problem that I corrected for the upcoming release; quite possible there's something else causing this issue for the speed test/performance API.

[JustOff]

Yesterday I managed to narrow it down to the following two files:

Code: Select all

netwerk/base/nsLoadGroup.cpp
netwerk/base/nsLoadGroup.h

Rolling them back makes the site work, but a new error appears in the console, and I don't have enough time to deal with it yet.

[adesh]

Thanks JustOff! That's a nice pointer. Because reverting changes to js/src and perfmonitoring did not fix the issue.

EDIT: I have a working build. Sending a PR soon.