WSJ issues

[csillag]

@dwhly ran into this a few days ago.

URL: http://online.wsj.com/articles/climate-science-is-not-settled-1411143565
Chrome. Latest live extension.

Steps to reproduce:

1. Go to the given URL.
2. Wait until the page finishes initializing (which might take longer than a minute)
3. Deploy H

Expected result: it works;
Actual result: it doesn't :)

In fact, deploying H on the page unleashes an unending torrent of error messages on the console, most of which does not even come from our code, but theirs.

I have not yet spent time to understand what the issue is; it might also be a problem with the site's code. (Maybe we can work around it, maybe not.) Or it might be our bug.

[csillag]

It turns out all the error messages come from their code, from a component called "tealium universal tag - utag.loader". (See http://tealium.com )

Looks like they are trying to harvest some data from the DOM, and that does not play nicely with that we are doing with the DOM. (This is just a hypothesis, for now.)

If this tool is widely used, we might want to look into how to avoid disrupting it.

[tilgovi]

I just loaded h on this page, in both FF and Chrome (extension in the latter), and see 38 annotations and everything works fine (I do see the console logs from utag.js). What am I missing?

[csillag]

What am I missing?

The unending cycle of error messages appearing on the console, which
start to flow as H is deployed. It looks like we break utag somehow.

We might still have H functionality, but at the cost of CPU resources
being constantly wasted, and some broken functionality of the host page.

[tilgovi]

Right, but you wrote:

Expected result: it works;
Actual result: it doesn't :)

That doesn't appear to be the case.

[csillag]

For me, with Chrome, latest extension, the CPU load is so bad that the
sidebar never appears.

1. okt. 23., cs 16.58-kor-kor, Randall Leeds
   [email protected] írta:

   Right, but you wrote:

   Expected result: it works;
   Actual result: it doesn't :)

   That doesn't appear to be the case.

   —
   Reply to this email directly or view it on GitHub.
   =

[dwhly]

For me, with Chrome, latest extension, the CPU load is so bad that the sidebar never appears.

For me, the sidebar shows up fairly quickly, but the annotations don't appear for about 45s.

[csillag]

Given the fact that it actually triggers the "page crashed" chrome dialog
for me, I think it's fair to day that it's broken.
2014.10.23. 17:16 ezt írta ("Dan Whaley" [email protected]):

For me, with Chrome, latest extension, the CPU load is so bad that the
sidebar never appears.

For me, the sidebar shows up fairly quickly, but the annotations don't
appear for about 45s.

—
Reply to this email directly or view it on GitHub
#1587 (comment).

[tilgovi]

I have no such problems.
On Oct 23, 2014 5:16 PM, "Dan Whaley" [email protected] wrote:

For me, with Chrome, latest extension, the CPU load is so bad that the
sidebar never appears.

For me, the sidebar shows up fairly quickly, but the annotations don't
appear for about 45s.

—
Reply to this email directly or view it on GitHub
#1587 (comment).

[csillag]

2014.10.23. 17:18 ezt írta ("Randall Leeds" [email protected]):

I have no such problems.

It might depend on CPU muscle.
(But we shouldn't rely on that.)

[aron]

@csillag is this related to #1624? Could you try this again with latest master.

[csillag]

I checked again with a local extension, build from latest master.
It's still very slow for me, and I also see messages like on the console:

"Blocked a frame with origin "chrome-extension://lcknipjijijafjbionbfpoehnlfdknhh" from accessing a frame with origin "http://online.wsj.com". The frame requesting access has a protocol of "chrome-extension", the frame being accessed has a protocol of "http". Protocols must match."

I don't see messages like that on other documents.

[tilgovi]

I think that might appear on any page with multiple frames. I see it not infrequently. I could check, but that's my guess without spending time on it now.

The reason would be the bridge plugin traversing all frames and sending beacons to potential guests so that they might acknowledge and connect to the sidebar. IIRC that traversal doesn't throw any exceptions or cause the code to halt anywhere but it does have the side effect of spitting out these warnings.

I think this may be the line that, harmelessly, causes this: https://github.com/hypothesis/h/blob/master/h/static/scripts/plugin/bridge.coffee#L167

[BigBlueHat]

@tilgovi any guess where that code lives now?

@csillag is this still an issue for you? Getting support requests that match this issue in both FF and Chrome on Ubuntu 14.04.

Thanks! ☕

[tilgovi]

Doesn't appear to be valid anymore. Not seeing any strange messages in the console.