Pavan Kiran Reporter
	Description • 7 years ago

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:57.0) Gecko/20100101 Firefox/57.0
Build ID: 20170920135025

Steps to reproduce:

1.I opened canva.com.
2.went to the bottom and clicked "open a new logo design"



Actual results:

A blank page with "about:blank" in the search bar was opened.


Expected results:

A new page should have opened for me to register into the site.

	YF (Yang)
	Comment 1 • 7 years ago

STR:
1. Open https://www.canva.com/create/logos/
2. Click the "open a new logo design".


Actual results:
New tab loaded, but there is a clear delay between opened and loading.

	YF (Yang)
	Comment 2 • 7 years ago

Source: "//Redirect to the link after tracking is completed" in https://about.canva.com/wp-content/themes/canva/js/scripts.js?ver=1.0.0.

	YF (Yang)
	Comment 4 • 7 years ago

Regression range:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3611ff7a79e6b616088bc4bf14e7ba8cdfe8bebf&tochange=5a73c211f58d6dca5b6938970311111aeb8c6692

	:Gijs (he/him)
	Comment 5 • 7 years ago

Moving this per the regression window & regressing bug. :bkelly, can you take a look / prioritize this?

	Ben Kelly [:bkelly, not reviewing]
	Comment 6 • 7 years ago

The code from comment 2 does this:

      //Redirect to the link after tracking is completed
      var windowReferece = window.open(); //Safari doesn't allow window.open in async calls, so we need to reference it earlier
      canvaAnalytics.trackEvent('topnav_login', trackingProperties, {}, function() {
        windowReferece.location = destinationUrl;
      });

So its:

1. Opening a new window in the foreground.
2. Executes some async code via setTimeout() in the window which is now the background.

As documented for a long time we throttle timeouts in the background window:

https://developer.mozilla.org/en-US/docs/Web/API/WindowOrWorkerGlobalScope/setTimeout#Timeouts_in_inactive_tabs_throttled_to_%3E1000ms

Bug 1371664 fixed some inefficiencies in how we handle background timers.  A side effect of this is we notice windows are in the background sooner and throttle their timers more effectively.  This code has always been racing that mechanism and just happened to beat it in the past thanks to bugs in our implementation.

Solutions:

1. Don't place your current window in the background if you know you have more work to do immediately.  Do your work and then open the new foreground window.
2. Or perform that work in the newly opened window instead of the current window.

Unrelated note:

This site is using a full core all the time for some reason.  (This might slow timers down as well.)  I turned on paint flashing and it seems to be spinning some kind of transparent loading layer on top of all the logo template grid items at the bottom.

	Dennis Schubert [:denschub]
	Comment 7 • 7 years ago

Sounds like something we could do outreach on.

	Ben Kelly [:bkelly, not reviewing]
	Comment 8 • 7 years ago

Also, edge has a similar 1+ second delay on navigation since they also do aggressive background timer throttling.

	Liz Henry (:lizzard Please n-i to RyanVM, jcristau, or pascal)
	Comment 9 • 7 years ago

OK, presumably, we think the tradeoff is worth it, for whatever we gained in bug 1371664. Outreach sounds like the right approach here.

	Adam Stevenson [:adamopenweb]
	Comment 10 • 7 years ago

This site is ranked 217. NI myself for follow up.

	Adam Stevenson [:adamopenweb]
	Comment 11 • 7 years ago

Created an issue on the website and reaching out on Twitter as well.

	Adam Stevenson [:adamopenweb]
	Comment 12 • 7 years ago

Got a response back:

"As per the team, they suggest that don't place your current window in the background if you know you have more work to do immediately. Do your work and then open the new foreground window.

We effectively did this but we open the link in the current window instead of the foreground window."

It seems to work for me, doesn't open a new tab or window. Can someone confirm please?

	Mike Taylor [:miketaylr]
	Comment 13 • 7 years ago

(In reply to Adam Stevenson [:adamopenweb] from comment #12)

> It seems to work for me, doesn't open a new tab or window. Can someone
> confirm please?

Yes, this seems fixed now. Thanks Adam!