930768 - [OS X] Going back/forward in history is much slower with swipe animations

Status: RESOLVED FIXED

(Core :: Widget: Cocoa, defect)

Description

[José Jeria]

Tested on MacBook Pro (early 2013) with Retina display
Build ID: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:25.0) Gecko/20100101 Firefox/25.0

This was reported in bug 860493, comment 11: 
"it takes about a second to get the actual page to show. It actually feels slow and lagging like this."

Steps to reproduce:
1. go to any webpage
2. scroll down a bit and click any link
3. go back using the trackpad with 2 fingers to trigger the swipe animation

Results:
The going back and forward now has a nice animation like Safari, but it also does make Firefox feel slower. The page you go back (or forward) to needs to be reloaded every time. This was not the case before (is "Back-Forward Cache"-feature no longer used?).

Not only does it have to reload the page, but I am also seeing this effect when going back:
1. page you are going back to is shown beneath the back animation
2. current page is shown again for a millisecond
3. page goes blank (white)
4. Top of the page of the page you are going back to is visible
5. Page jumps down to the position you were at when you left it

Comment 1

[Stephen A Pohl [:spohl]]

Hmm, the fact that we're seemingly no longer using bfcache is disturbing to me. I'm going to suggest backing out bug 860493 until we've figured this out. We shouldn't be refreshing pages where we haven't previously.

Comment 2

[Stephen A Pohl [:spohl]]

Attachment: Backout bug 860493

Comment 3

[Stephen A Pohl [:spohl]]

Comment on attachment 822073 [details] [diff] [review]
Backout bug 860493

Review of attachment 822073 [details] [diff] [review]:
-----------------------------------------------------------------

I should probably have referenced the changeset rather than the bug number that I'm backing out in the commit message. I've changed the commit message to "Bug 930768: Backout 6155ea7e8dea [...]" and will commit that instead.

Comment 4

[Stephen A Pohl [:spohl]]

(In reply to José Jeria from comment #0)
> Steps to reproduce:
> 1. go to any webpage

Could you reference a particular website that causes this issue for you? Just want to be sure that we're looking at the same thing. Thanks!

Comment 5

[Steven Michaud [:smichaud] (Retired)]

Comment on attachment 822073 [details] [diff] [review]
Backout bug 860493

Oh well.

Yes, it's probably best to back this out until you've had a chance to figure out where the trouble is coming from.

Comment 6

[Stephen A Pohl [:spohl]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a25982be666c

Comment 7

[José Jeria]

(In reply to Stephen Pohl [:spohl] from comment #4)
> (In reply to José Jeria from comment #0)
> > Steps to reproduce:
> > 1. go to any webpage
> 
> Could you reference a particular website that causes this issue for you?
> Just want to be sure that we're looking at the same thing. Thanks!

I tried this one www.dn.se, but it can be seen on basically any website. I coudnt reproduce the last 5 steps again at work, where we have a fast internet connection. I guess that you see the page going blank on a slower connection. I ran into a weird issue though when trying to reproduce this, see bug 930905.

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/a25982be666c

Comment 9

[Stephen A Pohl [:spohl]]

Hmm, I thought I could reproduce this bug, but I'm not so sure anymore. I find that http://www.dn.se reloads the page no matter what type of history navigation I use. Swiping with animation reloads, back/forward button reloads and swiping with animations turned off (by setting browser.snapshots.limit to 0 in about config) reloads the page too. I've even compared with Safari, and I'm seeing the same there.

I've also tried throttling my network bandwidth, with no effect.

José, would it be possible to get a screen capture from you? That would allow me to count frames and compare. Thank you!

Comment 10

[José Jeria]

Screen capture showing what was described in comment 0:
https://www.dropbox.com/sh/khv41eesjjcp3j0/sy5ifX9rUa?lst

i guess this happens because www.dn.se manipulates the history as described in bug 930905, comment 5

Comment 11

[Stephen A Pohl [:spohl]]

Thanks! Unfortunately, I can't really compare the speed of navigating the history from that screen capture. You would need to:
1. Ensure that the page is fully loaded before clicking on a link or swiping, i.e. no spinning wheel anymore.
2. Navigate history with swipe first, then go back and do the same with the back/forward buttons in the same screen capture.

We could then count the frames and compare.

Comment 12

[José Jeria]

(In reply to Stephen Pohl [:spohl] from comment #11)
> Thanks! Unfortunately, I can't really compare the speed of navigating the
> history from that screen capture. You would need to:
> 1. Ensure that the page is fully loaded before clicking on a link or
> swiping, i.e. no spinning wheel anymore.
> 2. Navigate history with swipe first, then go back and do the same with the
> back/forward buttons in the same screen capture.
> 
> We could then count the frames and compare.

I added 2 new screen captures: https://www.dropbox.com/sh/khv41eesjjcp3j0/sy5ifX9rUa?lst
swipe_dn and swipe_verge. Swiping works fine on the verge, its just dn.se causing issues.

Comment 13

[Stephen A Pohl [:spohl]]

(In reply to José Jeria from comment #12)
> I added 2 new screen captures:
> https://www.dropbox.com/sh/khv41eesjjcp3j0/sy5ifX9rUa?lst
> swipe_dn and swipe_verge. Swiping works fine on the verge, its just dn.se
> causing issues.

Great, thanks José! It looks like the speed of going back and forth in history is virtually identical. When you say "just dn.se causing issues", do you mean the additional page that is being added in between the first and the second? If so, I think your comment 10 is spot on, i.e. the history is being manipulated via pushState. I'll check and see if there's anything we can improve here.

Comment 14

[José Jeria]

Should this bug be reopened now that swiping animation has been re-enabled? Comment 1 is still an issue.

Comment 15

[Stephen A Pohl [:spohl]]

(In reply to José Jeria from comment #14)
> Should this bug be reopened now that swiping animation has been re-enabled?
> Comment 1 is still an issue.

I'm confused. I thought we've determined that the speed of going back-and-forth is virtually identical between swiping and the back/forth buttons (comment 12/comment 13). Instead, we seem to be running into bug 930905, comment 5. Is this not the case?

Comment 16

[José Jeria]

(In reply to Stephen Pohl [:spohl] from comment #15)
> I'm confused. I thought we've determined that the speed of going
> back-and-forth is virtually identical between swiping and the back/forth
> buttons (comment 12/comment 13). Instead, we seem to be running into bug
> 930905, comment 5. Is this not the case?

Sorry, please disregard my comment, I thought that the swiping caused "Back-Forward Cache" to stop working. This is not the case, I just verified it on bugzilla (list results page) and other websites. Sorry for the confusion.