Bug 90913 – Regression(121518) TextFieldDecorationElement formatting is broken

• Status: RESOLVED
• Resolution: FIXED
• Component: DOM

[Garrett Casto]

Before this change, a TextFieldDecorationElement was rendered on the right hand side text field. After it, it is rendered on the left side of the text field.

[Hajime Morrita]

Hmm. How can I try that?

[Garrett Casto]

Ah, sorry for not giving a way to reproduce. The only thing that I know of that uses this is a feature for generating passwords that I'm working on.

1) Add --enable-password-generation to the command line
2) Make sure that you are signed in with sync enabled for passwords.
3) Go to https://myaccount.nytimes.com/register

There should be a small key icon in the first password field on the left side of the field. It should be on the right side. It's possible that there is an easier way to test this, but I'm not sure what it is.

[Hajime Morrita]

I tried this on Chromium Linux ToT but I couldn't see any such icons. 
Am I doing something wrong?

[Garrett Casto]

Created attachment 153964 [details]
Linux UI

[Garrett Casto]

One other reason it would not show up is if you've every clicked on the "never remember passwords for this site" infobar for nytimes.com. To be safe, I'd just trying using a clean profile. I've attached an image of how the UI should look in case it's not obvious when it shows up.

[Hajime Morrita]

Oops. I accidentally used older binary. Now it shows the key icon.
Thanks.

[Hajime Morrita]

Created a reduction: http://jsfiddle.net/

[Hajime Morrita]

(In reply to comment #7)
> Created a reduction: http://jsfiddle.net/
Oops: http://jsfiddle.net/sggwj/

[Hajime Morrita]

Created attachment 155739 [details]
Patch

[Hajime Morrita]

Hi Dimitri, could you take a look?
This is happened to be a long awaited fix for me and Hayato-san :-)

[Dimitri Glazkov (Google)]

Comment on attachment 155739 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=155739&action=review

> Source/WebCore/dom/NodeRenderingContext.cpp:95
> +    ComposedShadowTreeWalker walker(m_node);

This area is super-hot. What are the performance implications of this change?

[Hajime Morrita]

(In reply to comment #11)
> (From update of attachment 155739 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=155739&action=review
> 
> > Source/WebCore/dom/NodeRenderingContext.cpp:95
> > +    ComposedShadowTreeWalker walker(m_node);
> 
> This area is super-hot. What are the performance implications of this change?
True. I observed certain regression on Dromaeo. 
I'm going to return here with ComposedShadowTreeWalker speedup.

[Hajime Morrita]

I want Bug 59816 to make it faster. IsInShadow flag will help up a lot.

[Eric Seidel (no email)]

Could you post the perf numbers to the bug?  Would be good to document how bad this is for perf.

[Hajime Morrita]

Created attachment 158540 [details]
Updated to ToT

[Hajime Morrita]

Created attachment 158541 [details]
Perf chart

Looks like I was wrong.
Attached chart shows no significant slowdown and show even slight speedup.
Now personally I'm rather optimistic
and would like to land this to see how it affects in Page cycler speed.

[Hajime Morrita]

(In reply to comment #16)
> Created an attachment (id=158541) [details]
> Perf chart
First two are patch applied.
Last two are original.

[Dimitri Glazkov (Google)]

Comment on attachment 158540 [details]
Updated to ToT

ok. Not flipping the cq+, because I would prefer if you land it and watch it :)

[Hajime Morrita]

Comment on attachment 158540 [details]
Updated to ToT

Fair enough. Let me flip it :-)

[WebKit Review Bot]

Comment on attachment 158540 [details]
Updated to ToT

Clearing flags on attachment: 158540

Committed r125739: <http://trac.webkit.org/changeset/125739>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.