Bug 100928 – [EFL][WK2] Change background color as white on pixel tests - 1

• Component: WebKit EFL
• Status: RESOLVED INVALID

[Kangil Han]

Splitted patch due to file size.

[Kangil Han]

Created attachment 171794 [details]
patch

[Gyuyoung Kim]

Comment on attachment 171794 [details]
patch

Looks good to me. Someone else might to take a final look.

[Laszlo Gombos]

I think this has impact on the Gtk port as well.

[Kangil Han]

(In reply to comment #3)
> I think this has impact on the Gtk port as well.

Yes, you are correct. GTK uses cairo as well.
One thing weird is expected result in GTK port has no gray color background.
I will check gtk port regression and be back.

[Gyuyoung Kim]

(In reply to comment #3)
> I think this has impact on the Gtk port as well.

Yes, right. Thanks.

[Raphael Kubo da Costa (:rakuco)]

I don't understand this change. The gray mask was added on purpose in r131941.

[Kangil Han]

(In reply to comment #6)
> I don't understand this change. The gray mask was added on purpose in r131941.

You've done huge rebaselining. :)

Don't get me wrong.
I will happily follow if decision has been made.
But, few questions here. :)

    Do we have special reason why we've selected gray mask?
    (Actually color is black but have transparent value though.)

r131941 has described all ports would have reached to agreement having gray mask.

    But, if so, why do we have port specific expected result?
    Can we simply replace the common expected png files?

From the regression testing point of view, gray background color has no problem.

    But, we can see the difference when we load the test page on MiniBrowser.
    Do we really want this?

[Raphael Kubo da Costa (:rakuco)]

(In reply to comment #7)
> But, few questions here. :)
>
>     Do we have special reason why we've selected gray mask?
>     (Actually color is black but have transparent value though.)

It's what all the ports such as chromium and mac use for the repaint tests.

> r131941 has described all ports would have reached to agreement having gray mask.
>
>     But, if so, why do we have port specific expected result?
>     Can we simply replace the common expected png files?

For the same reason we normally can't have cross-platform layout results: differences in font rendering, form controls etc. I don't doubt that we now have _some_ duplicate results; if you're willing to work on making sure if that's true or not, be my guest :-)

> From the regression testing point of view, gray background color has no problem.
>
>     But, we can see the difference when we load the test page on MiniBrowser.
>     Do we really want this?

What do you mean? Some pixel tests have no meaningful results without the gray mask.

[Kangil Han]

Okay, let's keep current implementation as rakuco mentioned here that other ports do the same.