[Webkit-unassigned] [Bug 48110] getBoundingClientRect: Do not truncate the coordinates to integers

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=48110





--- Comment #19 from Nikolas Zimmermann <zimmermann at kde.org>  2010-11-18 00:36:03 PST ---
(In reply to comment #18)
> Created an attachment (id=74204)
 --> (https://bugs.webkit.org/attachment.cgi?id=74204&action=review) [details]
> Fix and layout tests
> 
> > So why don't you change the expectations in the test file, so it says PASS?
> 
> I spent some time trying to figure out how to reconcile Chromium vs. non-Chromium WebKit results, until (silly me) it eventually occurred to me that the place to put font-dependent (therefore platform-dependent) coordinates is the expectation file, not the source HTML.
> So I added SKIP besides the usual PASS/FAIL and one of the tests, and I removed shouldBe() calls altogether in the other one.
> 
> In general, IMO, the shouldBe calls should be placed (no pun intended) only where there's a good reason to expect a certain number, or a certain formula. Everything else should be checked via expectation files.

Hrm, I don't (In reply to comment #18)
> Created an attachment (id=74204)
 --> (https://bugs.webkit.org/attachment.cgi?id=74204&action=review) [details]
> Fix and layout tests
> 
> > So why don't you change the expectations in the test file, so it says PASS?
> 
> I spent some time trying to figure out how to reconcile Chromium vs. non-Chromium WebKit results, until (silly me) it eventually occurred to me that the place to put font-dependent (therefore platform-dependent) coordinates is the expectation file, not the source HTML.
> So I added SKIP besides the usual PASS/FAIL and one of the tests, and I removed shouldBe() calls altogether in the other one.
> 
> In general, IMO, the shouldBe calls should be placed (no pun intended) only where there's a good reason to expect a certain number, or a certain formula. Everything else should be checked via expectation files.

Hm, you are also zooming non-text stuff like <div> or <svg> - I still don't see why the numeric values, after zooming eg. once or twice shouldn't be deterministic and non-font dependant?

Hah, I just realized that fast/dom/Element/getBoundingClientRect, doesn't zoom SVGs, I'm sorry overlooked this fact! So it's definately font dependant.

Sorry for the confusion :-)
Ok, new suggestion: Use the "Ahem" font, that should give you exactly the same results on all platforms (hopefully). Text will appear as black blocks (rects), all of the same width/height and advance.
Can you give that another try?

I'd love to avoid SKIP ... messages, or not checking something.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.