1426594 - getBoundingClientRect returns the wrong result for an SVG <tspan>

Status: VERIFIED FIXED

(Core :: SVG, defect, P3)

Description

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Test case

See attached test case which draws a red box over the area returned by getBoundingClientRect() when called in the <tspan> corresponding to DEF in that text run.

This is similar to bug 1052356, but not the same, because in this test case I see incorrect results for a grandchild (the <tspan>), but more-or-less correct results for a child (the <text>) of the root <svg>.

Comment 1

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Also, the test case uses Element.getBoundingClientRect() but I see the same result for Range.getBoundingClientRect().

Comment 2

[Robert Longson [:longsonr]]

Seems about right for me on Firefox 57 on a Mac.

Comment 3

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Screenshot

This is what I see.

(Sorry about the excessively large screenshot -- still need to find image editing software for this computer.)

Comment 4

[Robert Longson [:longsonr]]

What version of Firefox and what Operating System are you on here? Is this a regression at all?

Comment 5

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Nightly Firefox / Ubuntu 16.04. No idea if it's a regression -- I don't have any other versions of Firefox on this computer.

Comment 6

[violet.bugreport]

I couldn't reproduce the bug on Ubuntu with latest Firefox, thus I mark it as WFM. Feel free to modify if anyone has more precise info.

Comment 7

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Still reproduces for me on Nightly on Linux.

Comment 8

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Result in Ubuntu on Nightly from 2019-04-11

Comment 9

[violet.bugreport]

That's strange, my test was also on Ubuntu (Ubuntu 18.04) but it's fine. Is it a Ubuntu version problem or some particular setting on your machine?

Comment 10

[violet.bugreport]

Attachment: Cannot-reproduce-on-Ubuntu-18.04.png

Comment 11

[Brian Birtles (:birtles, Feb 26~ parental leave)]

I'm also on Ubuntu 18.04.2 LTS but the fonts we're getting are clearly different. I'm getting DejaVu Sans but you're getting some sort of serif font. I don't suppose you have any non-standard font settings? I checked mine but couldn't find anything obvious and running on a clean profile I still get DejaVu Sans.

Comment 12

[violet.bugreport]

I don't suppose you have any non-standard font settings?

I don't have non-standard settings. But I don't think that's relevant, since I still can't reproduce even if I add font-family: DejaVu Sans; on a new profile with nightly.

Comment 13

[violet.bugreport]

Attachment: Still-cannot-reproduce-using-DejaVu-Sans.png

Comment 14

[violet.bugreport]

Is it possible that your system font was corrupted somehow? Probably we can ask qa to test whether it works on their Ubuntu machine.

Comment 15

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Attachment: Windows result

I can reproduce on Windows too using Meiryo.

Comment 16

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Judging by the offset and dimensions of the misaligned rects, I wonder if it's a simple scaling difference due to display density. Both the Linux and Windows machines I am testing on have high dpi screens.

Comment 17

[violet.bugreport]

Attachment: Bug 1426594 - Should convert rect from TransformFramePointToTextChild to device pixel

We need device pixel, but TransformFramePointToTextChild returns in css pixel,
thus a scaling is necessary.

Comment 18

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/5c98e7a25bc0
Should convert rect from TransformFramePointToTextChild to device pixel r=longsonr

Comment 19

[Andreea Pavel [:apavel]]

https://hg.mozilla.org/mozilla-central/rev/5c98e7a25bc0

Comment 20

[Robert Longson [:longsonr]]

I just noticed you used the .1f multiplication trick here too. Could you replace that with float() instead please. r=me on that.

Comment 21

[violet.bugreport]

(In reply to Robert Longson [:longsonr] from comment #20)

I just noticed you used the .1f multiplication trick here too. Could you replace that with float() instead please. r=me on that.

Ok. Filed a new bug since I've also used that in other places (I personally found it more readable, but it's just some stylistic stuff ...)

Comment 22

[Brian Birtles (:birtles, Feb 26~ parental leave)]

Thanks for fixing this. Confirmed that it works for me now!

Comment 23

[Timea Zsoldos [:zstimi/tzsoldos], Desktop QA]

The TestCase works for me, I verified using Firefox 68.0b7 on Win 10 x64, macOS 10.13.6 and Ubuntu 18.04 x64.