1957469 - RFP: WindowDevicePixelRatio: chart distortion e.g. on kucoin

Status: NEW

(Core :: DOM: Security, defect, P5)

Description

[Thorin [:thorin]]

Attachment: example.jpg

from: https://codeberg.org/librewolf/issues/issues/2396

STR

• site: https://www.kucoin.com/trade/MAJOR-USDT
• load with out RFP
  • graph is populated, bars are thin, values are normal text size and fully visible etc
• load with RFP
  • graph is partially populated, x and y axes (plural of axis) are larger, text labels on axes are larger

Note: example is using nightly 138 but I can also reproduce in Mullvad Browser where we still report dPR as 1 and matching media queries (at 100% zoom)

Not sure what part of RFP is causing this (edit: PS: it's not canvas). Worth testing targets if someone has time

Comment 1

[Fatih Kilic [:fklc]]

I can't reproduce it :/

BUT if anyone would like to quickly find out which RFP target is causing the issue, they can use https://github.com/mozilla-extensions/fingerprinting-protection-debugger/releases/latest. It has a troubleshoot website button that binary "searches" rfp target similar to mozregression.

Comment 2

[Tom Ritter [:tjr]]

I can reproduce it, it's WindowDevicePixelRatio.. (It was actually really easy to reproduce because the chart redraws itself every second so you don't even have to reload when doing the troubleshooting.)

Comment 3

[Thorin [:thorin]]

but I can also reproduce in Mullvad Browser where we still report dPR as 1

me culpa - we report as 2, what we haven't backported is making it dynamic and aligning matchmedia

it's WindowDevicePixelRatio

indeed

Comment 4

[Thorin [:thorin]]

I've floated the idea with piero that we could return 1 or 2 based on the closest real value, so a binary outcome. This should then help mitigate sizing being overly-scaled, but still reduce blurriness (why we moved to 2 in the first place)

Comment 5

[BugBot [:suhaib / :marco]]

The severity field is not set for this bug.
:freddy, could you have a look please?

For more information, please visit BugBot documentation.

Comment 6

[Thorin [:thorin]]

(In reply to Thorin [:thorin] from comment #4)

I've floated the idea with piero that we could return 1 or 2 based on the closest real value, so a binary outcome. This should then help mitigate sizing being overly-scaled, but still reduce blurriness (why we moved to 2 in the first place)

FYI: https://bugzilla.mozilla.org/show_bug.cgi?id=1974034#c8

if limited to Windows

• 52.5% for 100% scale
• 16.8% for 125% scale
• 8.94% for 150% scale
• 1.32% for 200% scale

I think we should get stats per platform and control this per platform e.g. we're going to want 2 or higher for android and macs (and perhaps allow two values depending on the platform results)

Comment 7

[Thorin [:thorin]]

https://codeberg.org/librewolf/issues/issues/2598 shows another example of WindowDevicePixelRatio - I'm not a figma user, and I suspect the issues may run deeper than the image - but from what I can see in the images is the light colored "working area" is reduced in width with RFP