993325 - Size of radial gradient backgrounds incorrectly calculated

Status: NEW

(Core :: Web Painting, defect)

Description

[Sebastian Zartner [:sebo]]

Attachment: Test case

Open the attached file and resize the browser window horizontally.
Sometimes you'll see thin lines and sometimes not.

The radial gradient seems to be incorrectly positioned, causing this flickering.

This display error does not occur when the shadow is positioned at the left.

Sebastian

Comment 1

[Sebastian Zartner [:sebo]]

FWIW this display problem does not occur on Firefox 28.0 but in 29.0+.

Sebastian

Comment 2

[Mats Palmgren (inactive)]

Attachment: Testcase #2 - flexbox version

(just to rule out it's a table layout problem)

Comment 3

[Mats Palmgren (inactive)]

It appears to be a regression in FF29, I can't reproduce it in FF28 or 27.

Comment 4

[Alice0775 White]

Regression pushlog:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3635f2f0c4f6&tochange=13a5fb1e8525

Regressed by: Bug 700926

Comment 5

[Sylvestre Ledru [:Sylvestre]]

Tracking. However, if we want to fix for that, it should land before next Monday evening for the beta8.

Comment 6

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment 4 points at Nick - can you take a look at this?  We're too late for FF29 now unless you can safely backout bug 700926 and put us back to a known-good state.

Comment 7

[Lukas Blakk [:lsblakk] use ?needinfo]

Not seeing a huge user impact here based on the test cases, please do put up a forward fix (low risk) for uplift to 30, or look at a backout of 700926, but from what's in this bug this doesn't look serious enough to necessitate tracking.

Comment 8

[Sebastian Zartner [:sebo]]

FWIW this problem can also be seen within the column headers inside the Net panel of Firebug 2.0.

Nick, any update on this?

Sebastian

Comment 9

[Nick Cameron [:nrc]]

Hi, Sebastian, Lukas. Sorry this totally slipped off my radar. I haven't had a chance to look into this. I'm actually working on a totally different project now (Rust) and am busy with external events, etc. the next month or so, so I'm very unlikely to find time to look at this. I can try to fix this once that lot is over (end of August-ish). I'm also happy to work with someone else to fix this in the meantime.

Comment 10

[Sebastian Zartner [:sebo]]

> I'm also happy to work with someone else to fix this in the meantime.
So maybe you can CC someone else, which may help out?

Sebastian

Comment 11

[Nick Cameron [:nrc]]

Is anyone able to look into this? Matt, Seth, Markus?

Thanks!

Comment 12

[Sebastian Zartner [:sebo]]

Ping. Matt, Seth, Markus, can someone of you have a look at this issue?

Sebastian

Comment 13

[Matt Woodrow (:mattwoodrow)]

I probably won't have time to get to this in the near future.

Comment 14

[Sebastian Zartner [:sebo]]

Just tested this on 84.0.2, DevEdition 85.0b9 and Nightly 86.0a1 (2021-01-19) on Windows 10 with WebRender enabled and the problem still exists.

It seems Seth isn't active anymore, so I removed the related needinfo flag, but Daniel or Markus, maybe you can check what's going on here?

Sebastian

Comment 15

[Daniel Holbert [:dholbert]]

Sorry - too many things drawing my attention right now, so I'm not going to have time to look into this anytime soon, I don't think.

(It'd be great for someone to take a look at, perhaps as a good-second-or-third bug when getting into layout/graphics work. It doesn't seem especially high-priority to redirect people onto at the moment, though, given how long we've been shipping with this bug. If you know of sites that are affected by this, that might raise the priority a bit, though.)

Comment 16

[Sebastian Zartner [:sebo]]

The use case back then when I created this issue were the table headers in Firebug's Net panel. I currently don't have a concrete example for this issue, also because there are other ways to achieve a similar effect for that specific example.

It's still an incorrect calculation of the gradient's dimensions, though most of the time you won't notice it as you normally don't have such thin gradients. So, yes, it definitely doesn't have high priority.

Sebastian

Comment 17

[Clara Guerrero ( Need Info Brindusa Tot Please)]

Updating "has regression range" field accordingly as per comment 4.
Best,
Clara

Comment 18

[Sebastian Zartner [:sebo]]

I just tested this in the latest Nightly version (112.0a1) and it is still an issue there. For what it's worth, Blink doesn't have this problem.

Sebastian