1235093 - Specific styling (layerization) makes downscaled images flicker on hover

Status: VERIFIED DUPLICATE of bug 1225934

(Core :: Layout: Images, Video, and HTML Frames, defect)

Description

[arni2033]

Attachment: testcase 1 - Images flicker on hover.htm

>>>   My Info:   Win7_64, Nightly 46, 32bit, ID 20151223030323
STR:
1. Open attached "testcase 1"
2. Hover mouse over each image

Result:       Each image flickers   [please request a screencast if you can't reproduce]
Expectations: None of the pictures should flicker

Note:   I saw this hapen with Youtube suggestions (inside the player) and on some other sites

Regressed by bug 1207355
> pushlog_url:   https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=99a4fb4ba5c1ee96ac745c6ad11894bf0537f73a&tochange=2e19045ba652ca2a5a5fc0e20d6f95293acfa32d

Comment 1

[Chris Lord [:cwiiis]]

On first fresh load, it flickers white - on a reload, it flickers a wrongly-sized image (possibly not respecting pixel scale?)

Comment 2

[arni2033]

Please also have a look at this page:   http://translate.google.com/about/intl/ru_ALL/
When I slowly autoscroll it to the bottom, the smartphone pic is visually rotating several times.

Bug 1207355 added flickering during the 1st "rotation" (between 1st and second sections).
Then bug 1209970 added flickering during all subsequent "rotations", including minor ones.
Request a screencast if you couldn't reproduce. Also, tell me if that link deserves a separate bug.

Comment 3

[Daniel Holbert [:dholbert]]

I see something similar to this bug -- images flickering to a wrong size -- when I first hover the photos at http://www.yelp.com/biz_photos/j-and-m-hobby-house-san-carlos , too.

(I can't track down a regression range for this particular page, because it's been separately-broken by bug 1226904 until very recently. But I'm pretty sure the flicker is this bug.)

Comment 4

[Daniel Holbert [:dholbert]]

Marking Firefox 45 as affected, per regression range in comment 0, and requesting tracking. (Tracking reason: recent regression, affects several large sites (Google Translate, Yelp) and is perceptible enough to be annoying & look definitely-broken.)

Comment 5

[Daniel Holbert [:dholbert]]

Attachment: screencast of bug on Yelp

Comment 6

[Daniel Holbert [:dholbert]]

Attachment: testcase 2 (reduced)

Here's a reduced testcase with just a single image (included as a data URI), and the only styling being:
 (1) a smallish "width" to downscale the image
 (2) "will-change:opacity" on hover, to trigger layerization

Comment 7

[Daniel Holbert [:dholbert]]

From playing around with testcase 2 a bit:
 * This bug only happens in testcase 2 if the image is downscaled to have a *width* of < 158px.
    So "width: 157.4px" triggers the bug, while "width: 158px" does not.
 * The image in that testcase has an intrinsic *height* which just happens to be exactly 158px.
 * This seems a bit suspicious... I wonder if we're mixing up width & height somewhere.

Comment 8

[Daniel Holbert [:dholbert]]

Attachment: testcase 3 (img downscaled to 157px wide)

Here's an example demonstrating comment 7. The image here has a width of 157px, which triggers the bug for me...

Comment 9

[Daniel Holbert [:dholbert]]

Attachment: semi-reference 3 (img downscaled to 158px wide)

...and here it is again w/ width="158px" instead. (which makes the bug go away, for me at least)

Comment 10

[Sylvestre Ledru [:Sylvestre]]

Recent regression, tracking it.

Comment 11

[Liz Henry (:lizzard) (relman/hg->git project)]

Daniel or Jet, can you help find an owner for this bug ? Thanks.

Comment 12

[Daniel Holbert [:dholbert]]

Seth or tn (author/reviewer of the regressing bug 1207355) would be ideal.

I think seth might be on PTO, though maybe he's back or returning soon (not sure). Timothy, perhaps you could take a look?

Comment 13

[Daniel Holbert [:dholbert]]

(Given comment 7, I'm hoping this is just a typo somewhere & the fix might be trivial.)

Comment 14

[Timothy Nikkel (:tnikkel)]

Hmm, there was another bug involving showing the wrong sized image recently that had more investigation with seth and mattwoodrow both looking into it. I can think of the bug right now, but I thought that bug produced a patch at some point? Finding that bug and seeing if its a dupe might be helpful.

Comment 15

[Timothy Nikkel (:tnikkel)]

Ah, that bug was bug 1225934. This seems like a dupe of that bug. Unfortunately the patch in bug 1225934 doesn't seem to even fix the testcase in bug 1225934, so we can't test the patch on the testcase in this bug.

Comment 16

[Seth Fowler [:seth] [:s2h]]

Marking as a dupe of bug 1225934; I'm pretty confident that they're the same. I'll double check that the patch for that bug fixes this bug, though.

Comment 17

[Liz Henry (:lizzard) (relman/hg->git project)]

seth did you get a chance to check this?  bug 1225934 looks like it isn't fixed yet. If this was a regression in 45 I want to make sure we resolved it somewhere.

Comment 18

[Daniel Holbert [:dholbert]]

bug 1225934 is now fixed, and I've verified that this is fixed in latest nightly (though it wasn't fixed in yesterday's nightly, which didn't have the fix).

Clearing needinfo from comment 17, and marking as VERIFIED dupe.

Comment 19

[Liz Henry (:lizzard) (relman/hg->git project)]

The fix in 1225934 was also uplifted to 46 so I can mark this fixed. 
Thanks Daniel. It's always a nice feeling to un-bother Seth.