Issue metadata
Sign in to add a comment
|
Regression: Downloaded files appear blurred on downloads page on zooming out to 90%
Reported by
lpa...@etouch.net,
Jul 7 2016
|
||||||||||||||||||||||||
Issue descriptionChrome Version: 53.0.2785.8 (Official Build) c6d43518027ef223fe101efd8baa5a9b1bc979b4-refs/branch-heads/2785@{#37} 32/64 bit OS: Windows(7,8,10), Mac(10.10.5)(10.11.4), Linux(Ubuntu 14.04 LTS) Steps: 1. Launch chrome and press Ctrl+J. 2. Have at least two downloaded files in the downloads list. 3. Zoom out to 90% and observe the files below the first file. Actual: The files appear blurred on zooming out. Expected: The files should appear clearly on zooming out. This is a regression issue broken in 'M-48' and below is the manual regression range. Good Build: 48.0.2536.0 Bad Build : 48.0.2537.0 Narrow Bisect: https://chromium.googlesource.com/chromium/src/+log/e4471ed4dce8f2fc53e3e7f54aa8c4e18b93ea8c..e143ef44edfa1ea8cda564edd457f5ca644203e7?pretty=fuller&n=1000 Suspecting: r354221? Kindly help to re-assign if your change is not the cause of this issue.
,
Jul 15 2016
The download items have will-change: transform; because iron-list will fly them around, I wonder if this is because we now don't re-raster when you set will-change: transform and the scale changes. We totally should re-raster when zooming though.
,
Jul 15 2016
,
Jul 15 2016
I think that bisect is wrong?
,
Jul 15 2016
Zoom isn't implemented by transform, so I'd be surprised if that is the reason, but need to take a trace to see.
,
Jul 15 2016
,
Jul 18 2016
With response to comment #4 re-bisected again and getting the same range as mentioned. Kindly help to re-assign if your change is not the cause of this issue.
,
Jul 18 2016
If the bisect it right, can this be the patch from levi then?
,
Jul 18 2016
No the iron-list patch is the "cause". It made all of the download items composited because that's how iron-list does infinite scroll. Sounds like a compositor bug that the page just triggers now. Sorry I wasn't clear in my original reply, the bisect was correct.
,
Jul 18 2016
,
Jul 18 2016
ok thanks. fwiw reloading the page at 90% zoom stays blurry. investigating.
,
Jul 18 2016
Confirmed in a trace that the recording is not blurry. Both the first and second items in the list are in layers, however only the second one is blurry. The transform on the first one is 0,0,0 in the inspector. No partial pixel offset there. The transform on the actual layer is identity. The transform on the second one is 0,176,0 in the inspector. The transform on the actual layer in a trace is: [1, 0, 0, 0, 0, 1, 0, 158.39999389648438, 0, 0, 1, 0, 0, 0, 0, 1], Which is 176 * 0.9. So the page zoom is scaling the transforms but not snapping them to pixels. And non-integer-aligned layers produce blurry results. https://bugs.chromium.org/p/chromium/issues/detail?id=521364 is the bug for blurry under non-integer offset. I'm not sure if it is a bug that page zoom is not snapping to integer pixels on top of that. chrishtr@ can you comment, and dupe into 521364 if the latter is not a separate bug? (+oshima who's working on using page zoom for DSF fyi)
,
Jul 19 2016
Thanks, I confirmed that this happens on chromeos too.
,
Aug 1 2016
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by dbeam@chromium.org
, Jul 15 2016Components: Blink>Fonts Blink>Compositing
Owner: ----