1519756 - srcset is ignored in some cases

Status: NEW

(Core :: Layout: Images, Video, and HTML Frames, defect, P2)

Description

[formicant]

Attachment: FirefoxSrcsetBug.zip

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:64.0) Gecko/20100101 Firefox/64.0

Steps to reproduce:

1. Visit an XKCD page (e.g. https://xkcd.com/2097)
2. Set zoom to a value greater than 100%
3. Close the page
4. Open the page again

Note:
In some sites (e.g. https://webkit.org/demos/srcset — see ‘srcsetTest2x.png’ in the attached archive), srcset works correctly.

Actual results:

The src image is shown.
(See ‘xkcd1x.png’ in the attached archive.)

Note:
The correct srcset 2x image appears after changing zoom or opening the developer tools.

Expected results:

The srcset 2x image should be shown.
(See ‘xkcd1x.png’ in the attached archive.)

Comment 1

[Peter Magyari (Desktop QA)]

Hi,

I've managed to reproduce this issue on Windows 10 and Ubuntu 16.04 using Fx 64.0.2(2019-01-08) and Fx Nightly 66.0a1(2019-01-17)

I will add a component so that the developers can take a look at it.

Thank you for reporting!

Comment 2

[Alice0775 White]

Disabled e10s seems fix the problem on Nightly66.0a1.

And I can reproduce the issue on Firefox41.0a1 if forcibly enabled e10s.
So, I think this is not recent regression. But this seems to be e10s implementation bug.

Comment 3

[Cristian Baica [:cbaica], Release Desktop QA]

9:37.71 INFO: Got as far as we can go bisecting nightlies...
9:37.71 INFO: Last good revision: 09f4968d5f42 (2015-02-17)
9:37.71 INFO: First bad revision: 9696d1c4b3ba (2015-02-18)
9:37.71 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=09f4968d5f42&tochange=9696d1c4b3ba

In this case, for the 'good' revision, the image is blurry when zooming in for the first time an even after closing and re-opening the page.
For the bad revision however, the image is blurry when it is first zoomed in, but then it is clear(correctly rendered) when the page is re-opened.

---

30:08.89 INFO: Got as far as we can go bisecting nightlies...
30:08.89 INFO: Last good revision: 3cc3b1968524248450c465c4ea2ee5596ffa65f2 (2015-11-12)
30:08.89 INFO: First bad revision: 0eaf345983b3afc2b426e25a3be93ebf0d93e6c1 (2015-11-13)
30:08.89 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3cc3b1968524248450c465c4ea2ee5596ffa65f2&tochange=0eaf345983b3afc2b426e25a3be93ebf0d93e6c1

For this regression range, for the 'good' revision, the image is blurred only when the zoom is done the first time, so when the page is re-opened, the image is clear and correctly rendered.
For the 'bad'revision, the image is always clear (on the first zoom operation and when the page is re-opened).

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(In reply to Cristian Baica [:cbaica], Release Desktop QA from comment #3)

9:37.71 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=09f4968d5f42&tochange=9696d1c4b3ba

This contains https://hg.mozilla.org/mozilla-central/rev/f8cb71f5eda1

30:08.89 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3cc3b1968524248450c465c4ea2ee5596ffa65f2&tochange=0eaf345983b3afc2b426e25a3be93ebf0d93e6c1

This contains https://hg.mozilla.org/mozilla-central/rev/1c1cde95bfa6

Comment 5

[Tim Spurway [:tspurway]]

Sean, do you think this is something we'd want a fix in 66 for?

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

To expand on comment 4 -- this regression range isn't particularly useful since it basically just points to when the feature was implemented (in 2 phases) -- and it's also not clear from comment 3 whether the two regression points being described are improvements or regressions.

Comment 7

[Sean Voisen (:svoisen)]

This doesn't seem great, but I don't think we'll push to fix it in 66 given the severity and potential age of the bug. Will set as P2 and try to get to it soon.

Comment 8

[Patricia Lawless]

Hi Sean, will this be fixed for 68?

Comment 9

[Sean Voisen (:svoisen)]

We won't have bandwidth to fix this regression in 68. I will put on our radar for 69.