1352083 - Content shifts while scrolling on print preview page with APZ enabled

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect, P2)

Description

[Kartikaya Gupta (email:[e-mail])]

Linux Nightly, APZ enabled (i.e. default configuration). Go to http://conorlastowka.com/book/CitationNeededBook-Sample.pdf, select "Print" from the hamburger menu. This will open the print preview window. Scroll down using the mousewheel and observe where the content is on the page in the print preview. When I do it I see the content shifts down on the page and gets chopped off as I scroll.

If I turn off APZ (layers.async-pan-zoom.enabled=false and restart the browser) the issue doesn't happen and the print preview looks more correct.

Comment 1

[ovidiu boca[:Ovidiu]]

Do you have any idea from when is this regression?

Comment 2

[Kartikaya Gupta (email:[e-mail])]

No. It would be useful to try to get a regression window I guess.

Comment 3

[ovidiu boca[:Ovidiu]]

I will investigate this, and try to get a regression window.

Comment 4

[ovidiu boca[:Ovidiu]]

I did a regression and here are the results: 

19:00.49 INFO: Last good revision: a5510966f80b9b2f5abf59ab32cf4c92d66c60de
19:00.49 INFO: First bad revision: b5c1f385ee5209796b4808cc9c40c0cb0cb4521a
19:00.49 INFO: Pushlog:
https://hg.mozilla.org/integration/fx-team/pushloghtml?fromchange=a5510966f80b9b2f5abf59ab32cf4c92d66c60de&tochange=b5c1f385ee5209796b4808cc9c40c0cb0cb4521a


Please take a look and tell me if you can find something useful from this regression window.

Comment 5

[Kartikaya Gupta (email:[e-mail])]

Considering the pushlog contains a PDF.js update that seems like the most likely candidate, given the document we're trying to print is a PDF. It might just have exposed a pre-existing APZ bug though.

Comment 6

[Brendan Dahl [:bdahl]]

There were some changes to the PDF.js printing CSS in that bug (https://hg.mozilla.org/integration/fx-team/diff/b5c1f385ee52/browser/extensions/pdfjs/content/web/viewer.css). I haven't verified this, but I imagine this bug could be reproduced without any PDFs and using similar styling.

Comment 7

[Kartikaya Gupta (email:[e-mail])]

So I tried backing out the CSS change in comment 6 to verify, and that didn't fix the problem. Then I tried verifying the regression range in comment 4, and both of those revisions worked fine for me. I ran mozregression myself and got this regression range:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3460a22b7c62160047431f88eaca952f8a0e0624&tochange=6e02d48ef1226c6f8c7504c05dac113794d65f07

which is also quite plausible.

Comment 8

[Kartikaya Gupta (email:[e-mail])]

I verified by locally re-enabling accelerated layers on Linux, and confirmed that the issue went away.

Comment 9

[Florian Bender]

So can we close this then?

Comment 10

[Kartikaya Gupta (email:[e-mail])]

I'll check tomorrow if it's still an issue. If it is I think we should leave it open.

Comment 11

[Kartikaya Gupta (email:[e-mail])]

It's still an issue on my local m-c build.

Comment 12

[Botond Ballo [:botond]]

This appears to have since been fixed.

Comment 13

[Botond Ballo [:botond]]

It was fixed sometime in the 63 cycle. I got a 1-day fix window, but it's not clear what change in that window fixed it, and I can't get a narrower window:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=5a8107262015714d2907a85abc24c847ad9b32d2&tochange=690cb3015db6645b335ac4835a50073cb6a3e23c