1923150 - Text layer and canvas mismatch on Windows for people using an old version of pdf.js with Firefox 133

Status: NEW

(Core :: Graphics: Text, defect)

Description

[Calixte Denizet (:calixte)]

This bug isn't about the built-in pdf viewer in Firefox.

For the context, a lot of websites are using pdf.js and it's pretty frequent that they use an out-of-date one.
The fix for bug 1919906 has induced a regression: bug 1922063.
Because of this regression, the text selection and the find-in-page can be painful.
It's fixed in Firefox 133 for the built-in viewer and there is a fix too in the very last release of pdf.js (as a js lib).
But nothing is fixed for people using Firefox 133, on Windows, with an out-of-date pdf.js.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

:jfkthame, since you are the author of the regressor, bug 1919906, could you take a look?

For more information, please visit BugBot documentation.

Comment 2

[Jonathan Kew [:jfkthame]]

Simply reverting bug 1919906 would (obviously) reintroduce the bad rendering that it fixed.

https://bugzilla.mozilla.org/show_bug.cgi?id=1922063#c10 discusses some options, but none of them (except the "update pdf.js" option) seem risk-free.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

I think we should get bug 1920733 landed so that we can at least get a better picture of its risk

Comment 4

[Jonathan Kew [:jfkthame]]

Yes, let's get that into Nightly and see how things go. I'll try to take a look at the test failures there.

(We need to keep in mind, though, that until we decide we're comfortable shipping it to all Release users, that will in effect "hide" this issue on Nightly (and early Beta) channels, but end users will still experience the problem.)

Comment 5

[BugBot [:suhaib / :marco/ :calixte]]

The bug is marked as tracked for firefox133 (nightly). However, the bug still isn't assigned, has low priority and has low severity.

:marco, could you please find an assignee, increase the priority and increase the severity for this tracked bug? Given that it is a regression and we know the cause, we could also simply backout the regressor. If you disagree with the tracking decision, please talk with the release managers.

For more information, please visit BugBot documentation.

Comment 6

[Erich Gubler [:ErichDonGubler]]

:jrmuizel: Looks like some recent progress has been made on bug 1920733. Any thoughts on risk assessment now?

Comment 7

[Jeff Muizelaar [:jrmuizel]]

(In reply to Erich Gubler [:ErichDonGubler] from comment #6)

:jrmuizel: Looks like some recent progress has been made on bug 1920733. Any thoughts on risk assessment now?

bug 1920733 hasn't made it to nightly yet so we don't know anything new.

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

This is a reminder regarding comment #5!

The bug is marked as tracked for firefox133 (nightly). We have limited time to fix this, the soft freeze is in 10 days. However, the bug still isn't assigned.

Comment 9

[Donal Meehan [:dmeehan]]

:jrmuizel/:jfkthame next week is the final week of nightly before Fx133 goes to beta.
Any updates here on a possible fix?
We could consider backing out Bug 1919906 when we get to beta

Comment 10

[Jonathan Kew [:jfkthame]]

The current state of things, as I understand it:

As far as I know, bug 1920733 should have fixed the issue here, but the change there is currently targeting Nightly/Early Beta only, so as things stand the issue will reappear in later betas.

Whether we can un-gate the change from bug 1920733 and let it ship to Release will depend on any regressions that are discovered (as well as how the change is received in general, I guess; bug 1924346 is an instance of someone who noticed the change and doesn't sound happy with it).

Backing out bug 1919906 should be a "safe" option inasmuch as it will resolve the issue for pdf.js, but it will obviously re-introduce the poor canvas text rendering seen on Apple Maps (and presumably affecting other sites as well).

We could attempt to fix bug 1919906 in the "opposite" way: instead of ensuring that force-gdi-classic mode is consistently ignored for canvas2d text, we could allow it to be respected during both measurement/layout and rendering. This would in effect mean reverting bug 1730772, which introduced the inconsistency that results in bug 1919906. However, I have confirmed with a local build that doing this would regress the GDocs rendering that bug 1730772 was attempting to fix.

On balance, I think I'd lean toward backing out bug 1919906 from beta if we conclude that we're not ready to let bug 1920733 go to release yet. Jeff, does that seem like the right option at this point?

Comment 11

[Jeff Muizelaar [:jrmuizel]]

Yeah, I think backing out bug 1919906 from beta makes the most sense. I think we can afford to wait for bug 1920733 to cook a little bit.

Comment 12

[Donal Meehan [:dmeehan]]

Thanks all for the input, I will backout Bug 1919906 from beta after Fx133 merges to beta.

Comment 13

[BugBot [:suhaib / :marco/ :calixte]]

This is a reminder regarding comment #5!

The bug is marked as tracked for firefox133 (nightly). We have limited time to fix this, the soft freeze is in 3 days. However, the bug still isn't assigned.