1828757 - Snap Firefox is visually corrupted after updating to Ubuntu 23.04

Status: RESOLVED INCOMPLETE

(Core :: Widget: Gtk, defect)

Description

[nicocarbone]

Attachment: Captura desde 2023-04-18 17-50-07.png

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/112.0.0.0 Safari/537.36

Steps to reproduce:

I updated Ubuntu from 22.10 to 23.04.

Actual results:

Firefox window is visually corrupted. It seems to be working fine, but nothing is visible under the corruption. Screenshot attached.
No other application (even 3D games) are having any issues.
I have a AMR Radeon RX5500 using the default mesa drivers

Expected results:

Firefox window should be clean, as it was before the update.

Comment 1

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

The Bugbug bot thinks this bug should belong to the 'Core::Widget: Gtk' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[Darkspirit]

Thanks for the report! Please try to open about:support, click on "Copy text to clipboard" and paste it here.

Does the same problem occur with the Firefox tar.gz downloaded from https://firefox.com/new ?

Comment 3

[nicocarbone]

Sorry for the delay.

I can´t access about support because the interface is completely obfuscated by the artifacts.
The problem does not happen with the tar version.

And, as I just found out, neither with the beta channel of the snap firefox.

Comment 4

[Lee McFarland]

Could there have been something wrong within firefox (nightly) which didn't show until the latest release of Ubuntu?

Bug 1829447 got assigned by the BOT to component New Tab Page however, I wonder if this and 1829447 are really associated with a push for bug 1806766 on 12/25/22. At that time, Ubuntu 23.04 wasn't in use and, Firefox nightly didn't show any affects of this update. Suddenly with the availability of Ubuntu 23.04, mozregression is showing a failure starting Sun, 25 Dec 2022 21:52:26. If similar testing were done on this bug, would it show the same point at which the failure was introduced?

Comment 5

[Darkspirit]

(In reply to nicocarbone from comment #0)

I have a AMR Radeon RX5500 using the default mesa drivers

Snap on AMD seems to be affected by bug 1828196.

Comment 6

[:gerard-majax]

(In reply to nicocarbone from comment #3)

Sorry for the delay.

I can´t access about support because the interface is completely obfuscated by the artifacts.
The problem does not happen with the tar version.

And, as I just found out, neither with the beta channel of the snap firefox.

snap info snapd and snap info firefox ? Would be interesting to know if you are on core22 base.

Comment 7

[Darkspirit]

Please open about:support in the tar.gz Firefox and in Snap Firefox, and then try to find the position of the "Copy text to clipboard" button in the Snap Firefox.
Please also attach about:support of the tar.gz Firefox.

Comment 8

[nicocarbone]

Sorry again for the delay.

I have some developments. My GPU died a week after I reported this bug. I am running now on a spare Nvidia GPU I had laying around. So, I can't reproduce the issue.

But, more importantly, I am now considering if this bug was a result of a dying GPU... It only happened consistently in Firefox snap stable, but maybe it was related?

Comment 9

[:gerard-majax]

(In reply to nicocarbone from comment #8)

Sorry again for the delay.

I have some developments. My GPU died a week after I reported this bug. I am running now on a spare Nvidia GPU I had laying around. So, I can't reproduce the issue.

But, more importantly, I am now considering if this bug was a result of a dying GPU... It only happened consistently in Firefox snap stable, but maybe it was related?

Interesting, indeed from experience of dying GPU, that kind of corruption is quite expected. When you say it only reproduced with Firefox Snap Stable, do you mean you did not reproduce the issue with other versions of the Snap, or with non Snap Firefox ? Or are you just referring to your first comment and the fact that other processes would not repro ?

Also, is your spare GPU the same model ? No driver update in the meantime ?

Comment 10

[nicocarbone]

I couldn't reproduce the bug with the beta channel of the Firefox snap or with the stable tar version. Could a hardware issue only be triggered by an specific version of Firefox?

My spare GPU is a Nvidia GPU, the previous GPU was AMD. I had to install the proprietary Nvidia driver.

Comment 11

[:gerard-majax]

Ok, there's quite a lot of differences there. So AMD RX5500. If you cannot repro with same version of stable in snap and stable in tar, then I tend to think your corruption was unrelated and it was just a coïncidence your GPU died (last time I ran into a similar very-long-shot I tried my chance to the lottery, and won.).

Comment 12

[nicocarbone]

Yep, I agree. Seems weird that a hardware problem would differentiate between snap and tar versions of stable firefox. What does change between them is the mesa version used. I noticed that the snap version used a rather old version of mesa, 21.2 if I remember correctly. The tar version uses the system version of mesa, 23.0

Comment 13

[:gerard-majax]

Could be a bug on mesa that was fixed then.

Comment 14

[:gerard-majax]

Amin, I can't find the mesa version for core20 vs core22, do you know what versions we should be expecting on core22 ?

Comment 15

[:gerard-majax]

Checking on my system, 22.2.5.0 for core22. Can your GPU go through RMA? If so, once it comes back, would be interesting to check if you still have the corruption.

Comment 16

[:gerard-majax]

snap refresh --channel=candidate/core22 firefox

Comment 17

[nicocarbone]

On another system using mesa and running ubuntu 23.04, the firefox snap reports 21.2.6.0 in about:support. Maybe firefox stable is not using core22?

Sadly I can't RMA the card, it is way out of guarantee.

Comment 18

[:gerard-majax]

Yes, for sure core22 is not yet the default. Too bad for your RMA though.

Comment 19

[:gerard-majax]

Now core22 is there and the default