Bug 253446 - REGRESSION(2.39.90): [GTK] Rendering errors on cnn.com
Summary: REGRESSION(2.39.90): [GTK] Rendering errors on cnn.com
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: WebKitGTK (show other bugs)
Version: WebKit Nightly Build
Hardware: PC Linux
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-03-06 09:55 PST by Michael Catanzaro
Modified: 2023-05-08 14:01 PDT (History)
4 users (show)

See Also:


Attachments
Screenshot (562.88 KB, image/png)
2023-03-06 09:55 PST, Michael Catanzaro
no flags Details
Another screenshot (479.93 KB, image/png)
2023-03-06 16:07 PST, Michael Catanzaro
no flags Details
Another screenshot (765.29 KB, image/png)
2023-03-28 15:34 PDT, Michael Catanzaro
no flags Details
Yet another screenshot (381.78 KB, image/png)
2023-03-30 09:20 PDT, Michael Catanzaro
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Catanzaro 2023-03-06 09:55:16 PST
Created attachment 465317 [details]
Screenshot

It looks like tiles are getting jumbled around when rendering cnn.com (see attachment).

Unfortunately I do not have a reliable reproducer, but I've seen it a few times recently and I believe this never happened prior to 2.39.90.
Comment 1 Michael Catanzaro 2023-03-06 16:07:02 PST
Created attachment 465326 [details]
Another screenshot

Another screenshot, this time with only the Mapbox widget broken
Comment 2 Michael Catanzaro 2023-03-27 14:13:56 PDT
This appears to be fixed on main, and only broken in the 2.40 stable branch.
Comment 3 Michael Catanzaro 2023-03-27 14:17:16 PDT
(In reply to Michael Catanzaro from comment #2)
> This appears to be fixed on main, and only broken in the 2.40 stable branch.

Actually I'm not confident about this. It's much harder to reproduce in my development MiniBrowser than it is in Ephy Tech Preview so that might just have been luck.
Comment 4 Michael Catanzaro 2023-03-27 14:25:13 PDT
Ah, I've found the bug appears to be related to window size. With larger window size, I can reproduce with MiniBrowser using 2.40.0. Can't reproduce using main.
Comment 5 Adrian Perez 2023-03-27 14:30:11 PDT
I cannot reproduce this on a 2.40.x build from the tip of the release
branch, but this is using GTK3. I'm making a GTK4 build to check, but
there is some little chance that it's fixed by one of the patches that
have been backported 🤔️
Comment 6 Adrian Perez 2023-03-28 12:15:11 PDT
(In reply to Adrian Perez from comment #5)
> I cannot reproduce this on a 2.40.x build from the tip of the release
> branch, but this is using GTK3. I'm making a GTK4 build to check, but
> there is some little chance that it's fixed by one of the patches that
> have been backported 🤔️

I cannot reproduce it either with a GTK4 build of the current tip of
the 2.40.x release branch (commit 6dcdac4b3d5a). I've tried changing the
MiniBrowser window size a few times, but even maximized on a 1920x1080
monitor (no HiDPI, @1x scale) there are no glitches.

Michael, what size exactly is the window when you see glitches? Or,
alternatively, could you check a build with the commit mentioned above?
Comment 7 Michael Catanzaro 2023-03-28 15:34:58 PDT
Created attachment 465646 [details]
Another screenshot

Here is a screenshot taken with webkitglib/2.40 (commit 4524c8a9e1d4ae627ea2279ffed0ff05f806adf0) showing the bug. Definitely not fixed on the stable branch.
Comment 8 Michael Catanzaro 2023-03-28 15:46:17 PDT
Anyway, it's impossible to be certain because the bug does not occur reliably, but I'm once again convinced myself that this is fixed in trunk. I will attempt to bisect this, although there's a high chance of failure here as there's a lot of luck involved.
Comment 9 Michael Catanzaro 2023-03-28 15:51:42 PDT
(In reply to Adrian Perez from comment #6)
> Michael, what size exactly is the window when you see glitches?

I dunno. MiniBrowser does not remember its window size. I've just been resizing it each time I try to reproduce this bug, so the size will be different each time.
Comment 10 Adrian Perez 2023-03-29 01:44:36 PDT
(In reply to Michael Catanzaro from comment #9)
> (In reply to Adrian Perez from comment #6)
> > Michael, what size exactly is the window when you see glitches?
> 
> I dunno. MiniBrowser does not remember its window size. I've just been
> resizing it each time I try to reproduce this bug, so the size will be
> different each time.

Okay, so we are doing basically the same, and from your screenshot it
doesn't look like you have a HiDPI screen either. Maybe it's one of
these issues that, by chance, one person has more ease reproducing than
others—too bad we don't have a good reproducer :\
Comment 11 Michael Catanzaro 2023-03-30 09:20:34 PDT
Created attachment 465681 [details]
Yet another screenshot

Oooh, I found another affected website:

https://www.riverfronttimes.com/news/st-louis-to-extend-controversial-parking-management-contract-39741060

(still intend to attempt to bisect this)
Comment 12 Michael Catanzaro 2023-03-30 14:36:58 PDT
(In reply to Michael Catanzaro from comment #11)
> (still intend to attempt to bisect this)

I'm not going to succeed. It's too hard to reproduce.
Comment 13 Michael Catanzaro 2023-04-03 09:49:04 PDT
OK, I hit this today on reddit.com using 2.41.1, so I was wrong: it's not a stable branch bug after all. I still have not been able to reproduce on cnn.com but no question it's the same bug.
Comment 14 Michael Catanzaro 2023-04-15 10:40:30 PDT
Something changed and now this bug is happening drastically more often with 2.41.2 on reddit.com. It's so severe that I'll probably downgrade the GNOME runtime back to 2.41.1, which was also affected but much less often.
Comment 15 Michael Catanzaro 2023-04-15 10:41:29 PDT
On the other hand, I suspect I will be more productive if I cannot use reddit. Hmmm. ;)
Comment 16 Michael Catanzaro 2023-04-16 10:37:07 PDT
src.fedoraproject.org is now broken as well.
Comment 17 Michael Catanzaro 2023-04-16 14:43:49 PDT
OK so strangely enough, the rendering errors on reddit.com are occurring more commonly due to bug #255488. I know it seems weird but they are definitely related. To reduce bugzilla confusion, let's handle the 2.41.2 regressions in bug #255488 and keep this bug focused on the original regression from 2.39.90.
Comment 18 Michael Catanzaro 2023-04-17 09:58:24 PDT
OK, with bug #255488 resolved as something different: I haven't seen the original problem here with cnn.com since upgrading to 2.41.2. I think it's probably fixed. Shame we don't know when it got fixed to do a backport for 2.40.