Image-heavy page causes browser crash.

BenFenner · Post by **BenFenner** » 2025-02-20, 20:37

Operating system: Windows 7 (64-bit)
Browser version: Pale Moon 33.6.0 (64-bit)
Problem URL: See below for crashing URL
Installed add-ons:
deCDN 33.3
Pure URL 3.4
uBlock Origin 1.16.4.30
YouTubeNonStop 1.0.10

Installed plugins: (about:plugins):
application/x-silverlight npctrl scr
application/x-silverlight-2

(Forgot I had Silverlight from the Netflix days, would love to remove but not obvious how.)

about:support

Warning: Clicking the link below will likely crash Pale Moon if logged into the sr20-forum.
URL causing crash: https://www.sr20-forum.com/the-moshpit/ ... s-nws.html
Warning: Clicking the link above will likely crash Pale Moon if logged into the sr20-forum.

The link above is to an off-topic thread in an automotive forum with lots of images embedded from all sorts of sources. One of those sources is causing a browser crash, or maybe just the sheer amount of data is causing some issue. The crash is 100% repeatable on my end.
But as I type this I am realizing the page is off topic on the forum, so it is only viewable by forum members. *sigh*

This is a forum I administer, so let me create a temporary dummy account for people to use for testing. Hang tight.

Okay, anyone who wants to test can use these login details. I will perma-delete the account when testing is complete.
(Please don't use the account to make posts to the forum, send PMs, etc. That will just be more junk for me to hunt down and delete later.)

Username: Pale Moon Test
Password: The father died during childbirth.

Post by **moonbat** » 2025-02-20, 23:24

Logged in on my testing profile (an empty default settings one) and opened the page, no crash. I'm easily able to scroll the first page. The post goes on for 356 pages; are you using a script or something to expand them all into one page? I'm running KDE Neon, not Windows.

BenFenner · Post by **BenFenner** » 2025-02-20, 23:34

Ahh, I forgot my account has 4x higher posts-per-page settings (I see 89 pages of content). Let me apply those to this account. Hang tight.

Edit: Updated. Would appreciate another check. Thank you.

I may also set my own account to 1 post per page so I can narrow down if it is a specific post that causes the issue, then can find the content to really narrow this down (assuming it's not some type of OOM issue).

Post by **Moonchild** » 2025-02-21, 00:20

It would help if you could actually post appcrash info so we can see if it's OOM, if it's a crash in our code, in the (dated) GPU driver, or something else.

BenFenner · Post by **BenFenner** » 2025-02-21, 01:05

I was hoping someone else could reproduce and get the appcrash data. That's why I went through the trouble of making an account for others to test with. But if nothing happens tonight I will look up how to get appcrash details again (I always forget) and provide them tomorrow when I have availability to do so. I'd also have tested on more/newer machines here but I have way too much work (read: open tabs) to get through before I can risk the app crashing on me there. Another thing I will get to tomorrow.

BenFenner · Post by **BenFenner** » 2025-02-21, 01:28

Moonchild wrote: ↑
2025-02-21, 00:20
in the (dated) GPU driver

In that regard Waterfox on the same machine doesn't suffer from the same issue. That might be a clue going forward.
Again, I'll post actual new info tomorrow.

adoxa · Post by **adoxa** » 2025-02-21, 02:27

I viewed the first page (of 89) without incident (64-bit Basilisk, Win10, normal profile).

Post by **Moonchild** » 2025-02-21, 09:56

BenFenner wrote: ↑
2025-02-21, 01:28
In that regard Waterfox on the same machine doesn't suffer from the same issue. That might be a clue going forward.

That doesn't really matter. We're not Waterfox and not using the same layout and rendering issue. Pale Moon can tickle the driver "wrong", exposing the issue in it, while Waterfox does not.

Lucio Chiappetti · Post by **Lucio Chiappetti** » 2025-02-21, 12:08

I could login and navigate randomly among threads and posts without problem (latest Pale Moon on Linux, I have uBO but was not triggered at all on the site)

BenFenner · Post by **BenFenner** » 2025-02-21, 14:30

Well, I can't reproduce the issue anymore on the same machine. I'm on PM 33.6.0.1 now but don't think that has anything to do with it. (It was probably the restart I just did.)
I'd had Pale Moon open for 3-4 days with a 3.5-hour YouTube video loaded at various stages of watch time. This can cause Pale Moon to run very slowly from what I've seen in the past. The interesting part was that the same crash happened over and over again, even after Pale Moon had the chance to restart (although not gracefully I guess).
I would wager after another few days and more very-long YouTube videos and such I could get the browser to crash again, potentially using any web site, not just the one linked above.

So... Never mind. =/

I'll mark this solved. (I have deleted the test account mentioned in the first post.)

If I get into this (repeatable but rare) situation again in the near future I'll try to get an appcrash log and if I do then I'll bump this thread with that.

Lucio Chiappetti wrote: ↑
2025-02-21, 12:08
I have uBO but was not triggered at all on the site

That's how we do.

Post by **Moonchild** » 2025-02-21, 15:55

If the problem survived multiple browser restarts, but went away with a system restart, then it, by definition, can't be something internal to the browser.

BenFenner · Post by **BenFenner** » 2025-02-21, 18:33

Sure it can. Browser memory leak is my first guess (unless I guess that's no longer considered internal to the browser since it has leaked

). But we're not going to get to the bottom of it in this thread me thinks. Nothing to see here. Move along.

Post by **Moonchild** » 2025-02-21, 19:27

A browser memory leak would go away with a browser restart or it would also be reproducible after a system restart.
A browser restart will clear any and all application memory allocations, so it would have a clean slate. Since it could be reproduced with browser restarts, that by itself could indeed potentially be a leak in the browser, but combined with it being cleared up with a system restart rules it out, since it would lead to the same memory situation in-browser visiting the same site, and would logically lead to the same crash if it was an internal issue.

So, my best guess at this point is a memory leak in the graphics driver (or system library handling the graphics subsystem) having the GPU run out of VRAM, or potentially leading to (video) memory fragmentation, both of which could be cleared up with a system restart.

Goodydino · Post by **Goodydino** » 2025-02-21, 19:50

All I get is "Login failure".

BenFenner · Post by **BenFenner** » 2025-02-21, 23:32

I still maintain it could be a browser memory leak, which I think you're agreeing with?
I'd wager I can reproduce the issue after a restart, it would just take a week of slowly watching and pausing the same 3+ hour YouTube video and I don't have any of those slated for viewing right now. =/

Goodydino wrote: ↑
2025-02-21, 19:50
All I get is "Login failure".

See:

BenFenner wrote: ↑
2025-02-21, 14:30
I'll mark this solved. (I have deleted the test account mentioned in the first post.)

Post by **Moonchild** » 2025-02-22, 00:50

BenFenner wrote: ↑
2025-02-21, 23:32
I still maintain it could be a browser memory leak, which I think you're agreeing with?

No, I don't agree. Once more, if it was a browser leak then restarting the browser (which completely releases all memory allocated by the process) would either solve it (which it doesn't, since you said it would reoccur after restarts of the browser) or it would be immediately reproducible after a system restart (which it isn't) -- after all, the same interaction with the browser before or after reboot would lead to the same leaks in the browser and should lead to the same issue, regarless of how long the O.S. had been running -- so it is an issue outside of the browser that is simply exposed by the browser being in use in your particular way. O.S.es becoming unstable over time between reboots is not unheard of.

Put another way: if you start the browser, perform X tasks and it crashes, then you'd expect the X tasks to always crash it consistently if it was a memory leak inside the browser; same actions, same result. And that would happen no matter how old or "fresh" your O.S. session is. That's not what you're seeing, and the crashes seem to be related to the age of your O.S. session rather than the tasks performed, so it seems to be something in the O.S. that becomes unstable after your use of Pale Moon, that eventually makes the browser unstable, in turn.

BenFenner · Post by **BenFenner** » 2025-02-22, 04:10

Moonchild wrote: ↑
2025-02-22, 00:50
if you start the browser, perform X tasks and it crashes, then you'd expect the X tasks to always crash it consistently if it was a memory leak inside the browser; same actions, same result.

Yes, I'd wager I could repeatedly cause Pale Moon to crash by performing the same tasks I did earlier in the week. It just takes more time than I care to devote, and I agree it certainly could be a graphics driver issue or similar instead. Of course the OS is fallible too.

Moonchild wrote: ↑
2025-02-22, 00:50
... the crashes seem to be related to the age of your O.S. session rather than the tasks performed, so it seems to be something in the O.S. ...

I think you might have made some assumptions about my computing habits. I track my OS session uptime in months, 6-12 being common with regular/strenuous use. I count my Pale Moon session uptime in weeks, at best. Sometimes dictated by updates, but often dictated by run-away memory usage that slows the browser to the point where I have to close the browser to recover (often when I hit ~2GB of claimed memory with but a single tab open to about:blank -- keep in mind this happens not only on Windows which could be expected but also 64-bit Linux with 32GB RAM available which seems nuts). But it does close gracefully and recover to live another few weeks more until the cycle repeats.

In the crash that happened earlier in the week, PM did not have the opportunity to close gracefully, maybe leaving all of its leaked memory still allocated? Which is maybe a failure in the OS not PM? Which is why the crash kept happening over and over again? I did restore all open tabs and visit all the same sites/pages each time it crashed...
But the age of the OS session was completely unremarkable at this point, just in the middle of a normal 6-12 month stretch. And when I do, I don't reboot because the OS needs it. I do it because I'm going on vacation, or there was a power outage and the UPS didn't span it, or I've installed a new app that requires a reboot, etc.

Anyway, no need to continue this academic discussion if it takes time away from more important things. Onward and upward.

Bilbo47 · Post by **Bilbo47** » 2025-02-22, 19:39

BenFenner wrote: ↑
2025-02-22, 04:10
In the crash earlier, PM did not close gracefully, maybe leaving its leaked memory still allocated?

Nah, a functioning OS would not fail that way. Memory allocated by an app and then lost (either de-allocated or not, yet still non-usable by any app) will not remain inaccessible to the OS after the app is no longer running. Such a fail would be a major mem-leak flaw in the OS, manifesting in any app that fails to perfectly de-allocate, I believe.

BenFenner · Post by **BenFenner** » 2025-03-02, 01:40

I've got the PM crash, re-launch and restore tabs, crash again loop happening again. This time no image-heavy tabs, just a couple paused YouTube videos and some Slashdot pages. The crash happens when the final URL in the list below is opened/focused going in order from first to last.

The exact URLs open (don't judge me):
Window 1: https://www.youtube.com/watch?v=NDIXavk ... whlY3hmxsI (paused)
Window 2: https://www.youtube.com/feed/subscriptions
Window 2: https://www.youtube.com/watch?v=lAIVJxXoc_Q (paused)
Window 2: https://slashdot.org
Window2: https://news.slashdot.org/story/25/03/0 ... a#comments

However I don't think the combination of these URLs open in tabs is really the problem. I think they are just a coincidence. Or maaaaybe the YouTube videos are triggering a video driver bug? I'm happy to update video card drivers or whatever seems relevant.

I have marked this thread as no-longer resolved. I'd love to edit the thread title to be more descriptive, but am unable now.

I have AppCrashView.exe data now!

1st crash:

2nd crash moments later:

Post by **Moonchild** » 2025-03-02, 11:10

I checked the crash address and it looks like it's the same deCDN issue addressed in the deCDN topic, i.e. it is related to "OnDataAvailable" handling, only in this case it's in "imgRequest::OnDataAvailable" (so yes, it does have to do with loading images, and more likely on image-heavy pages).
I don't see an immediate issue with the code there, other than it may be less than optimal threading (mainthread/off-mainthread) so I'd rather wait and see if the patch landed for deCDN solves this issue than jumping right in and start messing with threading code.
Can you see if disabling deCDN solves these issues for you?

Pale Moon forum

Image-heavy page causes browser crash. Topic is solved

Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.

Re: Image-heavy page causes browser crash.