store page crashing browser

[impetus_maximus]

Pale Moon 29.1.0 (32-bit) in private mode (if that matters)
Win 7 64bit

simple to reproduce.just go here to CRASH→ https://paradisearcadeshop.com/collecti ... tton-green
checking the page with validator.w3.org shows a TON of errors, so the page is obviously poorly coded.

don't know if this counts as a bug, but perhaps it may help prevent future crashes like this?

[coffeebreak]

The crash seems to come from this script:

Code: Select all

https://files-shpf.mageworx.com/productoptions/source.js?mw_productoptions_t=1615776244&shop=paradise-arcade.myshopify.com

When the script is blocked (or else its domain mageworx.com is blocked), the page doesn't crash.

(Be aware, though, that mageworx.com seems to have a commerce function - I don't know if blocking the script will make some other problem for using the site.)

[gepus]

Can't reproduce the crash, neither with a fresh profile nor with the one I'm working with.
Pale Moon 29.1.0 64-bit/Win7 x64

[jobbautista9]

This doesn't seem to be a bug specific to Windows.

I can confirm what gepus said. It doesn't lead to OOM (which is the real issue here) in a fresh profile. However, having Pure URL installed will lead to OOM. So this might be an extensions issue.

[moonbat]

having Pure URL

Deactivate Pure URL by clicking its toolbar button and see.

[New Tobin Paradigm]

When are you people going to learn the difference between a hung UI thread and a program error crash?

Seems to be in a cycle of css js and cc likely due to that script causing constant reevaluation of dom and css shit locking us into a massive loop of watching responding reevaluating paint cyclecollect watching responding reevaluating paint cyclecollect.

It just keeps going on resorting in a hung ui thread constantly churning.

[jobbautista9]

Deactivate Pure URL by clicking its toolbar button and see.

I deactivated Pure URL as you said, and the browser still OOMs. So it's not a problem with Pure URL then.

I created a new profile again, and interestingly, it OOMs as well, even without any add-ons installed. I might have been lucky in the first try.

[moonbat]

Maybe what Tobin said, some unending loop that goes on allocating memory.

[New Tobin Paradigm]

I suppose on 32bit it would end in an OOM but OOM is not crash because runtime error.

[jobbautista9]

I sent an email to their customer support team. They responded and said that they "will definitely consider and try to fix this issue in the near future."

[impetus_maximus]

wow, thanks everyone for your efforts and input in the matter.

I sent an email to their customer support team. They responded and said that they "will definitely consider and try to fix this issue in the near future."

cheers for that. i emailed them as well along with a link to validator results for that page. they said it doesn't crash their browsers (chrome, firefox, opera)
seems pretty common practice that if a page 'works' for x browser the page creator doesn't worry about compliance.

P.S. the reason for those wondering... i'm using the 32bit version to try and limit memory used. now that i have 24GB (2x4, 2x8) i may switch.

again, thanks for everyone's efforts.
cheers

[vannilla]

it doesn't crash their browsers (chrome, firefox, opera)

Actually, I wonder if that's really true.
The multiprocess architecture, especially on Chrome, is designed to handle a crashing tab in a way to hide the event.
Essentially, at a high level (meaning the actual implementation might be more clever) it does something like this: a tab crashes, be it the actual process like in Chrome or the content process in Firefox; the "currently visible content" is saved like a snapshot and is shown to the user while the process is unavailable; the tab is restarted and user interaction can resume like normal.
Effectively, if there is an OOM issue, the tab would crash even in Chrome, but because the crash is hidden behind a piece of paper colored to look like the visited webpage, nobody realizes there is an actual crash.
So yes, the browser's user interface does not crash, but that doesn't mean the website also doesn't. Depending on the point of view, a crashing subprocess (= tab) can still be considered "crashing the browser".

[New Tobin Paradigm]

Chrome browsers (and Quantum) seem optimized for constant churning of dom and css creations, observations, mutations, and destructions but it does come at a very high cost.. They loose the ability to really do much outside the viewport and until the page is fully loaded and shit is done anything outside the viewport tends not to render or even exist.. Even then it can't actually handle old style flat html with all the shit per tag like say a very long standard content.

We of course can handle it but then we flop over sometimes at the newer shit.

Let me illustrate for you what I mean:

http://xref.palemoon.org/moonchild-central/source/platform/db/sqlite3/src/sqlite3.c

Please do not abuse this link or I will have to block it at the webserver but it is the PERFECT example.

This file WILL lock up the UI thread for a while and a couple of times BUT it will load in the end and you will be able to go up and down and all around from line 1 to line 230518 once it is fully loaded. 32bit MAY OOM.. I haven't bothered to try..

Now try it in a Chrome Browser.. it will chunk by viewport and never really seem to load the entire thing.. Getting to the bottom is almost impossible without a line number anchor and even then I don't think it will work properly. As it chunks the scroll bar chunks too.. IF you get to the bottom only THEN will you be able to scroll back and forth and THAT is whole different joy.

You be the judge of which is superior.

[Tomaso]

Here's another example, which I think demonstrates the same thing (sorry if unrelated)..

At first try, Chromium always fails to resize the article boxes at https://sol.no/, like it's supposed to.
The page needs to be refreshed, at least one time, before everything is shown correctly :



In comparison, Pale Moon loads it correctly, right away! :

[New Tobin Paradigm]

Leave it to a Gorn to miss the point I was trying to make.

Why do I bother.

[The Squash]

they "will definitely consider and try to fix this issue in the near future."

Don't smile yet. My experience shows that those messages are usually sent out by bots -- even totally legitimate companies usually employ bots for the task. Think of the classic scammer's ploy:

The [gold / bamboo / check] is [coming from the Sinai / traveling across the Silk Road / in the mail].

[Moonchild]

they "will definitely consider and try to fix this issue in the near future."

Don't smile yet. My experience shows that those messages are usually sent out by bots -- even totally legitimate companies usually employ bots for the task. Think of the classic scammer's ploy:

The [gold / bamboo / check] is [coming from the Sinai / traveling across the Silk Road / in the mail].

Pure speculation, unfounded, and unlikely to be in a personal response.

[The Squash]

32bit MAY OOM.. I haven't bothered to try..

I can verify that this does not OOM on 32-bit.

[Kris_88]

https://paradisearcadeshop.com/collecti ... tton-green
We can reproduce the problem that occurs on this page.

Code: Select all

<html>
<body>

<div id="test">test</div>

<script>

  var mutationObserver = new MutationObserver(function (mutations) {
      alert('Mutation event.');
      document.getElementById("test").style.display = "none";
  });

  mutationObserver.observe(document.getElementById("test"), {attributes:true});

  document.getElementById("test").style.display = "none";

</script>

</body>
</html>

The mutation observer is triggered when the attribute value has not changed.
Therefore, this script produces endless loop.
I think this is a problem in the Pale Moon.
I think this problem is causing it to freeze on a lot of sites.
You can check this example in other browsers.

[New Tobin Paradigm]

Could be. Thanks for the reduced test case.. It helps a FUCK ton actually. See your drive-by tinkering can be useful for more than extensions

As an aside, it would be more proper to call it a problem with UXP's JS Engine as it has nothing to specifically do with Pale Moon.