Palemoon crashes when accessing Devuan site

[frostknight]

devuan's anubis just freeze my little poor 8GB RAM i5-7500 Win10 machine in my workplace.
I think it used up all my machine's memory and start swapping to my old slow HDD.

That is strange, because I have an i7-3520M albeit it does have linux on it, but palemoon isn't the least bit slow.

Although I have an solid state drive.

Maybe that is your problem? regular hard drives are much slower. Maybe that is where your problem is.

EDIT: I just tried using it and it crashed my browser. Refer to my next response as for what I think happened.

[frostknight]

Unable to reproduce; I had no issue reaching and displaying the site. The site uses Gitea which we use for our own repo needs as well and works without crashes.


Please try to provide details about the crash.

Actually, I just tried it myself and the problem does in fact occur for me too. Weird stuff...

I think devuan's set up of Anubis might be the problem. Because if it works elsewhere and not on devuan, that means something broke.

[dinosaur]

I can confirm the (instant) crash, under Linux (no crash seen under Windows).

The PC is not the issue (Zen 9950X - 96GB of RAM).

Definitely a PM bug (no such issue seen with other browsers).

[frostknight]

I can confirm the (instant) crash, under Linux (no crash seen under Windows).

The PC is not the issue (Zen 9950X - 96GB of RAM).

Definitely a PM bug (no such issue seen with other browsers).

For me it didn't instantly crash, it took maybe 5-10 seconds then it crashed.

EDIT:

Actually it took like 15+ seconds then it crashed. Very odd.

[Moonchild]

Since I'm obviously not on Linux and this doesn't seem to happen in a virtualized environment, I'm unable to investigate without more info like a stack trace when the crash happens.

[frostknight]

Since I'm obviously not on Linux and this doesn't seem to happen in a virtualized environment, I'm unable to investigate without more info like a stack trace when the crash happens.

Here is something weirder Moonchild, I just tried using that website for the fourth time and I touched not the screen until anubis was done which took like a full 30+ seconds and it worked.

Extremely strange indeed....

EDIT:

Got some info from someone on Devuan forums, I don't completely understand what he meant completely, but I am sure you will:

https://dev1galaxy.org/viewtopic.php?pid=56781#p56781

Something about a javascript problem that firefox ESR has that palemoon probably has too? I don't know why this would exist in both though.

[Kris_88]

I can reproduce the problem on Windows 8.1, PM 33.8.0 32-bit. (PM 33.7.2 also crashes.)

https://git.devuan.org/LeePen/xlibre

Just try to open tools -> developer tools -> browser console while Anubis is calculating its hashes. Sometimes you don't even have time to get to the right menu item, the crash happens while the menu itself is opening.
The problem is probably in a graphic element on the page that is not displayed correctly. Sometimes you need to scroll down the page to see the bottom part and click the mouse on the page, otherwise the crash does not want to happen.

Code: Select all

Problem signature:
  Problem Event Name:	APPCRASH
  Application Name:	palemoon.exe
  Application Version:	6.7.0.9308
  Application Timestamp:	685da96e
  Fault Module Name:	mozjs.dll
  Fault Module Version:	0.0.0.0
  Fault Module Timestamp:	685daa17
  Exception Code:	80000003
  Exception Offset:	002b8039
  OS Version:	6.3.9600.2.0.0.256.48
  Additional Information 1:	5861
  Additional Information 2:	5861822e1919d7c014bbb064c64908b2
  Additional Information 3:	ff84
  Additional Information 4:	ff847098ee7efa49c70e06e469a2896b

Untitled.png

[Kris_88]

In short, yes, the crash is quite reliable. We open the page, scroll down to the end, try to open the browser console, and get a crash (always in the same module and with the same offset).

And yes, I managed to get the same crash here:
https://repo.palemoon.org/MoonchildProductions/UXP
Although, here Anubis uses difficulty 4 and the check is fast. Therefore, it is more difficult to get a crash.
Still, all these captchas and the like are downright crap, sorry...

[Moonchild]

I tried several times to reproduce on the current dev branch of the browser (64-bit build) and was unable to, no matter how quick or slow I scrolled down, opened/closed devtools/checked the console or what not.

If you get a crash on Windows, please capture a minidump that I can analyze.

As an aside, difficulty 5 is high and I really wouldn't advise it unless you are under a very heavy attack from very fast hardware attackers that get through.

[Moonchild]

Something about a javascript problem that firefox ESR has that palemoon probably has too? I don't know why this would exist in both though.

The post doesn't provide anything I can use. it's just vague observations "somewhere between 116 and 120". Mozilla's constant code churn will have thousands of code commits in such a time span.

[Kris_88]

If you get a crash on Windows, please capture a minidump that I can analyze.

I uploaded one to your server.
Thank you.

[Moonchild]

Thanks. Unfortunately it just tells me you ran out of memory, triggering a safe crash in the garbage collector.
Just in case it's something in the GC itself, you try to work around it by setting javascript.options.mem.gc_generational to false (and restarting the browser), but I'm not sure if that will help much.

[Kris_88]

try to work around it by setting javascript.options.mem.gc_generational to false (and restarting the browser), but I'm not sure if that will help much.

Unfortunately, it didn't help.

The computer has 16GB of memory, 12.5GB free.
Pale Moon takes up 185MB according to the task manager, uses the clean default profile.

Code: Select all

Problem signature:
  Problem Event Name:	APPCRASH
  Application Name:	palemoon.exe
  Application Version:	6.7.0.9308
  Application Timestamp:	685da96e
  Fault Module Name:	mozjs.dll
  Fault Module Version:	0.0.0.0
  Fault Module Timestamp:	685daa17
  Exception Code:	80000003
  Exception Offset:	002b8039
  OS Version:	6.3.9600.2.0.0.256.48
  Additional Information 1:	5861
  Additional Information 2:	5861822e1919d7c014bbb064c64908b2
  Additional Information 3:	ff84
  Additional Information 4:	ff847098ee7efa49c70e06e469a2896b

[Kris_88]

Sometimes another crash happens:
Fault Module Name: mozglue.dll
Exception Code: c0000005
Exception Offset: 0000d289

I uploaded another dump to your server.

[Moonchild]

Unfortunately, it didn't help.

I didn't think it would.

The computer has 16GB of memory, 12.5GB free.

Then why are you running a 32-bit version of Pale Moon?

[Moonchild]

Sometimes another crash happens:
Fault Module Name: mozglue.dll
Exception Code: c0000005
Exception Offset: 0000d289

I uploaded another dump to your server.

Yeah just another OOM error, only triggered from the devtools UI thread. It doesn't help.

All of these are deliberate OOM crashes for out-of-memory situations where there is no sane path forward for the JS engine due to memory exhaustion.

[Kris_88]

Then why are you running a 32-bit version of Pale Moon?

We've already discussed this once...
Celeron J1900 does not support AVX.

All of these are deliberate OOM crashes for out-of-memory situations where there is no sane path forward for the JS engine due to memory exhaustion.

How can there not be enough memory if the process does not take up even 200 MB and there is so much memory on the computer that I have never used it all? I have an indicator on the desktop and I have never seen even half of the memory used?

[Moonchild]

Celeron J1900 does not support AVX.

Oh, right, and you don't want to run Nuck-TH's builds.

How can there not be enough memory if the process does not take up even 200 MB and there is so much memory on the computer that I have never used it all?

"Memory" and "Out of memory" doesn't necessarily refer to the heap. If you run out of stack memory then that happens way before you could even notice any significant memory increase in a task manager or what not.
But that too is where 32-bit can cause issues, as it will inherently have a smaller stack space than 64-bit.

In most cases out-of-stack happens with improper recursion.

[Kris_88]

In most cases out-of-stack happens with improper recursion.

If I understand correctly, Java objects are not allocated on the stack. So if this is bad recursion, then the problem is not in the Java code, but in the browser itself, in the runtime.
Hmm, at least it is clear that there is a problem somewhere...

[Moonchild]

it's not java. it's javascript. And javascript by itself can cause stack exhaustion, depending on the calls made, although it's tough to achieve because we do checks for too much recursion etc - however, the asynchronicity of promises makes it difficult. Stack issues occur primarily because of the administration of function calls themselves. It's more of an interop issue than anything else, I think.