My memort filling up bug

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

Post by g***@aol.com
sites. Is there a known bug where scripts do not release the memory
when they end? It seems to be orphan files since the processes and
services are not getting that much bigger.
I noticed one of those click bait quizzes seemed to eat a gig or so I
never got back. It was full of pop up Google ads that the pop up
blocker missed.
W/7 64 bit pro & firefox. (But Slimjet isn't much better)

Is the memory still showing as loaded if you close the browser?

I know certain (I presume script-sodden) sites cause the CPU and memory
usage to skyrocket, even if they're just sitting there; my first
knowledge is usually the fan making more noise. But those both go down
if I close that tab, though may take a few seconds before they do (I
don't even have to close the browser). [The main offenders for me at the
moment are the "nextdoor" site (even if I'm not doing anything on it),
or viewing (especially thumbnails) on YouTube, though that latter is of
course understandable.]

(The old DOS mode) Xtree Gold also rails a core, more or less, even if
not apparently doing anything (I use it occasionally, mainly to see
what's really happening, or its hex viewer/editor), but again that stops
when I exit from it. (Puzzling, though!)

7-32 HP & Chrome.

--
J. P. Gilliver. UMRA: 1960/<1985 MB++G()AL-IS-Ch++(p)***@T+H+Sh0!:`)DNAf

... "Peter and out." ... "Kevin and out." (Link episode)

Paul

2024-04-10 12:32:15 UTC

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

He reported a computer with 16GB of RAM, filled to 13GB by program activity.
An abnormal kind of bloating.

Browsers have adopted various strategies, such as "tab sleeping",
as a solution. I find that runaway tabs now seem to stop at 3GB each.
They don't seem to go too much higher than that. I have not
attempted to study a sleeping tab implementation, to see how
much better it is.

Early on, when we referred to this as "leakage", I actually
had a computer crashed by this. A tab consumed all memory,
the OS was pinched until it crashed. Later, it was
the bloating you would notice, the wasted RAM, but not
so much wasted as to crash the computer.

You will find though, that the browser developers are "silent"
on the topic of why this happens. Which I find... weird.
It makes a browser company look bad, to run a computer
out of memory, and their reputation is at stake. You
would think they would say "the root cause is..."
so that the blame would fall on the shoulders of
the scumbags doing it.

Paul

J. P. Gilliver

2024-04-10 12:58:00 UTC

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

He reported a computer with 16GB of RAM, filled to 13GB by program activity.
An abnormal kind of bloating.

Indeed! (I can only have 4G with -32.)

Post by Paul
Browsers have adopted various strategies, such as "tab sleeping",
as a solution. I find that runaway tabs now seem to stop at 3GB each.
They don't seem to go too much higher than that. I have not
attempted to study a sleeping tab implementation, to see how
much better it is.

I've experimented with one - "The original great suspender" I think it's
called. I don't _think_ I've disabled it, but it doesn't seem to be
operating at the moment: I think it just stops working after a while
until prodded. When it is, it seems to work well - it's just difficult
what timeout to choose! 5 minutes is definitely too short; an hour
usually too long. I think I had it set to 15 or 30 minutes last time I
had it working.

Post by Paul
Early on, when we referred to this as "leakage", I actually
had a computer crashed by this. A tab consumed all memory,
the OS was pinched until it crashed. Later, it was
the bloating you would notice, the wasted RAM, but not
so much wasted as to crash the computer.

A lot of the time, when I look, my system is running with only a tiny
amount of free memory - but not zero; that's of course fine. I only
notice a real slowdown when something (more often than not a browser
tab) goes rogue and its usage just keeps climbing, but that doesn't
actually happen very often.

Post by Paul
You will find though, that the browser developers are "silent"
on the topic of why this happens. Which I find... weird.
It makes a browser company look bad, to run a computer
out of memory, and their reputation is at stake. You
would think they would say "the root cause is..."
so that the blame would fall on the shoulders of
the scumbags doing it.

They're not all as independent as they appear - certainly, the big
culprits, the browser developers don't want to "offend" (not least, of
course, as they're often part of the same company - Google/Chrome won't
complain about YouTube, for instance, but even when not part of the
same).

Post by Paul
Paul

John

--
J. P. Gilliver. UMRA: 1960/<1985 MB++G()AL-IS-Ch++(p)***@T+H+Sh0!:`)DNAf

"I'm a paranoid agnostic. I doubt the existence of God, but I'm sure there is
some force, somewhere, working against me." - Marc Maron

g***@aol.com

2024-04-11 00:00:48 UTC

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

He reported a computer with 16GB of RAM, filled to 13GB by program activity.
An abnormal kind of bloating.
Browsers have adopted various strategies, such as "tab sleeping",
as a solution. I find that runaway tabs now seem to stop at 3GB each.
They don't seem to go too much higher than that. I have not
attempted to study a sleeping tab implementation, to see how
much better it is.
Early on, when we referred to this as "leakage", I actually
had a computer crashed by this. A tab consumed all memory,
the OS was pinched until it crashed. Later, it was
the bloating you would notice, the wasted RAM, but not
so much wasted as to crash the computer.
You will find though, that the browser developers are "silent"
on the topic of why this happens. Which I find... weird.
It makes a browser company look bad, to run a computer
out of memory, and their reputation is at stake. You
would think they would say "the root cause is..."
so that the blame would fall on the shoulders of
the scumbags doing it.
Paul

Actually 8g but it boots at 1.6g that while I am watching the ram
graphic (on "performance" task mgr) in the corner of the screen it
just seems to slowly fill up when there is nothing in applications and
no real hogs on processes.
I have a screen shot of a fresh boot of resource manager and I am
waiting for it to fill up again to see if anything really changes
there. It's not like I am looking for 100 meg or something, this thing
gets to the top of the box and windows is just thrashing the page file
to keep from drowning. Doing anything, even one click, takes a minute
or more so choose wisely ;-)

I do know it is caused by script laden web sites and I never had my
other 4g machine do that. It would have been unusable. I am wondering
if it was one of those "Updates" I got from that back door guy we
talked about here. I updated this machine but I didn't do the old one.
They are both dual core intel machines running w/7-64 pro, Maybe even
the same disk ;-)

Paul

2024-04-11 02:07:52 UTC

Post by g***@aol.com

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

He reported a computer with 16GB of RAM, filled to 13GB by program activity.
An abnormal kind of bloating.
Browsers have adopted various strategies, such as "tab sleeping",
as a solution. I find that runaway tabs now seem to stop at 3GB each.
They don't seem to go too much higher than that. I have not
attempted to study a sleeping tab implementation, to see how
much better it is.
Early on, when we referred to this as "leakage", I actually
had a computer crashed by this. A tab consumed all memory,
the OS was pinched until it crashed. Later, it was
the bloating you would notice, the wasted RAM, but not
so much wasted as to crash the computer.
You will find though, that the browser developers are "silent"
on the topic of why this happens. Which I find... weird.
It makes a browser company look bad, to run a computer
out of memory, and their reputation is at stake. You
would think they would say "the root cause is..."
so that the blame would fall on the shoulders of
the scumbags doing it.
Paul

Actually 8g but it boots at 1.6g that while I am watching the ram
graphic (on "performance" task mgr) in the corner of the screen it
just seems to slowly fill up when there is nothing in applications and
no real hogs on processes.
I have a screen shot of a fresh boot of resource manager and I am
waiting for it to fill up again to see if anything really changes
there. It's not like I am looking for 100 meg or something, this thing
gets to the top of the box and windows is just thrashing the page file
to keep from drowning. Doing anything, even one click, takes a minute
or more so choose wisely ;-)
I do know it is caused by script laden web sites and I never had my
other 4g machine do that. It would have been unusable. I am wondering
if it was one of those "Updates" I got from that back door guy we
talked about here. I updated this machine but I didn't do the old one.
They are both dual core intel machines running w/7-64 pro, Maybe even
the same disk ;-)

You can run a 32-bit executable on a 64-bit OS.
The address space available to 32-bit executables
forms a kind of implicit "quota" and prevents
zooming off to infinity.

I've actually installed 32-bit software, for that
specific reason.

There is WIN32 and WIN64 here, for example.
The WIN32 can run on a 32 bit OS or a 64 bit OS.
When run on a 64 bit OS, it caps runaway behavior
(so even it you had tab sleeping, there would be
the additional protection of not running off to
infinity in any case).

http://releases.mozilla.org/pub/firefox/releases/115.9.1esr/

Otherwise, drivers and paged or unpaged pool leaks,
can lead to continuously increasing leakage.

Paul

g***@aol.com

2024-04-12 05:33:26 UTC

Post by g***@aol.com

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded

"This thing" - this computer? This browser?

He reported a computer with 16GB of RAM, filled to 13GB by program activity.
An abnormal kind of bloating.
Browsers have adopted various strategies, such as "tab sleeping",
as a solution. I find that runaway tabs now seem to stop at 3GB each.
They don't seem to go too much higher than that. I have not
attempted to study a sleeping tab implementation, to see how
much better it is.
Early on, when we referred to this as "leakage", I actually
had a computer crashed by this. A tab consumed all memory,
the OS was pinched until it crashed. Later, it was
the bloating you would notice, the wasted RAM, but not
so much wasted as to crash the computer.
You will find though, that the browser developers are "silent"
on the topic of why this happens. Which I find... weird.
It makes a browser company look bad, to run a computer
out of memory, and their reputation is at stake. You
would think they would say "the root cause is..."
so that the blame would fall on the shoulders of
the scumbags doing it.
Paul

Actually 8g but it boots at 1.6g that while I am watching the ram
graphic (on "performance" task mgr) in the corner of the screen it
just seems to slowly fill up when there is nothing in applications and
no real hogs on processes.
I have a screen shot of a fresh boot of resource manager and I am
waiting for it to fill up again to see if anything really changes
there. It's not like I am looking for 100 meg or something, this thing
gets to the top of the box and windows is just thrashing the page file
to keep from drowning. Doing anything, even one click, takes a minute
or more so choose wisely ;-)
I do know it is caused by script laden web sites and I never had my
other 4g machine do that. It would have been unusable. I am wondering
if it was one of those "Updates" I got from that back door guy we
talked about here. I updated this machine but I didn't do the old one.
They are both dual core intel machines running w/7-64 pro, Maybe even
the same disk ;-)

You can run a 32-bit executable on a 64-bit OS.
The address space available to 32-bit executables
forms a kind of implicit "quota" and prevents
zooming off to infinity.
I've actually installed 32-bit software, for that
specific reason.
There is WIN32 and WIN64 here, for example.
The WIN32 can run on a 32 bit OS or a 64 bit OS.
When run on a 64 bit OS, it caps runaway behavior
(so even it you had tab sleeping, there would be
the additional protection of not running off to
infinity in any case).
http://releases.mozilla.org/pub/firefox/releases/115.9.1esr/
Otherwise, drivers and paged or unpaged pool leaks,
can lead to continuously increasing leakage.
Paul

This is Firefox 115.9.1 ESR (64 bit)
The SlimJet I run is
Version 38.0.9.0 (based on Chromium 109.0.5414.74) (Official Build)
(64-bit)

Paul

2024-04-12 05:40:10 UTC

Post by g***@aol.com
This is Firefox 115.9.1 ESR (64 bit)
The SlimJet I run is
Version 38.0.9.0 (based on Chromium 109.0.5414.74) (Official Build)
(64-bit)

Neither of those is capped on RAM usage then.

Firefox uses multiple processes, and any process
could go nuts.

If we had a definitive statement from the browser
manufacturers, exactly what the bloat problem is,
we could work on it from our end. But without
a hint, where would we look. Addons ?

Paul

Tim Slattery

2024-04-12 14:17:55 UTC

Post by Paul
You can run a 32-bit executable on a 64-bit OS.
The address space available to 32-bit executables
forms a kind of implicit "quota" and prevents
zooming off to infinity.

Umm??
32-bit programs in Windows (either 32-bit Windows or 64-bit Windows)
get a 4GB virtual memory space to run in. Why 4GB? Because that's how
many bytes you can address with 32 bits. It has nothing to do with
quotas or runaways.

64-bit executables currently get a 128TB virtual memory space. 64 bits
will address far, far, more than 128TB. But current hardware
constraints must be taken into consideration.

Still, the huge VM space for 64-bit programs makes it much easier for
them to handle huge data sets - video, or large pictures or whatever
you need.

--
Tim Slattery
timslattery <at> utexas <dot> edu

Paul

2024-04-13 05:45:43 UTC

Post by Tim Slattery

Post by Paul
You can run a 32-bit executable on a 64-bit OS.
The address space available to 32-bit executables
forms a kind of implicit "quota" and prevents
zooming off to infinity.

Umm??
32-bit programs in Windows (either 32-bit Windows or 64-bit Windows)
get a 4GB virtual memory space to run in. Why 4GB? Because that's how
many bytes you can address with 32 bits. It has nothing to do with
quotas or runaways.
64-bit executables currently get a 128TB virtual memory space. 64 bits
will address far, far, more than 128TB. But current hardware
constraints must be taken into consideration.
Still, the huge VM space for 64-bit programs makes it much easier for
them to handle huge data sets - video, or large pictures or whatever
you need.

Not quite correct, Obiwan.

The 4GB address space, is split into userland and kernel space.
By default the split is 2GB:2GB, and when certain things are
taken into account, a 32 bit process "cannot use more than
roughly 1.8GB of RAM". That's a rough rule of thumb.
That's the actual address space limit.
When you run Photoshop x86 (a large application), then it
might report usage up to 1.8GB.

The kernel address space is needed so you can call into the
kernel for your various low level things.

If you alter the split (I've done this on WinXP, long enough
to finish a build of Firefox), you can do the split
as 3GB:1GB. But if you do that, there are some side effects
that hint maybe leaving things alone full-time was best.
You should not run a daily driver with WinXP 3GB:1GB .

https://learn.microsoft.com/en-us/windows/win32/memory/4-gigabyte-tuning

"The /3GB switch makes a full 3 GB of virtual address space
available to applications and reduces the amount available
to the system to 1 GB.' [The "system" being kernel space]

The total address space is 4GB, but it's split.

Programs can be Large Address Aware. This allows a program
to actually use 3GB when you set the split to 3GB:1GB.
This might have had something to do with editbin.exe
(dumpbin.exe to verify).

editbin /LARGEADDRESSAWARE:NO program.exe # Turn it off
editbin /LARGEADDRESSAWARE program.exe # Turn it on

I'm having some trouble with my test programs that show
the effects. I don't know if this is a side effect of being on
W11 or what. So I redid my editbins, and dumpbins, and
had a go at it. First the canonical article, now a bit uglified
by web monkeys. I ran eight test cases for actual test, and
copied in the six that were important (met the conditions
mentioned in the table). The left column tests were done
in a 32-bit VM, the right column tests were done native in Terminal.

https://learn.microsoft.com/en-us/windows/win32/memory/memory-limits-for-windows-releases

Limit on x86 Limit on x64

32 bit process 2GB 2GB
3GB with LAA 4GB with LAA

64 bit process --- 2GB
License-limit etc with LAA

******************************** Actual test **************************************

C:\Downloads>malloc-no D:\MALLOC\NOT>malloc
01916 megabytes t=001.190557 01898 megabytes t=000.308925

bcdedit /set IncreaseUserVA 3072
C:\Downloads>malloc-yes D:\MALLOC>malloc
02879 megabytes t=001.889975 03821 megabytes t=000.613203

D:\MALLOC\NOT>malloc64
01901 megabytes t=000.317975

D:\MALLOC>malloc64
30632 megabytes t=005.741883

The last one stopped at 30GB, only because other RAM in the computer
is occupied right now, but not doing computing. The RAM license on this
machine does not interfere. I had to buy Windows 7 Pro on the other machine,
to get enough memory license to use all the RAM.

If there wasn't a RAM license, the upper limit on 64-bit is controlled
by CPU arch. One of the CPUs was 2^43, a slightly later one (perhaps
the thing I'm typing on) is 2^48. There is a hardware trick for the
other bits. On AMD, even a desktop CPU, has the same address bit width
as the server CPUs. Intel likely does things a bit differently, but
this is not visible in normal use anyway. It's just an academic note.
For example, an AMD machine could stuff 6TB in it today, a server,
but I doubt that's clunking 2^48 over the head. I'm not even going
to calc. But when machines have limits that matter to Intel or AMD
business plan, you may find the address decoder spoils your fun.
Even if 6TB would fit in my desktop consumer machine, it would be
neatly snipped off and unusable, above 128GB.

So actually, while you can show me a 64-bit register value for an
address on your screen, if you dump enough of them, you might notice
something peculiar about the top bits. The top bits are used for
some other purpose I would have to look up. Irrelevant to use,
but important to "someone". The hardware, the high speed comm links
between cores (coherent cache and so on), the address field on the
links might show six bytes of address bits, not eight. Since carrying
the extra two bytes, slows down the link a bit, it's in your best
interest to "only carry what the arch predicts is needed".

*******

Getting back to my original comment, 01898 megabytes is
what I was referring to as an implicit quota. I would
have to check specific executables to see if editbin has
been used on them.

In the past, using LAA was a waste of time, since
no user would be setting the /3GB switch or using
bcdedit like that for a later OS. But using dumpbin
you can check.

H:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\bin\Hostx64\x64>

dumpbin /headers D:\MALLOC\malloc.exe

Microsoft (R) COFF/PE Dumper Version 14.30.30706.0
Copyright (C) Microsoft Corporation. All rights reserved.

Dump of file D:\MALLOC\malloc.exe

PE signature found

File Type: EXECUTABLE IMAGE

FILE HEADER VALUES
14C machine (x86)
F number of sections
53D49DBF time date stamp Sun Jul 27 02:35:43 2014
10800 file pointer to symbol table
45B number of symbols
E0 size of optional header
127 characteristics
Relocations stripped
Executable
Line numbers stripped
Application can handle large (>2GB) addresses <=== LAA is Yes
32 bit word machine

Paul

2024-04-10 18:25:07 UTC

Post by g***@aol.com
This thing is still loading up the memory after going to script loaded
sites. Is there a known bug where scripts do not release the memory
when they end? It seems to be orphan files since the processes and
services are not getting that much bigger.
I noticed one of those click bait quizzes seemed to eat a gig or so I
never got back. It was full of pop up Google ads that the pop up
blocker missed.
W/7 64 bit pro & firefox. (But Slimjet isn't much better)

Firefox unfortunately, have a browser memory leak in its core. Much worse
than Chrome. It's not possible to keep a clean Firefox installation (i.e.
without any browser extension) running all the time and used for days to
weeks or more (depending on usage), without noticing its gradually
increasing memory consumption even when there's only one blank browser tab
left after each browser usage session.

Bad browser extension design can make things worse. It is the most common
cause of web browser memory leak or worsening the leak, regardless of which
web browser is used. It assumes that the computer has infite memory, and
doesn't care about memory usage efficiency. Worse thing is that it has
become a trend in nowaday softwares (including websites).

g***@aol.com

2024-04-11 00:15:22 UTC

Post by JJ

Firefox unfortunately, have a browser memory leak in its core. Much worse
than Chrome. It's not possible to keep a clean Firefox installation (i.e.
without any browser extension) running all the time and used for days to
weeks or more (depending on usage), without noticing its gradually
increasing memory consumption even when there's only one blank browser tab
left after each browser usage session.
Bad browser extension design can make things worse. It is the most common
cause of web browser memory leak or worsening the leak, regardless of which
web browser is used. It assumes that the computer has infite memory, and
doesn't care about memory usage efficiency. Worse thing is that it has
become a trend in nowaday softwares (including websites).

I always close FF when I am not using it. That flushes the cookies
from most sites, anything not white listed. I very seldom use FF at
the same time on different sites and never on incompatible sites.
(like anything "my money" and anything spam/malware driven like
Facebook or "back"door.)
I am not sure how much leakage there may be between tabs.

Anyway ... when this thing gets full I start looking around
When I look at processes , there is no Firefox running. some svchosts
are pretty big but not eating 5 or 6 gig. I really want to look at
resource manager and compare it to my clean boot screenshot.
I still don't expect to see anything. I think this is just memory that
was taken and never given back. It's not really allocated to anything
Windows reports but it is still marked unavailable.
I ran a pass of RAM test early on OK but I may crank it up and go to
bed. Let it churn all night and see if anything hardware pops up. That
would be too easy, I think it is software ;-)

Paul

2024-04-11 02:15:23 UTC

Post by g***@aol.com

Post by JJ

Firefox unfortunately, have a browser memory leak in its core. Much worse
than Chrome. It's not possible to keep a clean Firefox installation (i.e.
without any browser extension) running all the time and used for days to
weeks or more (depending on usage), without noticing its gradually
increasing memory consumption even when there's only one blank browser tab
left after each browser usage session.
Bad browser extension design can make things worse. It is the most common
cause of web browser memory leak or worsening the leak, regardless of which
web browser is used. It assumes that the computer has infite memory, and
doesn't care about memory usage efficiency. Worse thing is that it has
become a trend in nowaday softwares (including websites).

I always close FF when I am not using it. That flushes the cookies
from most sites, anything not white listed. I very seldom use FF at
the same time on different sites and never on incompatible sites.
(like anything "my money" and anything spam/malware driven like
Facebook or "back"door.)
I am not sure how much leakage there may be between tabs.
Anyway ... when this thing gets full I start looking around
When I look at processes , there is no Firefox running. some svchosts
are pretty big but not eating 5 or 6 gig. I really want to look at
resource manager and compare it to my clean boot screenshot.
I still don't expect to see anything. I think this is just memory that
was taken and never given back. It's not really allocated to anything
Windows reports but it is still marked unavailable.
I ran a pass of RAM test early on OK but I may crank it up and go to
bed. Let it churn all night and see if anything hardware pops up. That
would be too easy, I think it is software ;-)

This behavior became more evident, once compositing was introduced
for window rendering. The way a browser pixels get to the screen
today, is not the same as how it was done in Firefox 2.

The compositor runs at 60FPS, and even when a page is "content-frozen"
and not changing the screen, the compositor is *still* running.
Then, when the compositing process loses touch with the Windows
display subsystem... bad things happen, such as 3GB of memory leaks.
I lucked out one day, went to a Firefox window, the "content" on the
screen was stale, which means the path from Firefox to the screen
was broken, and this was at the same time as Firefox had consumed 3GB
of memory. Presumably there were 3GB of queued pixmaps headed for
the display. My conclusion is "they don't use double-buffering"
which would reduce the queue to nothingness. Earlier display
implementations used double buffering, which has a quite-small
memory foot print. Absolutely essential when your Matrox
card only has 4MB of VRAM.

Paul

Newyana2

2024-04-11 12:36:36 UTC