Palemoon becoming more sluggish as the day progesses? Topic is solved

[yami_]

Well if you want to disable this feature, that is fine, you can do it. Making this a default however is wrong and should not be done, because this is a useful feature that other browsers also have. BTW there is no evidence of this pref being supposed to be used only for caching the back and forward web page navigation:

Code: Select all

// Fastback caching - if this pref is negative, then we calculate the number
// of content viewers to cache based on the amount of available memory.
pref("browser.sessionhistory.max_total_viewers", -1);

I must say that claiming that the program uses its internal setting incorrectly is quite unusual...

[anvakl9]

BTW there is no evidence of this pref being supposed to be used only for caching the back and forward web page navigation

I got it from http://kb.mozillazine.org/Browser.sessi ... al_viewers
also Firefox doesnot use it for undo closed tabs. I tried the steps mentioned in my earlier post on firefox and it cleared the memory after closing just one tab.

Making this a default however is wrong and should not be done

if you mean that browser.sessionhistory.max_total_viewers should not be set to 0 by default on a fresh install, I agree with you. The default value should be -1. People who are having issues of high memory usage and freezes can set it to 0 until a patch is released. But you appear to claim that this setting should be used to cache closed tab too and I disagree with that.

[Moonchild]

This setting is supposed to be used when pressing Back and Forward buttons but PM used it to undo cosed tabs(Ctrl+Shift+t).

the total viewers is the total amount of pre-rendered pages Pale Moon keeps in RAM to prevent unnecessary repaints. This does use memory but does not include anything that would still be running in terms of scripting (which seems to be the issue here). I don't think a lack of memory is the problem for the people having this issue. Please let me know if I'm wrong though and if it's a memory inflation issue, instead.

It is only used on a per-tab basis for fast rendering when pressing the back button. It's not browser-wide and not used for closed tabs.

b) browser.cache.compression_level -> 0 (this is just to fill up the RAM faster)

Nonsense! Cache compression has nothing to do with viewers and does not increase RAM usage.

I tried the steps mentioned in my earlier post on firefox and it cleared the memory after closing just one tab.

Really? so how did you look at memory usage for Firefox, considering it is using content processes? Shutting down a content process is NOT the same as releasing memory in a process, so of course that will be faster.

[yami_]

Off-topic:
I did this test using Firefox with e10s disabled and for me while Pale Moon with default settings freed ~700 MB, Firefox only managed to free ~70 MB. And obviously both browsers did not return to their initial memory usage.

[Harvest Moon]

I know its not my thread but thanks to all who responded!

two quick comments:

1. Splat's suggested IDL disabled method...I think that was a nice touch to find that! But I need to research what else is that Intel process doing in the computer especially since I have WD's latest, largest SSD now (still SATA of course---just not as familiar as with the HDD's)

I will wait until I hear from others before I take any permanent suggestions or the patch is created---if someone were to do that...before taking the offered "work around" for this problem!

2. and second comment is for anvakl9's method...of duplicating it..
I have 8GB RAM so that would explain why I don't get "hit" right away, when launching PMB with the sluggishness until I've had it open and opening and closing tabs for an hour or more---so just when I need everything to be working at their best is when I lose control of PMB...

Oh I see yami and anvakI9 commented back and forth with Moonchild, on a new page 6, with more details....

You both are beyond my understanding of how browsers work in relation to memory requirements and background repaints, renderings etc etc as Moonchild was clear in pointing out---but again I leave it to everyone here to hash these things out and let me know what works for the majority before I go clicking on values and changing them...
Hopefully we're on the trail/track to pinning this down!

Thanks everyone!
Harvest Moon

[PalleP]

By default, the CC uses incremental collecting which can be disabled by setting dom.cycle_collector.incremental to false.
People suffering from the issue could see if flipping that pref changes the behavior for them or not. It could be better or could be worse if there is a change - please report.

I did that when you made the post, and it actually seems to help.
This is said with reservation because of the diffuse nature of this problem, but up till now I have not been able to stress PM and memory usage seems to act different (load/unload).
In a few days I will try logging into gmail and see what happens
Have noone else any feedback on this

[Hmm81]

I tried briefly using other firefox forks to see which ones exhibit the same problems, seamonkey 2.9.x versions which forked firefox 52 esr appears to consistently have similar weird issues with gmail. Icecat ver. 52.6 sometimes has the same issues, which I presume is based off of 52 esr as well. I also tried firefox esr ver. 52.9 and 52.6 and both had the same issue. Firefox ver. 53, 54 and 56 doesn't seem to have the same problem.

Perhaps this information will help track down the problem.

[Cassette]

I tried briefly using other firefox forks to see which ones exhibit the same problems, seamonkey 2.9.x versions which forked firefox 52 esr appears to consistently have similar weird issues with gmail. Icecat ver. 52.6 sometimes has the same issues, which I presume is based off of 52 esr as well. I also tried firefox esr ver. 52.9 and 52.6 and both had the same issue. Firefox ver. 53, 54 and 56 doesn't seem to have the same problem.

Perhaps this information will help track down the problem.

If you really want to dig into it, you can try the nightlies. It looks like the 52.x builds go between 9/19 and 11/14.
https://ftp.mozilla.org/pub/firefox/nightly/2016/

Are you able to consistently get it to happen? If so how?

[yami_]

Would not it be easier to just use mozregresion?

[Cassette]

Would not it be easier to just use mozregresion?

I had never used the tool. Would it not be more helpful to not assume others know everything you know?

https://mozilla.github.io/mozregression/
The link has a helpful video explaining the function for anyone interested.

[PalleP]

I tried briefly using other firefox forks to see which ones exhibit the same problems, seamonkey 2.9.x versions which forked firefox 52 esr appears to consistently have similar weird issues with gmail.

If you only test for the Gmail problem, you can do brief testing of the other browsers since it is so obvious that PM goes banana once logged in to Gmail.
The sluggishness can not be tested by briefly trying other browsers since the problem is so inconsistent.
Maybe the 2 problems are connected, maybe not.

[Rickkins]

Oh I'm pretty sure it has something to do with ghost windows and the cycle collector, as well. But what -exactly- is still the question.
Unfortunately it's not that straightforward to tweak because none of the involved CC configuration variables have preferences to change them.
By default, the CC uses incremental collecting which can be disabled by setting dom.cycle_collector.incremental to false - I just looked at the timing involved and the incremental slice of it may simply be too short (Mozilla set it to 5 ms) which -could- result in too much overhead. But that's just a bit of speculation right now.

People suffering from the issue could see if flipping that pref changes the behavior for them or not. It could be better or could be worse if there is a change - please report.

I made the change last night before bed, so I could have a clear run at it today. By supper time I should have a good idea if it helps...

Thanks.

[Harvest Moon]

Just a newbie after all these years of computer classes and self-help teachings!

I tried briefly using other firefox forks to see which ones exhibit the same problems, seamonkey 2.9.x versions which forked firefox 52 esr appears to consistently have similar weird issues with gmail. Icecat ver. 52.6 sometimes has the same issues, which I presume is based off of 52 esr as well. I also tried firefox esr ver. 52.9 and 52.6 and both had the same issue. Firefox ver. 53, 54 and 56 doesn't seem to have the same problem.

Perhaps this information will help track down the problem.

If you really want to dig into it, you can try the nightlies. It looks like the 52.x builds go between 9/19 and 11/14.

https://ftp.mozilla.org/pub/firefox/nightly/2016/

Are you able to consistently get it to happen? If so how?

Amazing! Cassette!

Thanks for proving that someone is busy writing down each time firefox is updated what happens to the various computer resources/memory/storage
but...

I can't seem to open these...W10 does not have a native function and being so nice and clean with new install of W10 I'm installing very very few third party items ---don't want a virus etc...
[My edit: I see re reading it does not make sense---OF COURSE I can unzip the files...but many of them check sum etc... and dir files that open to dead ends keep asking me for the name of the program to open it--does that make sense now? I'm' certain if I had my W7 fully loaded with all my years of collecting all types of opening programs I'd be able to open it...but the commercial place that saved the W7 over and opened W10 for me in same drive overlay---new init etc boot files---said their new 2018 rules forbid them from copying any "programs" over so I lost so many things which are still trapped in the old W7 drive---somewhere hiding away from view {oh oh sluggishness is on again I have to stop editing} :bangshead: ]ended edit in middle]

So can anyone jump me over to what program actually opens these files to view? (without just searching the internet as auto-app function I'm forced to rely upon in new W10 is asking me to identify a app without giving me any choices!

and for the other link offered by Cassette:

Would not it be easier to just use mozregresion?

I had never used the tool. Would it not be more helpful to not assume others know everything you know?

https://mozilla.github.io/mozregression/

The link has a helpful video explaining the function for anyone interested.

Again I'd not like to go searching the web for third party items...does anyone have a direct, virus free, unwanted program free link to go to and get something that installs this (if I wanted it--still thinking of its use in future...I'm curious though to see what I can see in the files)

This single thread is a great teaching tool for debugging for me---to update my knowledge skills... and personal (in mind) kb.

Thanks again all for responding!
Harvest Moon

[Hmm81]

I tried briefly using other firefox forks to see which ones exhibit the same problems, seamonkey 2.9.x versions which forked firefox 52 esr appears to consistently have similar weird issues with gmail. Icecat ver. 52.6 sometimes has the same issues, which I presume is based off of 52 esr as well. I also tried firefox esr ver. 52.9 and 52.6 and both had the same issue. Firefox ver. 53, 54 and 56 doesn't seem to have the same problem.

Perhaps this information will help track down the problem.

If you really want to dig into it, you can try the nightlies. It looks like the 52.x builds go between 9/19 and 11/14.
https://ftp.mozilla.org/pub/firefox/nightly/2016/

Are you able to consistently get it to happen? If so how?

Hmm... I guess when I have some time I can look into the nightlies to see when it was fixed in 53.

The gmail problem is basically one where (1) you login to gmail (2) logout of gmail (3) close the tab (4) open a new tab (5) do an about:memory. Usually you will notice a memory leak of some sort with a detached window or ghost window related to gmail still. It's been proposed the issue that many are having is that when garbage collection occurs, it can't properly get rid of the leak, causing occasional sluggishness. Perhaps others can check when they are having sluggish issues if there are other detached windows from other sites and not just gmail. I just know that gmail almost 100% consistently has a memory leak, which I also see as a potential security issue, that's more important than perhaps the occasional sluggishness, although that's annoying as well.

[yami_]

does anyone have a direct, virus free, unwanted program free link to go to and get something that installs this

If you are talking about mozregression then you can download it here: https://github.com/mozilla/mozregressio ... on-gui.exe. mozregression is a official Mozilla tool designed to find both regressions and bugfixes. If you want a mozregression tutorial I think I could write something.

[Hmm81]

So I checked the nightlies for ver 53. Prior to 12/1/2016, they simply crash when trying to load up a webpage. But it appears after 12/1/2016, ver 53 nightlies are already working correctly with gmail, despite being only 15 days into development. Not sure this helps in anyway.

[Harvest Moon]

does anyone have a direct, virus free, unwanted program free link to go to and get something that installs this

If you are talking about mozregression then you can download it here: https://github.com/mozilla/mozregressio ... on-gui.exe. mozregression is a official Mozilla tool designed to find both regressions and bugfixes. If you want a mozregression tutorial I think I could write something.

Thanks for trying to help me understand yami!

My WIndows 10 "defender" program stopped me from running the exe program...

https://i.postimg.cc/7YWFLDPK/Cant-run-mozregression-gui-anyway-W10-is-defending-against-it.jpg

Guess I'm going to have to listen to those Microsoft designers as I don't want to make any more of this than just having sluggish typing annoying my Email and other web activities..

[Hmm81]

So the first beta release of 52 doesn't work with gmail correctly, showing the same issues as current palemoon versions show. The last alpha version 1/23/2017, doesn't seem to have that issue. Anyone want to try to confirm? Took me a couple hours to test various versions, so hopefully the info may lead to something.

edit: I didn't realize there was a difference between aurora and central builds, assuming that a2 was alpha 2. Anyways, after getting central nightlys to stop crashing by unchecking the multiprocessor option, I found out 10/25/2017 works although sometimes can be slow closing gmail's tab, stalling firefox for at least half a minute to a minute, but it looks like if there's a detached gmail window, hitting the garbage collection buttons works. 10/26/2017 and onwards don't appear to garbage collect correctly, although it looks like the stalling part doesn't seem to occur when closing the tab.

Not sure why the aurora developer mode builds seem to work...

[Rickkins]

Oh I'm pretty sure it has something to do with ghost windows and the cycle collector, as well. But what -exactly- is still the question.
Unfortunately it's not that straightforward to tweak because none of the involved CC configuration variables have preferences to change them.
By default, the CC uses incremental collecting which can be disabled by setting dom.cycle_collector.incremental to false - I just looked at the timing involved and the incremental slice of it may simply be too short (Mozilla set it to 5 ms) which -could- result in too much overhead. But that's just a bit of speculation right now.

People suffering from the issue could see if flipping that pref changes the behavior for them or not. It could be better or could be worse if there is a change - please report.

I made the change last night before bed, so I could have a clear run at it today. By supper time I should have a good idea if it helps...

Thanks.

Ok, so it's the end of the day, and I can report back.

Now, I realize that one day and one person barely rises to the level of anecdotal, but...

... at this point, I have not had to restart the browser at all. Not even once after the whole entire day, and that is certainly progress.

You have my thanks.

[yami_]

Thanks for trying to help me understand yami!

No problem.

Off-topic:

My Windows 10 "defender" program stopped me from running the exe program...
https://i.postimg.cc/7YWFLDPK/Cant-run- ... nst-it.jpg
Guess I'm going to have to listen to those Microsoft designers as I don't want to make any more of this than just having sluggish typing annoying my Email and other web activities..

So in Windows 10 they started to call it Windows Defender SmartScreen? I guess this sounds scarier that Windows SmartScreen used in Windows 8. SmartScreen was originally a part of Internet Explorer that functioned similarly to Chrome Safe Browsing file scanning component. It works by sending some kind of application signature to Microsoft. After receiving the signature Microsoft servers will check if other Windows users downloaded this application and if the application is digitally signed the servers will check if the other applications signed with the same certificate where downloaded by Windows users. Based on those two checks a application reputation is created. If that reputation is to low Windows will display the scary warning dialog that you saw. This kind of download checking can be used to defend the user from fresh malicious applications that will not be recognized by a normal AV. So why was SmartScreen integrated in Windows Defender? You see, in Windows 8 Microsoft decided to give application developers an option to use Extended Validation certificates of code singing. EV certificates are more secure and more expensive versions of normal certificates. Because of that if a EV certificate is used to secure a TLS connection the web browser will display a green bar containing the organization name in the URL bar. If you use a normal certificate the bar will be blue and instead of organizations name a domain will be displayed. But what changes from the user perspective when you sign an application using a EV certificate? Nothing at all. And this is where SmartScreen comes in. When SmartScreen was integrated in Windows Defender it was modified to automatically give an application the reputation if the application itself is signed with an EV certificate. The application is still checked but if the same certificate was not used for malicious things the SmartScreen warning will never appear. Was SmartScreen integrated in Windows Defender to force developers to use EV certificates? I do not know, and Microsoft denies such accusations:

Detractors may claim that SmartScreen is “forcing” developers to spend money on certificates. It should be stressed that EV code signing certificates are not required to build or maintain reputation with SmartScreen. Files signed with standard code signing certificates and even unsigned files continue to build reputation as they have since Application Reputation was introduced in IE9 last year. However, the presence of an EV code signing certificate is a strong indicator that the file was signed by an entity that has passed a rigorous validation process and was signed with hardware which allows our systems to establish reputation for that entity more quickly than unsigned or non-EV code signed programs.