[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Firefox 104 on Sid unusable because it blanks open tabs after a few minutes



Hi David,

 

thank you for spending your time to ponder and write about my FF problem.
I try to answer some of your question below.

 

On Donnerstag, 8. September 2022 11:27:01 -04 David Wright wrote:

> On Wed 07 Sep 2022 at 17:52:09 (-0400), Eike Lantzsch KY4PZ wrote:

> > On Mittwoch, 7. September 2022 16:23:28 -04 Alex King wrote:

> > > On 8/09/22 04:16, Eike Lantzsch KY4PZ wrote:

> > > > I admit that I usually have ~ 70 tabs open.

> > > > FF does not blank all, just most of them.

> > > > Then I have to reload the pages and enter all information again

> > > > if

> > > > the pages contain forms like snail-mail tracking, tracking of

> > > > packages, tracking of processes with the local communications

> > > > state

> > > > agency, following geomagnetic storms, RF-Propagation, blogs etc

> > > > etc.

>

> So are we to understand that an even greater majority of your tabs

> have information entered by you?

>

Yes that is so, but it is no sensitive information.

For example I use to have this dynamic page open.

 https://www.dxmaps.com/spots/mapg.php?Lan=E&Frec=50&ML=M&HF=N
If it shows up (out of cache) it sits on 'VHF & up' and on '6m'
I change it to 'LF - HF' and '10m'.
It then updates the information (spots and daylight) automatically.

With this static page, when (or if) it is shown out of cache I Ctrl-R and it does show the updated information of DX-expeditions. That is only necessary once a dayhttps://www.ng3k.com/Misc/adxo.html
But when FF decides to (or stumbles over its feet) blank the page I have to Ctrl-R several times until the page 'stays put'. Several times because if this happens, FF does it again and again for 3 to 4 times.


On this government page I used to enter my code 'Código de Expediente' once a day and then, by entering the simple captcha text, am able to see how far my document/form has advanced in the beaurocratic process and in which department it is stuck right now.https://staf-servicios.conatel.gov.py/ConatelJavaEnvironment/servlet/com.conatel.consultaexp
But if FF blanks the page I have to reload and enter the code and then enter another captcha text. Thus FF is causing me more clickety-click-work than necessary.

> > > > My time is too valuable to waste it reloading and reloading web

> > > > pages

> > > > again and again.

> > > > It happens with dynamic pages and with static pages. There is no

> > > > consistency at all.

> > > >

> > > > I looked into the FF settings for a setup - but no - there is

> > > > nothing

> > > > about this behaviour.

> > > >

> > > > OK, so which browser will I now use instead of Firefox?

> > >

> > > A couple of FF Extensions you might benefit from:

> > >

> > > Auto Tab Discard: although it does the opposite of what you want,

> > > (I.e. the purpose is to discard or "unload" tabs,) it gives a UI

> > > into

> > > the settings for when tabs are discarded.  (Although I'm on FF

> > > 91.13.0esr, and I don't know if it works on FF 104.)

> > >

> > > Textarea Cache:  Allows to save automatically the content in a

> > > text

> > > input field.  Great for retrieving text you wrote into a website,

> > > e.g. if the submission failed for some reason.  Not sure if it

> > > stores text from text input fields on discarded pages or not...

>

> Does it know the distinguish fields that might contain sensitive

> information (like a password)?


It does and, provided the page is shown at all, this page used to be shown after startup exactly as I set it. But when FF decides to blank the page I have to reload and adjust the size of the map and the focus again. Also have to click the buttons to get the information of the desired band (range of frequencies) and the modes of modulation.

 https://pskreporter.info/pskmap.html
Tedious to say the least ... especially if this happens several times.


> > Opera has a setup option "Snooze inactive tabs to save memory" which

> > can be disabled. I didn't find anything similar in the settings of

> > FF but

> > there is  in about:config

> > browser.tabs.min_inactive_duration_before_unload

> > I am testing this right now with a higer number of seconds (I

> > believe that it is set in seconds).

>

> I just rotated through all ~150 of my FF tabs, and at the end of that,

> there are about a dozen FF processes running (top calls them Web

> Content). I noticed some of the tabs date from back in March. I

> suspect many of the pages redisplayed themselves from cached files

> rather than downloading the content again, but I'm just guessing.


The pages are first redisplayed from cache - that is right.
Dynamic pages then are updated automatically - at least that was the case until this problem started to appear.

Static pages have to be reloaded manually. But because my individual setup of that page is retained a simple Ctrl-R or click on the reload button solves that.

 

When the page has been unloaded the tab just appears blank but the URL still appears in the . When I Ctrl-R it loads the page from the website, wich Ctrl-R is supposed to do. There is no shortcut or button to display a page from cache.

 

> > Weird thing is that FF tends to unload the tabs agressively when

> > starting to use the browser but later when I've gone through the

> > process of reloading the pages and filling in requested data,

> > adjusting the focus of maps and whatnot *several times* then it

> > seems to succumb to the will of the user and maintain the pages

> > loaded. Absolutely weird behaviour ...

>

> If "unloading tabs" means that they have to be repainted, rather than

> instantly appearing again when you switch to them, then my experience

> might tie up with that. I never close my main browser: it gets killed

> when I either kill X or shutdown (by command or power button).


The pages have to be repainted from cache when starting FF. But once they are blanked they have to be reloaded from the website.

>

> I'm happy with the ?unusual situation that when I start FF with the

> URL of our weather forecast, I get two tabs. The second contains the

> forecast itself, which I close, and FF displays the following in

> the remaining tab:

>

> --✄✄--------

>

> Sorry. We’re having trouble getting your pages back.

>

> We are having trouble restoring your last browsing session.

> Select Restore Session to try again.

>

> Still not able to restore your session? Sometimes a tab is causing the

> issue. View previous tabs, remove the checkmark from the tabs you

> don’t need to recover, and then restore.


I'll try that.

>

> [View Previous Tabs ↓]          [Start New Session] [Restore Session]

>

> --✄✄--------

>

> I press Return (or click on the highlighted [Restore Session]

> button, and the rightmost old tab is painted. If I now click on

> some other tab at random, FF will restore it to the same contents

> as when I left it, eg scrolling down to a particular point in

> the text, google map, street view, etc. Each tab will take a moment

> or two. Some definitely reload from the web every time, like my

> grocery coupons, which I guess is a dynamic page. (I will typically

> have no tabs left open with forms on them.)

 

I have some pages open with forms on them but no sensitive information. No Logon and such ...

I think this is the only one:

 https://www.karhukoti.com/maidenhead-grid-square-locator/?grid=GG14eq
I only enter a so called grid square into the form and get the geografical location.
Right now it is 503 "Service unavailable".

Yes, what you describe is the behaviour which I expect and which always worked until FF developed the problem with randomly loosing its marbles and showing blank pages with the pertinent URL still in the address (or URL) field.

 

>

> Does this mean that "FF tends to unload the tabs agressively when

> starting" is just a consequence that a new instance of FF only

> loads tabs if and when you switch to them? (I think my computer

> would be very unresposive for quite a while if FF were to try to

> open ~150 tabs at once each morning, and I know people here for whom

> 150 is not that many.)

That is indeed cause one to lose a lot of time because if the problem occurs then each tab has to be reload from the website and that takes some time.

 

FF used to start with all the tabs loaded from cache. That is quick. Then I use to reload (mostly just click on the pages which I want to be updated). Dynamic pages update by themselves.


This has been my "work-flow" for years. I really don't want to change the way I work to accomodate a misbehaving software.
Opening one website at a time in just one tab is no solution for me.

(Anecdotal: Anybody has experience with SAP? They use to remodel organizational structures to accomodate their humongous software packages. That is why they use to lose business with some huge organizations which already have a work-flow which works well.)


>

> > Textarea Cache is a fine extension for what it's worth and for what

> > it's intended but unfortunately does not help in my usecase

> > (entering package codes, adjusting maps for size and focus, filling

> > forms of state agencies. It does not even cache the text entered on

> > WhatsApp or Google Translate. Answer is "No cache here".

> > Anyway I leave it installed because the possibility to save text

> > comes in handy when e.g. answering to blog posts.

>

> Sorry not to have answers, but I was trying to replicate (or really

> just follow) your narrative and didn't understand all the terms.


No reason to be sorry. I really appreciate your comments. If there are no answers you at least helped me to recapitulate and to develop a clearer picture of what is happening and how to deal with it.

Weird thing is: since I posted my first message of this thread, I adjusted about:config
browser.tabs.min_inactive_duration_before_unload

to a million seconds and the problem did not appear anymore.
I don't know yet if "this is it!".

I got enough RAM and am observing my os' swapping so I think that it is not a memory problem.

 

Since I also did some package updates in between I really don't know if the problem is solved now or if it will again, at some time, rear its ugly head.

 

Thank you again and all the best to you
Eike


--

Eike Lantzsch KY4PZ / ZP6CGE



Reply to: