Thursday 18 January 2018

KDE 4 vs Plasma 5 - The Farce Awakens

Well, that might be a little harsh.  But the move to Plasma 5 is still a pain.  As I've written previously, I'm running a Gigabyte BRIX with a Celeron N3150 1.6GHz 64 bit CPU (about 2GHz in burst mode) and 8GB of installed RAM.
It is working pretty hard for a low end processor, but it wasn't really all that long ago (1984) that I was using an $8,000 Epson QX-10 with a Z80 8-bit processor at about 4MHz and about a quarter of a Meg of RAM for all my heavy business work.  So I'm pretty happy with what the $300 BRIX can achieve.

Ok, so back to 2018.  I've used most of the main Linux distributions since we first converted all our business computers to Red Hat in 1998. Before the end of the year I had installed Mandrake on one computer and that was the beginning of my experience with KDE.  I've used almost everything else available in Desktop Environments, but I always end up back at KDE.

I settled on Linux Mint as my favourite distro quite a few years ago after it was suggested by a friend.   Over the last five years it has really improved and although I often mess with other distros, I keep a copy of the latest Mint KDE in my main root folder.  The second root folder is used for either testing new releases of Mint, or dabbling in other distros, while enabling me to immediately reboot into something stable and familiar if anything goes wrong.

And that is what happened when I installed Mint 18 point something last year.  Mint KDE had moved to Plasma 5 and so many things didn't work any more.  Now I had tried NEON in the early days of Plasma 5, and given up on it.  So my main distro is Mint 17.3 with Plasma 4.13.2, and my spare distro is Kubuntu 17.10 with all the latest bits of Plasma 5 and QT.

Here's the How the desktops look:

Mint 17.3 with KDE 4 (above) 


Kubuntu 17.10 with the latest Plasma 5 and all the beta bits (above)

So how do they compare?  KDE 4 is mature, slick and almost bullet proof.  Even on the Celeron, it is snappy with no lag, and no compositing problems.  Everything just works.  A bit like KDE 3 was when KDE 4 first came out.  And back then we all complained about how slow and clunky KDE 4 was.

That's what Plasma 5 feels like compared to KDE 4.  Slow and clunky.  Plasma 5 takes forever to boot, although the latest is a little faster to shut down.  It is also a little laggy when switching work spaces, even with the timing set to zero.
I also tried it with Wayland instead of X11/Kwin.  That was a mistake.  Almost nothing works properly, but luckily CTRL+ALT+Backspace got me logged out and I could log back into a standard Plasma session.

That will all change though.  Work seems to have more or less stopped on KDE 4.x, and pretty well everything is being focused of trying to sort out Plasma 5.  Each time I upgrade it, there seems to be a few less bugs.

One of the worst is the apparent memory leak in 'heap'.  If I run either the Screensaver Wallpaper option on the desktop, or if I set a Media Frame Applet on the desktop, I can run KSysGuard and watch as the plasma-desktop process just grows and grows every time an image changes.  Left alone it will soon eat all of my 8GiB of RAM and start using my 16GiB of Swap space.

The plasma-desktop process also eats RAM when doing something as simple as clicking the Application launcher on the Panel, then clicking the desktop to close it.  Doing that over and over, but never clicking on a menu item, it can leave up to 100Kib or more in plasma-desktop each time.  It soon takes the used RAM from say, 1.8GB to 2.4GB.  Little things like that need to be fixed before Plasma 5 will be ready for everyday use.  But as long as you set up the Plasma Restart icon in the Panel

(see an earlier post about 'kquitapp5 plasmashell && kstart plasmashell')

 it only takes a moment to free up RAM.  And the problem happens much more slowly now, especially if you remember NEVER to leave Desktop Wallpaper running or install the Media Widget.

Luckily I have a script on the panel that reboots Plasma without affecting other running apps, and without having to log out and back in.

This update is being written as a follow up, because I did a major upgrade of Plasma 5, Frameworks and also Qt today and There are a few slight improvements.  I still boot into Kubuntu 17.10 with Plasma 5 frequently, but the moment I finish messing around there, I sigh with relief when I'm back in Mint 17.3 and KDE 4.

Sadly, that can't last.  Eventually KDE 4 will be phased out like KDE 3 was.  I just hope Plasma 5 is matured enough by then to take over the heavy lifting.

The one good thing is that both are still very easy to customise.  As you can see in the screen shots, I can still have my translucent title bars and panels, and I have Cairo-Dock on KDE 4 and Latte-Dock on Plasma 5.





Saturday 13 January 2018

Kubuntu 17.10 Working. Solution seems to be - DON't use 'Slideshow Wallpaper'


Kubuntu 17.10 with Plasma 5 working at last!
Solution seems to be - DON't use 'Slideshow Wallpaper'

After so much time trying to fix memory issues with Plasma 5, I finally seem to have Plasma 5 and Kubuntu 17.10 working happily - for now. 
The worst culprit seems to have been using the Slideshow option for Wallpaper in Configure Desktop.
I'm sure that could change, and I'm keeping Mint 17.3 with KDE Plasma 4 in the spare root partition for Justin.

I have mentioned before that I have a spare panel on the right side of my desktop mostly to make desktop switching easier by simply scrolling in the blank space at the right of the screen.  But I always liked having Docky running in hidden mode at the bottom for quick access to a few odd things that I use most often.  

In plasma 5, when I installed Docky it had some issues, so I did some snooping around and found 'Latte Dock'.

Check it out HERE



 Latte Dock was in the Kubuntu repository, and it works similarly to Docky, although it doesn't seem as versatile so far.  It did however blend nicely with my theme and in the top screen capture you can see it taking up most of the bottom of the screen - like Docky in Panel Mode.  It doesn't eat much of the system's resources and unlike Docky when I tried it, Latte Dock sits nicely on all my Virginal Desktops AND all my Activities.

Today I have been running most of my usual programs and even with KODI and Waterfox running RAM used after five hours was only up to 2.1GiB and Swap used was zero.

This is a huge change because previously after a couple of hours the RAM was often up to 6GiB used and swap around the same!

Just for now, I am impressed with Kubuntu 17.10 and Plasma 5 for the first time ever in the years that I have been trying to use Plasma 5.

Below is the System Information for the Brix Celeron n3150 system I am using.


Day 2 of the Kubuntu 17.10 / Plasma 5.11 experiment and I have learned a few more things.  I kicked in a browser I have been messing around with - Vivaldi.  There are some things I don;t like about Vivaldi.  In fact I hate them /  One is that I can't find any way to open Speed Dial entries in new tabs with a left click.  Another is that I cannot use my double click actions on the Title Bar, for example Double Click to toggle shade does not work..

However in most other things, Vivaldi is faster and using a few less resources than most of the other high end browsers I have at the moment - even with some extensions installed.

Here's the RAM use comparison yesterday with Waterfox, and today with Vivaldi.



Friday 5 January 2018

The 'MyRepublic NBN Myth' - BUSTED !!!

A little over a year ago I was sick of my new NBN broadband connection being about the same speed as my ols ADSL2+ broadband.  Despite my connection being Fibre To The Premises (the best possible option for speed), it was no better and sometimes not even as good.  I was still paying about the same - $75.00 per month, and even though I had a 25/5 Mbps plan compared to 22Mbps ADSL, there was no noticeable improvement.

Then someone told me he was moving to a company called MyRepublic to try them out.  So I did some checking on Google and found that they offered speeds 'up to' 100/40 Mbps.  No wild promises - just that they would supply the best speeds they could to my home, depending on the infrastructure available.

I expected perhaps around 50/20 or if I was really lucky, 75/30.  I also read a lot of complaints from people about a terrible lack of customer service, written by dissatisfied customers.  Not a great start, but I have been using the Internet since before we had it publicly available in Australia, so I'm fairy resilient.  Based on my research I revised my expectations and decided that if I got consistent speeds as low as 35/5 Mbps, I would be better off, because as well as being at least twice the speeds I got now and also having unlimited downloads, at $59.99 a month the plan was $15 a month cheaper.  As it turned out, the speed was more than four times better.

MyRepublic suggested it 'should' be about 20 days between my order date, and an active service.  And checking my log, I see that I ordered and paid the first payment on the 9th December 2016.

And here, I got to experience the so called 'lack of customer service' that people complained about.  I heard nothing from MyRepublic.  In the mean time, and with no notice from MyRepublic, the NBN contractor arranged to finish installing my NTD, and a modem had arrived.  Out of curiosity I plugged it in, expecting to hear more from MyRepublic directly.

One night about two weeks after I had ordered my plan, and about a week after the modem arrived I glanced at the modem and noticed that a red LED had changed to green.  So I disconnected my ADSL modem and tried my shiny new Technicolor NBN modem.  I was online!  So I ran a speed test and got 95Mbps download speed and about 35Mbps upload.  This was far better than I had hoped for and pretty close to the 100/40 I had signed on for.  I knew once I fired up my VPN it would drop significantly, but I was pleased.

However, other than an email on the 9th acknowledging my order, I had still heard nothing from MyRepublic. No notice that the modem was being sent, and not advice that i was connected.

That was until the 29th December (the 20 days they had estimated) when I got an email saying:

Dear Ross,

Congratulations, your MyRepublic broadband service is now active.
To connect your Wi-Fi Hub+ modem, follow the simple steps in the Quick Start Guide.
If you ordered a Home Phone service, you will receive a separate notification, telling you when it’s time to plug in your home phone.

By that time, I had been happily Internetting away on my lovely fast connection for about a week.

So what about the lousy customer service everyone was complaining about?

Well, the reality seems to be that there is a small number of unhappy people and they are very vocal.  The rest of us are not as noisy, because we are quietly sitting enjoying speeds the rest of Australia only dream of.

MyRepublic saw an opportunity to give Australians what nobody else would give them.  The fastest speeds possible on the new NBN, for the price of an average ADSL2+ connection.  Plus, they included UNLIMITED downloads.

But to do that, they had to cut a few corners.  They seem to have decided that delivering on their promise of the best speeds they could provide, while keeping the price as low as possible, was more important than setting up a huge customer service centre.

And I think they have the correct priority.  Here's why.  I have had problems with the service provided by pretty well all my ISPs and RSPs over the years.  And, being a technician with networking, hardware and software skills, as well as a knack for troubleshooting - I have developed some patience.  I have a procedure for solving Internet problems.  You might need help from a tech savvy friend, or your computer guy.

1)     I check my hardware.
2)     I check my network settings.
3)     I wait to see if the problem resolves itself.
4)     If the problem doesn't improve in 24 hours I contact customer support.

And here is the problem with MyRepublic.  There is usually a LONG wait for phone support.  The online chat for tech support doesn;t have as long a waiting time, but unless you have an alternate Internet connection - that is not an option.  And the same applies to Facebook messaging.

However, on the few occasions when there has been a problem, MyRepublic customer support have been great.  They are pretty well aware now that I don't ring up to whine.  I tell them what is happening and what I have done to investigate the problem at my end.  Then they usually send me a message on Facebook Messenger or online chat.

That works, because usually the Internet problem is not a lack of service, but rather a very slow service.  Last November (2017) I had speeds as low as 3Mbps for a while.  But while it was horribly slow, I could still message them.  So instead of jumping up and down and crying in my beer, I simply sent daily    updates so they would be informed while they tried to find the issue.

After about a week they contacted me and said there was a network upgrade happening in my region (Mackay, Qld) and they apologised, but assured me it would be fixed as soon as possible.

By the late November I had speeds around 85 Mbps without my VPN and speeds of about 55 Mbps with my VPN.  That didn't change for a while, and I still don't know if there was a problem with my VPN service.  But in the last days of November, I was back to around 90 Mbps even when running through my VPN.

Now, there is one more thing I often hear in various forums, particularly Whirlpool.net.au and that is the way MyRepublic has OOKLA Speedtest 'rigged' so it shows much higher download speeds than you can actually achieve using MyRepublic.

This is the 'MYTH' referred to in the post title.  There will always be a small difference between the speed shown on various speed test sites.  But there are ways to actually verify your 'REAL' download speed.  For Linux users it is very easy.  Open a terminal (the thing Windows users call a Dos Prompt) and use the wget program to download an 'iso' file of a Linux distribution like Mint or Ubuntu.  These iso files are typically 1.5GB to 2GB in size.  The terminal will show the download speed in MiB/second (we used to call it Megabytes per second), and at the end it will say how long it took to download.

If you miss the download speed, just divide the finished file size in MB by the number of seconds and you have the speed.  Then enter the speed in MiB (MB) into a MiB to Mbps converter (you can find one using Google).

The result is your REAL download speed.  Compare this to the speedtest result from Ookla or your preferred speedtest.  Below I have an actual test I did earlier to illustrate the point.  This download and the speedtest were done while my computer was accessing the Internet through my VPN (Total Sollutions LLC).


203MB downloaded in 19 seconds



203MB divided by 19 seconds = 10.68MB/second


 10.68MiB/second into the converter = 89.59Mbps  (REAL) download speed


 Run OOKLA speedtest to see how MyRepublic supposedly has the speedtest results 'rigged' to show a false reading. 0.44


Yep!!  It's rigged!!  The speed test 'might' be 0.44 Mbps faster that the REAL result - or could it just be a lag to do with the distance between Sydney NSW and Mackay, Central Qld?
Either way. I can't see anything to justify claims that MyRepublic 'rig' the speedtest results.

This test was done on a Gigabyte BRIX with a Celeron BACE3150 (1.6Ghz) Processor and 8GB RAM running Mint Linux KDE 17.  Where it says Total Server Solutions LLC, shows it was running through my VPN in Sydney.  If it was not running through the VPN, it would show MyRepublic, and the speed would usually be between 95Mbps and 98Mbps.

Just to show the consistency of my MyRepublic connection, here are my last speed test for 2017 and my first speed tests for 2018 over three days at different times of the day.  about 2pm, about 3pm and about 8pm.


And here is a collection of individual downloads, again showing how consistently I get around 10 Megabytes or more per second. 


Finally, if you have an issue with MyRepublic, why not take the time to realise they are trying to do what nobody else was willing to do.  Give us cheap, affordable FAST broadband on the NBN.

Sure, there are some other companies offering similar pricing, and a few even do 'unlimited downloads'.  But the pricing and unlimited downloads were pioneered by MyRepublic.  Without them, we would probably still be paying $200 if we wanted 100/40 speeds and the downloads would be capped.

Be patient.  When you do get through, don't rant and rave like a lunatic.  Try explaining the problem and get a channel of communication going with your customer service rep.  They WILL help, once they can find the problem.  Often it is not something they can actually fix.  Much of the infrastructure is outside their control.  But if they know there is a problem, they need it fixed just as much and as soon as you do.  They WANT YOU to be HAPPY.

So I wrote about this on Whirlpool.net.au forum and linked to this post.  And someone wrote in reply to a comment:
Also the link you included from Ross, he only tested speeds in Australia, not speeds overseas.
Providers like Aussie Broadband, Mungi, Telecube and Leaptel provide good speeds in Australia as well as overseas.
Let's be frankly honest here, how many of us just use Australia based websites?

So I decided to add a speed test from Mackay, to my VPN in Sydney, to a Speedtest served in the USA.
NOTE - this IS running through a VPN as can be seen by the originating server 
'Total Server Solutions, LLC
137.59.252.154'  
and it was done at 4pm.   Even through the VPN and testing the USA, it is not bad for a 100/40 account.  Upload is a bit dismal, but it is download speed we count on for viewing Netflix, Youtube and other content.



Thursday 4 January 2018

Kubuntu 17.10 and Plasma 5 memory problems - solution?

For a couple of years now I have been trying to find a way to move to a version of KDE using the latest Plasma desktop.  I have had to revert to earlier releases using Plasma 4.x because every time I used a release with Plasma 5.x I would end up with a crashed system after a few hours, with plasmashell and some indexing processes eating up all available RAM and even most of my swap space.

At last, after tracking down and disabling most of the indexing, I thought I had the problem sorted, but all it did was slowed the gradual increase of RAM use.  Even with a clean boot and NO programs running today, my RAM use went from around 1.2GB to 7.2GB, whole I was out shopping for a couple of hours.

This was the amount of RAM used by 'plasmashell' (basically the main part of the KDE desktop) when I rebooted the computer.  Then I went out to get some medication and food.


I left the computer with NO user programs loaded other than the system monitor.
Plasmashell was using about 70KB of RAM and my entire system was using about 900MB of RAM and about 8.5MB out of 16GB of swap file.

 I came back an hour or two later and looked at the system monitor.  What I found on my return is shown below.


Still only a tiny bit (7.3MB) of swap used, but a whopping 4.5GB of my RAM was being used by plasmashell (the KDE desktop).

Now, bear in mind that I had not opened any other windows or programs.  I had simply left the computer turned on when I went out.

I killed plasma, and restarted it, and my RAM in use when back to about 1.2GB.  I opened Chromium web browser and Dolphin and a few other things and got my RAM use up to a bit over 2GB, then closed each program, and RAM stayed at about 2GB.  And it crept up, while I waited.

So I killed and restarted Plasma again.  There is a handy little script for any time you need to restart Plasma without rebooting Linux.  Simply type this into a terminal and press ENTER:

kquitapp5 plasmashell && kstart plasmashell

By this time I was bloody annoyed and started searching Google yet again.  I had already found the indexing problem and confirmed it was an issue.  But this was something else.  Eventually I came across some discussions among the developers and maintainers of Plasma5 regarding memory leaks.

It seems that some 'plasmoids', little applications that do small jobs within the Plasma shell, can leave bits of themselves in RAM when they refresh.  Effectively I suppose they sort of shut down, leaving a bit of themselves in memorym then when they restart, they load a full version of themselves, then leave a little bit of themselves again, thus building up a stash of useless 'dead' code in RAM each time.

And one of the worst offenders seemed to be the plasmoid that draws a little picture slideshow on the desktop so you can see slides from your pictures folder or elsewhere.  But I don't use that, so I kept digging and found...

Slideshow Wallpaper!!

Yep, this is built in to KDE and allows you to set a folder of images as wallpaper and have it change at an interval you choose.  Anything from seconds to hours between changes.  I had the slideshow set to change images each 30 seconds on one of my 'Activities'.  Now, Activities are what Plasma 5 wants us to use instead of the 'Virtual Desktops' we had in Plasma 4.

The problem is, that in KDE with Plasma 4, Virtual Desktops worked perfectly and Activities worked perfectly.  Then some dickhead decided to 'fix' them.  And neither has worked properly since!

Anyway, I found the Activity that was set for a Slideshow Wallpaper and changed it to a single image.  I've been running the system continuously for some hours since then and my system monitor shows the result below, even with Chromium open so I can type this post..


Chromium Browser however is another program that doesn't play nicely with Plasma 5, and if I keep it open, opening and closing new tabs will gradually add more and more to the amount of RAM being used.  It is now around 1.7GB, but in an hour or two it will be up around 2.5GB.

Fortunately I used that little terminal script to create a shell script to kill and restart plasma, made it executable, and dragged it to my Panel.  I can now restart Plasma desktop with a single click.

!#/bin/bash
# Restart Plasma 5.x to free up memory. 
# Copy this script to the Panel. 
#Change the icon once it is there,
#for convenience.
kquitapp5 plasmashell && kstart plasmashell