Jonathan Holvey's questions

I'm in the process of migrating data to a bcache partition. I've created the backing disk and cache disk, and I used dd to copy data from another partition (of the same size) to the new /dev/bcache0 partition.

Unfortunately, it appears that even though the backing partition /dev/sdb1 is the same size as the source partition, the actual bcache partition /dev/bcache0 is slightly smaller, and the dd operation failed.

Checking the disk reveals the difference between the physical and superblock sizes of /dev/bcache0:

# fsck /dev/bcache0
The filesystem size (according to the superblock) is 122086400 blocks
The physical size of the device is 122086998 blocks

I've been able to increase the size of /dev/sdb1 using parted, however I'm reluctant to do the same with /dev/bcache0 without knowing the implications, and without calculating the new size very carefully.

What's my best course of action from here to get the /dev/bcache0 partition mountable?

There shouldn't have been any data at the end of the source disk, so I'm assuming the data on /dev/bcache0 is intact, but inaccessible due to the partition misconfiguration. Unfortunately I also accidentally deleted the data from the source disk, so I can't migrate it a second time. I have a working backup of all the critical data, but I'd rather see if I can recover the bcache partition before resorting to that.

Google Chrome has an option to launch a web page as a windowed 'Chrome app', which, when used with Gnome Desktop, will create a special .desktop file which can be used to run Chrome in a stand-alone mode with its own icon displayed in the dash/dock and window switcher.

The problem with this mode is that Chrome tries to impersonate the GTK theme resulting in a pretty ugly window title. It also takes up unnecessary space displaying the address bar and other controls at the top of the window. This is especially true when using an insecure connection, which is typical for accessing locally-hosted applications.

An alternative is to launch Chrome with a --app=http://example.com option, which results in a much nicer experience. A comparison of the appearance of the two modes is at the end of this question.

I can create a .desktop file to launch Chrome with the --app option and an icon I specify, but this results in the new window being grouped with other Chrome instances, which all share the same generic Chrome icon entry in the dash/dock and window switcher.

Is there an option, either in Chrome or Gnome, to force the window that results from the .desktop file to behave as a completely separate application?

Google Chrome app:

Google Chrome with --app option:

I'm having issues getting Gnome's Night Light feature to work correctly with my monitor setup. I've recently switched to a new graphics card that has three DisplayPort outputs (in addition to one HDMI and one DVI). My monitors only support DVI and HDMI, so I bought three cheap DisplayPort to HDMI cables to connect them. Everything seems to be working correctly, except the Night Light filter in Gnome.

I initally thought I may be experiencing a known issue in Gnome 3.34.1 where Night Light stops working on Wayland, however switching to Xorg didn't help for me. Anyway, I believe that issue is now fixed. I'm using Gnome 3.34.3 on Ubuntu 19.10.

I switched back to the old graphics card (AMD HD 5870) and the issue went away, making me suspect either the new graphics card (AMD RX 570) or the cables. I haven't changed any drivers or firmware - I have the packages firmware-amd-graphics and xserver-xorg-video-ati installed. The output of xrandr --verbose is at https://pastebin.com/E9cvPdZ0.

I tried various combinations of cables for my three monitors and this is what I found:

Connections             | Result
------------------------+-------
HDMI    -       -       | Y - -
DVI     -       -       | Y - -
DP-HDMI -       -       | Y - -
DVI     HDMI    -       | Y Y -
DP-HDMI DVI     HDMI    | N N N
DP-HDMI DP-HDMI DP-HDMI | N N N

In each case I restarted the computer to ensure the result could be consistently reproduced.

What technology does Night Light use to adjust the colour temperature of a display, and how could this be affected by the DisplayPort to HDMI cables I have? Should the feature be able to work with such a cable, and how can I find one that will work?

I've been using a TBS6281SE TV tuner without issues for a couple of months, after installing TBS's fork of the LinuxTV open source drivers from GitHub.

Today the tuner has stopped showing up in the system. It is listed by lspci, but not by lsdvb (the latter reports an error opening /sys/class/dvb). There is no obvious issue reported in dmesg, and, naturally, the tuner isn't recognised by TVheadend.

I haven't made any changes to the system recently. I tried re-building and installing the driver, and installing available software updates from apt upgrade, but nothing has helped.

I'm using Ubuntu 16.04.3, with kernel version 4.4.0.

What steps can I take to determine what has gone wrong? How should I attempt to get it working again?

I have several services running on my home Ubuntu server, which have HTTP web interfaces for configuration. Naturally, each is accessed using the machine hostname and a different port number, such as http://media-server:9981.

Since these port numbers are difficult to remember (I currently use suggestions from my browser's history to access them) is there a method I can use to access them all using port 80, but with a different DNS name for each, such as http://tvheadend?

A simple command line application, or IP table configuration, if possible, would be perfect. Note that I would only need this to work on my local network, not over the internet.

I'm connecting to an OpenVPN server from Ubuntu 16.04, with OpenVPN 2.3.10 from the command line using sudo openvpn --config client.conf, or as a service using sudo service openvpn@client start.

I want the resulting connection to be used as lowest priority, so I'm trying to set the metric of the default route it creates to 1000, however I'm not sure how to achieve this. The OpenVPN manual states:

--route-metric m
Specify a default metric m for use with --route.

I'd hoped that adding route-metric 1000 to my config file would affect the default route that OpenVPN creates automatically, however I have to assume that it's only used with the --route option, and the route in question still has a metric of 0:

Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         x.x.x.x         128.0.0.0       UG    0      0        0 tun0

I can see three ways round this problem. How can I go about achieving one of them?

Option 1: Tell OpenVPN to use a different metric when creating the default route

Option 2: Prevent the route from being created in the first place, and use the --route option to create my own using the vpn_gateway variable provided by this option (I think)

Option 3: Modify the default route using the --route option. This will require knowing all the parameters of the route, so some scripting may be required.