serhii.net

xdg-settings gets the award for least intuitive interface ever.

• xdg-settings get default-web-browser was firefox.
• xdg-settings set default-web-browser qutebrowser.desktop is quiet
• xdg-settings get default-web-browser is still firefox.
• man page says that the errors are shown as …return code??
• echo $? returned 2, which is file not found basically.
• Bonus points for not accepting -h (only --help), and having --list as a parameter, but get/set as commands.

> xdg-settings set default-web-browser

xdg-settings: invalid application name

oh well.

Created a file with -> in the name, it didn’t appear on mobile, checked sync logs - not there because the name is “illegal”. Is not allowing > a global thing or only for Android?

For an executable (..qutebrowser.sh) to be an ‘application’, it has to have a .desktop file in ~/.local/share .¹

For qutebrowser, created this:

[Desktop Entry]
Name=Qutebrowser
Comment=Qutebrowser
Exec="~/.local/bin/qb %f"
Terminal=true
Type=Application
StartupNotify=true
MimeType=application/x-www-browser;
Keywords=python;

• To test it, desktop-file-validate qutebrowser.desktop
• To refresh the db, sudo update-desktop-database
• sudo desktop-file-install qutebrowser.desktop then put it in /usr/share/applications ²

This describes all the things needed to set qb as default browser: New option for open link with browser · Issue #716 · RocketChat/Rocket.Chat.Electron

At the end, symlinked /usr/bin/qb to it’s location in my home folder, maybe the universe will come crashing on me but then I don’t have to mess with the usual creation of bash runner scripts in ~/.local/bin/.. to have it globally available. Including for things like update-alternatives that seem to want a global thing.

• Screen screens:
  • screen -R screename attaches a screen with this name or creates it.
    • Tab completion works!
    • You can only write part of the name, will work if it’s enough to identify it
  • <C-a> :sessionname newscreenname renames an existing instance
• ~/.screenrc exists. Some useful settings:
  • defscrollback 200000 for “infinite” scrollback
  • deflog on to log everything automatically
• Using screen when no screen is installed¹ : connect to it with ssh from any other server that does have screen installed.

• ssh -L 6006:127.0.0.1:6006 servername -p 1234 maps port 6006 of servername to localhost:6006, using ssh that’s running there on port 1234
• multiple ports are possible by passing multiple -L arguments

If you do it often, you can add these settings to ~/.ssh/config:

 Host pf
 Hostname servername
 LocalForward 6007 localhost:6007
 LocalForward 6006 localhost:6006
 Port 1234   

…and then you connect to it as ssh pf.

sshfs mounts a remote folder to one on the local filesystem.

• sshfs server:/data/me ./local-folder -p 12345
• sshfs -o Ciphers=aes128-ctr -o Compression=no server:/data/me ./local-folder -p 12345 may be faster

When I tried it at the beginning it was horribly slow, the problem was the zsh prompt that had info about the current git repo. Disabling it or using bash solved the issue.

To Export settings, File -> Manage IDE Settings -> Export Settings ¹

Interestingly the first google result was the similarly named Share your IDE settings | PyCharm, which is a feature in Pycharm Professional and is closer to syncing than to exporting.

If you copy a directory, there may be symlinks there, that will also show fine when you tree or cat or whatever. What saved me was their different color in the terminal.

.. How did people do this in b/w terminals?

TODO How can I avoid this in the future, given my heavy symlinks use?

In Python 3.10+, Unions (Union[str, Path]) can be also written as str | Path¹

… And the syntax str or Path I’ve been using and getting no errors from, apparently, doesn’t exist at all. TODO - why did it work?

Pycharm froze, killed it with killall I think, didn’t see it in the process list even (ps aux | grep pycharm) but couldn’t start it either because it detected an already running instance and refused to start.

The Internet¹ suggested pkill -f pycharm killed whatever was remaining, and I could start it after that. Still no idea what happened though.