Should have been there when the great Documents and Repositories split
was done.

The Redis CLI history file was moved to the Documentst folder so it
would synced using Nextcloud. A symlink was left in its place to point
to the new location. However it's a dangling link until the Redis CLI is
first run, which upsets the pre-commit broken link check. Instead, set
the REDISCLI_HISTFILE environment variable to the new location and
remove the symlink.

Have Nextcloud backup and sync them.

Like before, Nextcloud!

I don't want to fiddle with the Nextcloud client which directories to
ignore in my Documents folder. So create a new directories,
"Repositories" for repositories that are in some form of a version
control system and are backed up else (so Nextcloud shouldn't sync
them).

When using a remote Docker host, the images are build locally.

I think I added it to expand aliases with `git exec`. But now `git rm`
fails because it expands the `rm` alias.

There's no code I worked on there, it's like the node_modules directory.
Move it under .local so it's not copied by Nextcloud.

Use the one from the Flatpak in Flathub.

Make it available to others.

It's about time.

Nextcloud will back it up!

Google SDK CLI profiles for Shore and Endless.

This reverts commit 393f2782.

It's breaks a few things.

Use the variable file matching the workspace when validating. Also when
validating and formatting, show the diff.

The prune_ssh_sockets shell function would always delete the
ControlMaster sockets (the open file handlers would have a random suffix
and the check would always show that the socket wasn't open by any
process). Instead add a prefix to the ControlMaster sockets and ignore
them when pruning. Those sockets are always deleted when the process
exists. Also, mark the vim filetype for the ssh config fragments as
sshconfig.

Don't delete sockets that are still in use (have an open ssh client
listening).

Per
https://docs.ansible.com/ansible/latest/reference_appendices/config.html#the-configuration-file
the ~/.ansible.cfg file is ignored if another is found first. So use
environment variables instead so they work with all of my setups (the
internded purpose). Also, for the ansible-local and
ansible-local-playbook aliases, use the Python3 interpreter that was
used to run Ansible in the first so that modules (like postgresql_privs)
that require Python module have access to them.

- Rename them to tf* instead of t* (tr is already taken, so avoid that).
- Don't use a variable file for the workspace if not variable file
exists.

In some cases (like a chroots or containers with minimal /dev)
/dev/stderr is not available. Replace outputting to fd 2 instead.

git grep -l '>> /dev/stderr' | xargs sed -i 's@>> /dev/stderr@>\&2@g'

The checks for open ports/ sockets was timinig out after 1 second which
isn't reasonable.

Too many changes in one commit that aren't relevant to each other all
because I didn't commit these for some time and they piled up, exactly
what I shouldn't do.

- Remove some Git submodules. They're not needed.
- Remove the REPREPRO_BASE_DIR environment variable, it's set with
direnv inside that project.
- Use Docker Buildkit.
- Use auto Python interpeter with Ansible, prefer Python3.
- Better ls aliases.
- Netcat aliases to listen and check ports and sockets.
- Aliases and functions for Terraform.
- Use Black to format Python code.
- Show the run duration and exit code in a line before the prompt, maybe
better with Bash earlier than 5.0, where there's an annoying with line
wrapping.
- Valid name for the flatpak_kill shell function.
- Delete user-dirs files, didn't do anything anyway.
- Convert README from reStructuredText to Markdown.
- Don't install gsutil with Python2, use snap to install the Google SDK.
- More pre-commit hooks, update the exiting ones, format the file for
yamllint, Use flake8 from it's repo, the one from the hooks repo is
deprecated.
- Clean the Shore SSH config.
- Add the shared AWS keys to the Endless SSH config.

Back to ElementaryOS, Flatpak is older there.