When moving the configuration files to $XDG_CONFIG_HOME I forgot to
change the line in .bashrc that sources the tmux configuration so bash
was still trying to source ~/.tmux.conf, which no longer exists.
This commit fixes that. Additionally, the color setting code has been
factored out of tmux.conf into tmux-colors.conf (which can be sourced
from bash profile and tmux.conf). This way the rest of the code in
tmux.conf doesn't have to be executed again when changing the background
color.
The previous implementation failed to prevent duplicates if the entry
being added was the only entry already present in the path. The reason
is that it always required a : to be present in order to remove the new
entry before adding it.
The new implementation is clearer and solves that. However, if the new
entry is already present it will be kept where it was, not moved to the
front of the list. This is fine at the moment.
SC2123 warns that the PATH variable is being manipulated, but this is
exactly the intention here.
SC1091 warns that /etc/profile cannot be checked. This is fine, as this
file is provided by the platform and there is no need to check it here.
shellcheck(1) is warning that the single quotes prevent the expression
from being expanded. However, this is the desired behavior because this
expression is meant to be interpreted by vim(1), not bash(1).
This makes sure the return code of extracting the path from manpath(1)
is not ignored. Also, unset the MANPATH environment variable instead of
just assigning an empty string to it, so that manpath(1) determines a
suitable default hierarchy to search for manual pages.
The way I was doing the test is incorrect, as [ is the test(1) program
and the shell builtin command is not something it understands. Instead,
I now just let the shell execute `command` and then set the variable if
`brew` is available.
This requires the local key (fernandos@N17FERNANDOS3) to be added to
~/.ssh/authorized_keys but uses a more standard and straightforward way
to connect via SSH.
Also, we now connect via a host-only network. This makes us independent
from any port forwarding later.
`bash` will not save history if this directory doesn't exist - it will
create the file but not the directories in the path.
Just to be safe, we do the same for `less`.
As I am creating new keys with the ed25519 type, the constraint was
inconvenient. I now let SSH use the default, which will use either RSA
od ED25519, whatever is available.
Previously the global namespace of all interactive python session were
polluted with all the stuff we did in this initialization script. For
example, globals()['__doc__'] would show the docstring from this file.
While none of this was ciritical, it could be confusing sometimes. So
this commit fixes that.
This bug caused entries to not be removed if they were already present
in the beginning of the list before being prepended. This could lead to
duplication of entires, usually this ended up being /usr/local/bin.
Since the old ports I was using are now no longer accessible from work,
I had to change to the following:
- 143/imap for SSH logins and tunneling
- 587/smtp for access to git repos over SSH
Hopefully these ports will stay open because they are common services.
The default value is:
let g:ctrlp_working_path_mode = 'ra'
which means:
a - like "c", but only applies when the current working directory outside of
CtrlP isn't a direct ancestor of the directory of the current file.
r - the nearest ancestor that contains one of these directories or files:
.git .hg .svn .bzr _darcs
So this makes it more convenient to work within repositories, which is
what I mostly do anyway.
