- FORK @joesuf4/home (github.com/joesuf4/home) on GitHub to
$your_github_account - cd ~
- git clone https://github.com/$your_github_account/home
- cp -a home/.git .
- git checkout .
- rm -rf home
- bash ~/.bootstrap
- wsl.exe —-shutdown
- install Docker Desktop (or google instructions for installing docker via Ubuntu's apt).
That will get you the whole ball of wax, that you can customize and commit to at will.
You don't need to publish your subsequent commits to GitHub, unless you want to do a PR against the @joesuf4/home repo (or to just share them publicly as I am now).
You can resync your fork with @joesuf4/home future changes and pull them into your ~/ checkout as you see fit.
Below documents what's involved in getting the bcs assume-role $foo $bar
to function correctly for your workplace, assuming you aren't yet a colleague.
Post-Op: (OBSOLETE) The strings ### XSLTPROC ### and ### XSLTPROCREV ### in the
.bootstrap-produced .bcsrc script need to each reflect your
$aws_account_name) echo $aws_account_id;; ...
mappings, written as shell case clauses, concatenated on a single line. The
only difference between the former string and the latter is that the roles are
reversed for ### XSLTPROCREV ###:
$aws_account_id) echo $aws_account_name;; ...
The other tunable is the ${_bcs_roles[@]} array at the top of .bcsrc, to
set the range of assumable AWS roles for your company.
I use this primarily to distribute my home directory's essential files across several servers. The sources are found on my laptop, are pushed to my github account, and from there are cloned to my home directory across various hosts, typically with ansible.
Most of my worktime is spent in zsh running inside a pty-driven (@SunStarSys
has the pty git repo) tmux session on my laptop, and that's reflected in
my config. Obviously I use emacs a lot, and the .emacs file contains a few
essential elpa/melpa packages; I couldn't get by without some of them. Of
particular note is ccls, which is an excellent piece of software to drive
clang-based LSP IDE apps (like emacs).
The configs are messy and not really aimed at public distribution, but I'd be happy to consider appropriate patches if there's something you'd like to see me make use of.
The reusable (static) binaries and scripts are over in the bin/ dir. The were
compiled against Ubuntu 20.04's glibc, and most will not work on newer glibc
operating systems. The screen binary should be suid root for session reattachment
to have any hope of working. The only nontrivial script is bin/pffxg.sh -
a parallelized, filtered recursive grep - targeting very large source trees typical
for embedded development work. I've included some primitive documentation for
it below the next section.
Copy and paste linter.{rc,sh} into your local repo, run ./linter.sh install, enjoy.
Workable for any PR/MR-based workflow and programming language framework imaginable.
I have python, emacs, mingw+git, and Docker Desktop locally installed in Windows itself. No Windows IDE/compiler whatsoever. I launch an emacs daemon at Windows Login Start Up, which I realistically only ever interact with via the Ubuntu emacsclient symlink in ~/bin. All of my emacs-related LSP daemons run within WSL itself, (via Ubuntu-clang-based custom software builds,) through *.bat shims in ~winhome/bin.
I have no emacs install on Ubuntu; or any X11 client/server support for that matter. I want a modern, DRM/4K+HDR enabled, secure, professional graphics/font engines that are promised to work with modern hardware and backed up by a support contract. I insist on paying, or having my employer pay, Microsoft (or Apple, once up on time) for the privilege. X11 ain't that. Not even close.
I keep ~winhome under this git repo, as well as Ubuntu's ~ dir. I have .zshrc
aliases to make synchronization via git push/pull trivial. I rely on
Powershell for CLI Terminal sessions, and drop into my Ubuntu install via the
wsl command.
This is my preferred business environment to work within, and my employer gives me amazing hardware and software support, so I don't need Windows admin privs to get it done. Fuck Apple, this is the cat's meow.
I have static builds of my Linux observability toolchain living in ~/bin along
with a lot of other interesting things. These are stand-alone tools that will
execute on any reasonably current Linux bare-metal host, or running Linux-based
container, or WSL (on a 5.x kernel built w/ CONFIG_IKHEADERS=yes in the config)
itself.
Update (12/22/2021): Experimenting with wslg on Windows 11 (X11 has no alpha channel support yet). Emacs for Windows has some pain points on WSL, so I may need an attitude adjustment against my X11 loathing now the MS is in the game.
Modeled on the API for The Silver Searcher (ag), pffxg.sh is a simpler, faster,
and more feature-rich product in one-tenth the source code line count. Comes
with native support for a tmpfs-based (optionally lzo-compressed) cache, which
provides consistency in search result efficiency unavailable in other similar
products. Portable to any platform that supports bash + standard GNU fileutils,
plus lzop if you need a cache, including Windows 10 and Mac OSX.
To search pwd for foo:
% pffxg.sh foo
% pffxg.sh --help
Parallel Find Filename Xargs Grep (pffxg) VERSION 2.2
Usage:
$0 [ --version | --help | `ag` file extension class | --all | --unzip \
| --conf* $PFFXG_CONF
| --list* | --no-* \
| [--args|--files|--batch] $PFFXG_ARGS \
| --cmd $PFFXG_CMD \
| [--workers|--jobs] $PFFXG_WORKERS \
| --max* $PFFXG_MAX_STATUS \
| --cache $PFFXG_CACHE [--re*]? [--compress]?
| --level $PFFXG_LEVEL \
| --compressor $PFFXG_COMPRESSOR \
| [--excl*|--ignore] $PFFXG_EXCLUSIONS]* [$filename|--] ${cmd_args[@]}
$filename is passed to the -name argument of find(1), so it can be a glob
pattern. The remainder of "$@" are arguments to be passed to $PFFXG_CMD,
which defaults to grep(1) with a few additional --prefixed arguments.
Special case parasitic behavior for `ag` file extension classes:
This script treat unrecognized '--' prefixed args as
file extension classes within `ag` and will limit the search
to files in those extension classes. The default behavior
is to search the entire list of extension classes as a whitelist.
If you really want all files searched (including binary files), pass
the --all flag instead.
# customizable env vars:
# Fully-Qualified Path to (optional) config file
: ${PFFXG_CONF:=~/.pffxg.conf}
# $PWD directories to fully ignore
: ${PFFXG_EXCLUSIONS:=out release prebuilts kernel external hardware}
# The default grep(1) command plus appropriate base arguments
: ${PFFXG_CMD:=grep --color=always --with-filename --line-number --ignore-case}
# Tolerable exit status threshold for $PFFXG_CMD
: ${PFFXG_MAX_STATUS:=1}
# Maximum number of files to list on $PFFXG_CMD command line
: ${PFFXG_ARGS:=10000}
# File extensions to limit the search to; do not prefix these with .'s
: ${PFFXG_EXTENSIONS:=}
# Maximum number of concurrent jobs to run at any given moment
: ${PFFXG_WORKERS:=$(awk "BEGIN { printf \"%d\", 1 + sqrt($(nproc)) }")}
# Location of cache directory
: ${PFFXG_CACHE:=}
# You can enable/disable cache compression at any time and the cache will adjust (with a --refresh)
: ${PFFXG_COMPRESS:=}
# Cache compression level (happens in background so can be relatively expensive)
: ${PFFXG_LEVEL:=7}
# Compression prefix (lzop -U is faster than gzip is faster than pigz here)
: ${PFFXG_COMPRESSOR:=lzop -U}
% cat .pffxg.conf
PFFXG_EXCLUSIONS='[.]ccls-cache [.]*[.]min[.][a-z]*'
My employer has me on a top-of-the-line Dell Precision 5550 i9 vPRO w/8 cores,
which presents as 16 CPU to WSL/Linux. Only WSL kernel builds, eks report global . -n 1000 and broad pffxg.sh searches, light up the fans (for 2-3
min tops). I've had it with "*.min.{js,css} as text-file" greps, so pffxg.conf
excludes them all.
Copy and paste into ~/.emacs to install the pffxg elisp function into emacs:
(defvar pffxg-command "bash pffxg.sh ")
(defun pffxg (command-args)
(interactive
(progn
(list (read-from-minibuffer "Find filenames and grep them (like this): "
pffxg-command nil nil
'grep-find-history))))
(let ((null-device nil)) ; see grep
(grep command-args)))
