June 2018


Rebooting this .plan after a long absence. Again.

More work on switching to Linux.

Switched to pass from 1Password so I can use it everywhere. 1Password has been pushing the cloud version pretty hard, so it's probably only a matter of time before they deprecate the real app. This was painful. I had to hack apart the 1password2pass.rb file to handle my naming conventions, and of course the version of that script in the pass repo was out of date.

Got weechat up and running. I can communicate again.

Got offlineimap running, but didn't finish the syncing yet (I have a lot of mail).

Tore Roswell out of my lispindent and clhs scripts so it just builds with vanilla SBCL now. So much easier and less brittle.

Backlight support (specifically xbacklight) required editing xorg.conf as described in the Arch wiki. The backlight keys on the keyboard still don't work, but at least I can dim the screen.

xcape remains busted in stumpwm for now. Will debug later.


Getting VPN set up. Mostly used the network manager GUI to handle the settings. To disable saving of the password I had to click the silly little icon in the right of each password field.

Got my xrandr bullshit mostly tamed, thanks to some Lisp code from katco. It's really nice to be able to script my window manager to add hotkeys with Common Lisp.


More math review. I'm rusty.

More homework from the Prolog class. This time it was the classic Family Tree style exercise. Wasn't too tough, even though my Prolog is really rusty.

Still on OS X for personal stuff until the rest of the parts for my Linux/Windows machine arrive.


Math review.


Finished the Trains exercise for the Prolog class. It was harder than I expected — my Prolog skills definitely need work. But I guess that's what the class is for, right?

July 2018


The quest to make my new Linux machine usable continues. Finally managed to build Weechat from source after screwing around in dependency hell for a while (apparently there's no libgnutls30-dev for Ubuntu 18.04?). Also managed to get the sound coming out the right ports after some dark incantation with pactl. Not using Gnome as a WM apparently means I'm going to be perpetually confused about how to configure anything on this damn machine, I guess.

Catching up in the Prolog class. Thankfully tomorrow is a holiday so I can hopefully make some good progress there too.


Managed to get Weechat to work with Unicode support. I had installed libncursesw and reran cmake .. but apparently that wasn't enough, I had to fully blow away the build/ directory and redo the cmake from scratch to get it to link against libncursesw (rather than libncurses). What a shitshow. I really need to take a month off and write my own IRC client in CL.

Looked into an SBCL bug, but it turns out the standard is just kind of ambiguous, and the current behavior is probably fine.


Did the Prolog modules tutorial for the Prolog class. This made me really appreciate Common Lisp's package system... the Prolog module system seems really crufty in comparison.

Fixed a display issue for my blog, but I don't have Hugo set up on this machine yet, so it'll have to wait to be deployed.


Got mutt running on the new Linux machine. The quest to convert my dotfiles fully over to Linux continues.

Got ABCL, ECL, SBCL, and CCL all installed and working on the Linux machine. I had originally installed SBCL (for bootstrapping a new build of it) and StumpWM through the package manager and forgotten about them. This meant I had a /usr/share/common-lisp laying around which was getting loaded for other installs too, which borked a few things. I had to remove the old packages, blow away ~/.cache/common-lisp, and everything works now. ECL is still installed through the package manager, the rest aren't.

Started getting my Lisp test infrastructure up and running on the Linux box. Previously I used figlet and lolcat to print nice headers during the test runs. I don't want to install Ruby on this machine if possible, so I looked for a replacement for lolcat and found toilet, which includes the functionality of both. Great! Had to download the figlet contrib fonts manually, which was annoying (Homebrew on OS X did it automatically, but the Ubuntu package doesn't). Toilet's rainbows aren't as nice as lolcat's, but the tradeoff of not needing ruby is worth it.

Got Hugo installed on the Linux box and made a build of my site. I hope I didn't break anything. Made a couple of tweaks to fix a couple of issues I noticed in Firefox on Linux. Hand-compiled the LessCSS to avoid having to install NodeJS on the machine. Removed the timeago jQuery plugin -- the quest to exterminate all JS on my website continues.

Wrote a blog post. Will post it tomorrow.

Looked into why StumpWM's remove command is making the widths weird. Gotta dive in more when my mind is fresh.


Figured out why StumpWM wasn't noticing my timezone changes. I tracked it down to SBCL itself not noticing the timzeone changes — e.g. you can run (get-decoded-time), change your timezone, and run (get-decoded-time) again and SBCL will still return the old timezone until you restart it. Eventually I narrowed this down to get_timezone in SBCL's time.c, which uses localtime_r. The problem is that localtime and localtime_r don't check for a timezone change, you have to call tzset yourself, according to man 3 localtime. So the solution is to run (cffi:foreign-funcall "tzset") in your StumpWM process whenever you change your timezone.

Watched the Prolog class presentation from a couple days ago. I'm still a little bit behind, gotta hopefully catch up this weekend.


Got notmuch set up on the new box. Forgot how dang fast it is.


Got pass sharing sanely on my phone and desktop.


Installed the Arduino IDE on my desktop. The quest to switch fully to Linux continues.


Started setting up my new Yubikeys. The process is... not friendly. I'm mostly following this guide.


Finished setting up the Yubikeys. I think.

There's still a few fiddly bits -- handling multiple different Yubikeys seems like it's gonna take a bit of scripting grease, and I don't think there's a way to time out the PIN on the Yubikey after a certain amount of inactivity. I think this should mostly be alright, since the only one I'll leave plugged in is the Nano: that's plugged into my desktop monitor that I use for a KVM, so whenever I switch the computers or turn off the monitor it'll lock. I think that's a good enough mix of practicality and security for what I need.


Finally got around to fixing StumpWM's frame splitting/removing/balancing. Got a work-in-progress/proof-of-concept PR at https://github.com/stumpwm/stumpwm/pull/481

Debugged why my scrot keyboard shortcuts that work fine on Debian weren't working on my Ubuntu machine. It took a while because shell commands you run through StumpWM's key mappings have their output blackholed to god only knows where. Eventually I split the commands into a separate shell script and redirected all the output to a file (I should have done this much earlier), which let me see giblib complaining about the keyboard being busy. Once I found that error it led me to https://bbs.archlinux.org/viewtopic.php?id=86507 which shows a solution: add a sleep 0.2 before the call to scrot. Jesus.


Did some coding interview-style programming exercises. I'm rusty at this stuff.

Started writing a blog post. It's turning out to be longer than I expected it would be. By now I should expect this.


Cleaned up my Vim bundle folder. Removed a bunch of things I never use.

Figured out how to do wildcard blocking in uBlock. Again. Here's for my future self: the uBlock rule should look like ##[class^="prefix-"] or ##[class*="part"].

I think I found a setting to eliminate the screen tearing I've been seeing in Ubuntu with my Radeon card. It was described here.

Got Inkscape installed in preparation for setting up my AxiDraw again this weekend. Used the PPA to install the latest version (0.92). EMSL recommends 0.91, but apparently I can use the prerelease version on one file for the AxiDraw (plot_utils.py) and it should work okay. I went ahead and version controlled .config/inkscape/extensions this time for when I inevitably have to dick around with the source again — I learned my lesson last time.


Got stumpish set up so I can poke at stumpwm from within scripts. Hacked my pass binary to pop up the copied message with stumpish so it's a little less opaque when I use the stumpwm shortcut to grab a pass.


Finally got my VPN setup on the Linux machine. Had to disable IPv6 at the router because PIA doesn't support it, and leaving it enabled will leak my actual IP, and I wanted to use OpenVPN instead of PIA's custom app thing. Networking is a mess.

Got Project 1999 set up on Linux. Fonts are a mess, but otherwise it's working fine. Had to add this to ~/.wine/user.reg to get it to stop fucking up my gamma:

[Software\\Wine\\X11 Driver] 1269299093

August 2018


Tried to get Folding at Home working, but after about an hour of screwing around with GPU drivers on Linux I gave up. Sorry, I'd like to help, but I can't invest the time to debug a basic install process.

Set up my own lightweight personal pastebin as outlined here. The hardest part was getting a working pbcopy and pbpaste inside a bash script. I was using xsel which works at my main shell but wouldn't work at all from inside a bash script. I assume this has something to do with $DISPLAY, but then I tried xclip and it worked just fine, so fuck it. Clipboards on Linux are such a mess.


Traveling. Trying to use my second Yubikey on my work machine, and of course it's a fucking mess because why would anyone ever possibly want to use more than ONE single smartcard in their life, right?

The problem is that even though the secret keys for decrypting my pass archives are on this second Yubikey, GPG always wants me to insert the original Yubikey and pops up the "Please insert smartcard with serial number ..." dialog. Even after I run gpg --card-status to forcibly tell its dumb ass to notice the new card.

I tried the usual "delete the secret key and reimport the pubkey, then run --card-status to make it notice" dance, but this time even that didn't work. I ended up having to delete some files by hand: https://donncha.is/2014/07/problems-using-an-openpgp-smartcard-for-ssh-with-gpg-agent/

How is any normal person supposed to actually use this software?


At the office, trying to get my Thinkpad set up again. Linux is a Journey.

Had trouble getting the external monitor to work. Again. First attempt: used a USB-C to MiniDP dongle to a MiniDP to DP cable. It saw the correct resolution, but something in the dongle prevented it from calculating the correct timing, and the monitor wouldn't work.

Second attempt: scavenged a raw USB-C to DP cable. This actually worked.

Also had to scavenge a mouse. Took a while to find the not-bluetooth USB dongle. I checked in all the USB ports on the monitors but didn't see it, til a coworker pointed out the Apple keyboards also have a spare USB port and that's where it was.


Back in Rochester, and GPG is being an asshole once again. Much the same problem as on 8/12 — I'm trying to switch back to my normal Yubikey. The problem:

  • I have two Yubikeys, A and B, which hold my GPG key K.
  • I normally use A.
  • I want to switch to using B.
  • GPG still thinks the private keys for K are stored only on A, even when I plug in B.

The solution is:

  1. Blow away the "keygrip" files in ~/.gnupg/private-keys-v1.d corresponding to the keys on the Yubikeys, but not other keygrip files. An easy way to do this is something like grep -rl shadowed-private-key ~/.gnupg/private-keys-v1.d/ | xargs rm.
  2. Run gpg --card-status so GPG will notice the missing keygrips and realize they're on this new Yubikey.

I should probably wrap this up into a script.


Tad says I need some kind of "udev rule" for my Switch controller under Linux:

KERNEL=="hidraw*", ATTRS{idVendor}=="057e", ATTRS{idProduct}=="2009", MODE="0666"

I don't know what this means, but I'm dumping it in here for later.

Published http://stevelosh.com/blog/2018/08/a-road-to-common-lisp/ after many plane rides and weekends.


Another Ubuntu setup. Practice makes perfect I guess.

Things I needed to apt install to get something usable:

  • arandr
  • aspell-en
  • aspell-is
  • autoconf
  • build-essential
  • chromium-browser
  • cmake
  • curl
  • dunst
  • exfat-fuse
  • exfat-utils
  • fish
  • git
  • gnupg-agent
  • htop
  • hugo
  • inkscape
  • libx11-dev
  • mmv
  • msmtp
  • neomutt
  • neovim
  • network-manager-openvpn-gnome
  • notmuch
  • offlineimap
  • pcscd
  • python-neovim
  • python3-pip
  • restic
  • rlwrap
  • scdaemon
  • scrot
  • slock
  • texinfo
  • tmux
  • toilet
  • tree
  • vlc
  • w3m
  • wget
  • xautolock
  • xcape
  • xclip

Updated my dotfiles bootstrap script to include the new linux symlinks.

Bootstrapping the system is still an uphill fight.


Continuing bootstrapping.

Installing Dropbox makes things a lot easier because I can easily sync little bits of state between computers. But the Dropbox site sure doesn't make it easy. To my future self: here's how to install Dropbox on Ubuntu:

  • Download the .deb file from their site.
  • sudo dpkg -i thefile.deb
  • dropbox start -i

Important: do not run the last command with sudo, because if you do your entire installation will be totally fucked (~/Dropbox will be owned by root) and you'll have to start all over.

Building Mercurial from source. Had to install python-dev first.

As always, hg-git is fucking broken on install. Had to symlink dulwich/dulwich into the Mercurial directory, but it was still broken. This is the problem. Mercurial's lack of a stable plugin API is why I no longer really maintain my hg plugins. It sucks. For now I'm just gonna downgrade Mercurial.

September 2018

I quit my job at the end of August and spent September traveling and not thinking about computers.

October 2018

I spent the first three weeks of October in SF at my new job, to onboard in meatspace before starting remote work.


Continuing game jam work.


More work on the Lisp Game Jam. I don't think my entry will be ready in time, but if nothing else at least I've accomplished a couple of things:

  • Done some actual testing of Mansion. I really need to come up with a more stringy syntax, and also start thinking about how to define colors (just use constants?).
  • Figured out how to use esrap. It's a little clumsy, but better than doing everything by hand. I still like Smug better, I think, though it has its own clumsiness.


Finally got back to the electronics book. Learned about 555 timer chips.

November 2018


Did a couple of Rosalind problems after work:

  • cl-ppcre's all-matches only returns non-overlapping matches, not all matches. Had to write my own. Dammit.
  • ~* will skip over format arguments.
  • Drakma follows HTTP redirects by default, which is nice.
  • Uniprot is slow until you warm up their caches.


Added some support for arbitrary data to CACL. After using jq and being miserable, I think it might be possible to do RPN JSON processing. (Re)learned how to use the CL pretty printer in the process, from the CLR book, so I can print hash tables much more nicely now.


Changed the battery in my Thinkpad. Mostly straightforward, except that the original was missing some screws and one of the plastic tabs had broken off and was jamming things when I tried to reassemble it. Once I figured that out it went back together fine.

Did some more Rosalind problems. One was trivial. Implemented longest-monotonic-subsequence for another, which was tricky.


Got the backlight control working in my Thinkpad. Had to edit /etc/X11/xorg.conf.d/20-intel.conf to contain:

Section "Device"
    Identifier  "Intel Graphics" 
    Driver      "intel"
    Option      "Backlight"  "intel_backlight"

Then reboot, and xbacklight would finally work.

Played some DCSS for the first time in a long while. Still fun.