Goodbye Roswell

Roswell is a multi-platform Lisp installer and launcher; if the reader is familiar with Clojure it could be seen roughly as the equivalent of Leiningen. Yesterday I've decided to say goodbye and thank for all the fish to Roswell and proceeded to remove it and install SBCL the old trusted way, directly from source.

Roswell has some neat offerings, like installing with a single command a Lisp implementation like CCL or SBCL, download and configure Quicklisp or build an executable from a project.

When the installation works.

In my case, more than a single time, when I've tried to upgrade it I ended up with some errors. Nothing that I couldn't solve with some pronging, but that elided the usefulness of the tool¹.

The biggest problem I've found was the installation from scratch. Quite often it ended with a somewhat strange error in a bizarre place of the system, the hostname-util.c file, or in other words, the fail pointed to the famous systemd.

The bane of IT

Well, not exactly.

As I've said before, contrary to the loud majority I'm OK with systemd. Frankly, creating an init system better than a bunch of shell scripts isn't placing a bar really high.

Anyway, this is the error message:

$ Assertion 'uname(&u) >= 0' failed at ../src/basic/hostname-util.c:55, function gethostname_malloc(). Aborting.

I was a little puzzled because there aren't a lots of thing that could go wrong in the gethostname_malloc() function, unless the error is caused by lack of RAM. To deepen the mystery, on the same system I can build SBCL inside a tmpfs partition², so RAM is aplenty.

I've looked at the bug reports and found at least one with the same error message and probably digging deep enough I could have isolated the problem, but sometimes the pragmatism wins. In this case, installing it from sources didn't create a lot of friction, unfortunately this means that deploying has became more complicated.

Installation Notes

I decided to put the SBCL installation in my home directory. For the compilation it requires an ANSI compliant Common Lisp implementation, other than that I don't usually build it with --fancy features enabled as I stick with the defaults:

$ mkdir -p ~/.local/share/sbcl
$ sh ./make.sh
$ INSTALL_ROOT=~/.local/share/sbcl sh ./install.sh

I'm using Zsh so I've added the binaries path in my .zshenv file:

# add SBCL common lisp to PATH
if [[ -d "$HOME/.local/share/sbcl/bin" ]]; then
    export PATH="$HOME/.local/share/sbcl/bin:$PATH"
    export SBCL_HOME="$HOME/.local/share/sbcl/lib/sbcl"
fi

If it's not already installed, I then proceed to download and configure Quicklisp, that means adding my local project directory path in the ~/.sbcl.rc file:

#+quicklisp
(progn
  (pushnew #P"/path/to/sources/lisp/"
           ql:*local-project-directories*))

Profit?

Edit

The Stow version of the instructions above became:

$ sh ./make.sh --with-sb-core-compression
$ SBCL_HOME="" INSTALL_ROOT=/usr/local/stow/$(basename $(pwd)) sh ./install.sh
$ stow -D sbcl-currently-installed-version -S $(basename $(pwd))

where --with-sb-core-compression is applied to add zlib compression for core images and $(basename $(pwd)) usually expands to sbcl-version-number, if upstream doesn't change the naming format of their archives. Invoking Stow from the current directory should work when configured properly.³

Hiding Emacs' Mouse Pointer

Greetings folks. Another afternoon spent shaving yaks.

Out of the box Emacs is configured to hide the mouse pointer when the keyboard is used, a sane default if you ask me, with only a small detail in its current implementation: the pointer is hidden only when is typed a self-inserting-character. In other words, moving the cursor with the keyboard doesn't hide the mouse pointer.

And in my case, the mouse pointer always lands where I want to read.

I've checked the available documentation and I can't find anything that can help me, so it seems that I need to hack the sources. So the other day I've tried to modify the move_point¹ function in cmds.c with a tentative patch:

diff -u emacs-25.3/src/cmds.c emacs-25.3-mod/src/cmds.c
--- emacs-25.3/src/cmds.c
+++ emacs-25.3-mod/src/cmds.c
@@ -71,6 +71,8 @@
       xsignal0 (Qend_of_buffer);
     }

+  frame_make_pointer_invisible (SELECTED_FRAME ());
+
   SET_PT (new_point);
   return Qnil;
 }

Yay! It worked after the first attempt.

Well, it's not the cleanest solution as I've introduced at least one side effect: when used in combination with Gnome's windows sloppy focus mode² the window focus event hide the pointer without touching the keyboard. For the moment it doesn't bother me and it's good enough, probably I need to put a condition to not hide the pointer if it's an X focus event.

SSH keys precedence

Greetings readers.

Summer is officially over, so I can finally bear to be in front of a computer monitor without the help of an AC unit, the thunderstorm of the last week lowered the temperatures to a human level so going outside for most of the day is pleasurable, if one can find a place not infested by mosquitoes.

Enough ramblings, today¹ I've verified that the OpenSSH client doesn't try keys in alphabetical order when attempting public/private keys logins, but instead tries them according to their timestamp, starting with the oldest.

Good to know.

This reminds me a Gerald Sussman quote²:

Nowadays you muck around with incomprehensible or nonexistent man pages for software you don't know who wrote. You have to do basic science on your libraries to see how they work, trying out different inputs and seeing how the code reacts.

I should be honest, I haven't checked thoroughly the OpenSSH man pages, anyway those words resonated with me.