| Age | Commit message (Collapse) | Author |
|
It works better on the NC10 because the touchpad is small, but on the
Dell 3190 the edge is extremely narrow, and the touchpad itself is big
enough to make two-finger scroll bearable.
|
|
|
|
Replace .md/.org extension with .html.
The output of "make" is now ugly as sin.
|
|
|
|
|
|
|
|
Maybe not the best idea, since the dependency chain will trigger a
site-wide rebuild everytime the README is edited. Ah well.
|
|
|
|
|
|
|
|
Still a few wrinkles to iron out, e.g. the toplevel page's <title>.
|
|
|
|
|
|
Seems kind of wasteful to recompute it every time. Also, now
subsequent scripts don't have to care about file extensions anymore.
|
|
Make generate-deps.py compute the list of indices instead of relying
on the list of source folders that contain text files, otherwise we
will miss intermediate folders that do not contain any file.
Remove TODO entry to maintain Makefile dependencies to scripts: that
sounds too tedious. Let's assume that at some point the Makefile and
these scripts will be bundled together into a proper package.
|
|
If an intermediate folder contains no leaf page, the previous code
does not add it to the parent directory's subfolders.
Still needs support from generate-deps.py.
|
|
It's annoying to have to spell out "apt-get", and I don't want to
configure APT not to use the fancy progress bar because I happen to
like it.
Alternative approach: https://oremacs.com/2019/03/24/shell-apt/
|
|
|
|
I had spent the better part of a day composing a 50-line message to
help-gnu-emacs asking for advice, and I was a heartbeat away from
hitting C-c C-c, until I began wondering if the list returned by
alist-get was a copy or a reference.
Turns out it's a reference. Remembered that setcar is a thing. Whipped
up the following:
(let* ((indicator (alist-get 'compilation-in-progress mode-line-modes))
(old-props (text-properties-at 0 (car indicator)))
(new-props '(face compilation-mode-line-run)))
(setcar indicator (apply #'propertize "⚙" (append new-props old-props))))
|
|
|
|
Also don't run pandoc in a pipeline within a Makefile recipe: pipefail
is not set, so generate-index.py failures will not be bubbled up.
|
|
|
|
|
|
I dimly remember that it was interesting, but I don't seem to have
taken any notes. Oh well.
|
|
So that I can re-use generate-index.py for READMEs.
|
|
This way I can concatenate the output with regular indices.
|
|
So that I can add a special recipe for them, where I'll concatenate
the directory index.
|
|
|
|
|
|
|
|
|
|
It's not clear whether or not Python supports '\0' for null: the
reference[1] says nothing about this specific escape sequence, but
Python 3.7 seems to recognize it…
repr('\0') says "'\\x00'", so let's use that.
[1]: https://docs.python.org/3/reference/lexical_analysis.html#string-and-bytes-literals
|
|
|
|
|
|
|
|
I prefer thinking of slashes as concatenation operators, always
accompanied by operands on both sides. I don't like them hanging
there at the end of my variables, which is what $(dir …) causes.
|
|
"Uppercase = public API, lowercase = internal plumbing".
|
|
|
|
Since it already computes the list of pages.
I'm not comfortable with the pattern substitution, the implicit "slash
management" is not obvious.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
By giving them meaningful names and adding a master script to call
them all.
|
|
|
|
|
|
This folder will be used to hold all website-building code.
At least until I change my mind and figure it's better to do it
otherwise.
|
|
|