Known issues ============ * `man(1)` completion doesn't work on OpenBSD as `manpath(1)` isn't a thing on that system; need to find some way of finding which manual directories should be searched at runtime, if there is one. * The checks `gscr(1df)` makes to determine where it is are a bit naïve (don't work with bare repos) and could probably be improved with some appropriate `git-reflog(1)` calls * `dr(6df)` is probably more practical in awk * Running the block of git(1) commands in the prompt leaves five "stale" jobspecs around that flee after a jobs builtin run; only saw this manifest after `90dcadf`; either I understand job specs really poorly or this may be a bug in bash * I can't find a clean way of detecting a restricted shell for ksh instances to prevent trying to load anything fancy (works for Bash) * Zsh, either! $options[restricted] is "off" within the startup file * Would be good to complete the Makefile variables for NAME, EMAIL etc with educated guesses (`id -u`@`cat /etc/mailname`) etc rather than hardcoding my own stuff in there * Need to decide whether I care about XDG, and implement it if I do * Need to decide whether I'm testing the shell snippets for MPD, Keychain etc, and if so how. * The custom shell functions really should be documented, but it's not clear to me exactly where this should happen, because the commands' availability depends on which shell you're using; the `sd` function isn't available when you're not using Bash. Maybe I should try to extend `help` without breaking it? * The `b:undo\_indent` definition for the `perl` filetype can probably be pushed upstream. * The `_text_filenames` completion handler for Bash won't work on files with newlines in their names. Can it be made to? * Typing the normal mode mapping for `paste_open.vim` *twice* causes an error.