Codify the canonical C style and add reindent tooling

AGENTS.md described style in one paragraph that said little more than
"follow the surrounding code", which was not actionable once files had
drifted apart. Replace it with an explicit specification.

The canonical style is Emacs cc-mode "stroustrup" with tabs enabled: a
4-column offset with 8-column tabs, so depth 1 is four spaces, depth 2 is
one tab, depth 3 is a tab plus four spaces. Most of src/ already followed
this; what looked like randomly mixed tabs and spaces was simply cc-mode
output. Document the byte ladder explicitly, since editors that expand
tabs or assume a 4-column tab silently corrupt it.

Rules beyond indentation are derived from counted majorities in the
existing code rather than invented: errctx over e (92 to 45), padded
control parens (140 to 7), pointer binding to the identifier (486 to 5),
and always-brace (only four unbraced bodies exist). Brace and else
placement are called out as house conventions that cc-mode does not
enforce, so "run the indenter" and "follow the guide" cannot conflict.

Also record the naming, error-handling, and API-surface rules that a
consistency sweep showed were being broken silently -- in particular that
a *_RETURN macro inside an ATTEMPT block skips CLEANUP.

Add the tooling to apply it:

  .dir-locals.el          applies the style to every c-mode buffer
  scripts/reindent.el     batch reindent via Emacs
  scripts/reindent.sh     reindent or --check the tree
  scripts/hooks/pre-commit  reindents staged sources

reindent.el deliberately avoids Emacs' tabify: its tabify-regexp is
" [ \t]+", which is not anchored to the start of a line, so it rewrites
runs of spaces anywhere and destroys the hand-aligned value columns in
the bit-flag tables in actor.h and iterator.h. Only leading whitespace is
ever rewritten, and lines starting inside a string literal are skipped.

The hook checks staged content rather than the working tree, so what is
committed is what was verified. It re-stages a fixed file only when the
index and working tree agree, otherwise a partial `git add -p` would
sweep unstaged work into the commit.

Enable with: git config core.hooksPath scripts/hooks

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-30 18:17:26 -04:00
parent bc782fbffe
commit 8d21d5c7dd
5 changed files with 549 additions and 3 deletions

87
scripts/reindent.el Normal file
View File

@@ -0,0 +1,87 @@
;;; reindent.el --- batch reindent to cc-mode "stroustrup" -*- lexical-binding: t -*-
;; Reindents each file named on the command line, in place, to the project's
;; canonical style: cc-mode "stroustrup", 4 columns per level, tabs 8 columns
;; wide. See AGENTS.md -> "Coding Style".
;;
;; The style is set explicitly here rather than read from .dir-locals.el so that
;; the result does not depend on where the file lives -- the pre-commit hook
;; runs this over temporary copies outside the project tree.
;;
;; Usage: emacs --batch -Q -l scripts/reindent.el -- FILE...
;;; Code:
(require 'cc-mode)
(setq make-backup-files nil
create-lockfiles nil
auto-save-default nil
vc-handled-backends nil
inhibit-message t)
(defun akgl-retab-leading-whitespace ()
"Rewrite every line's leading whitespace as tabs-then-spaces at `tab-width'.
`indent-region' fixes the column a line starts at, but `indent-line-to' leaves
a line alone when it is already at the right column even if the bytes are
wrong -- eight spaces where the canonical form is one tab. This pass closes
that gap.
Deliberately not `tabify': that function's `tabify-regexp' is \" [ \\t]+\",
which is NOT anchored to the line start, so it rewrites runs of spaces
anywhere on the line and destroys the hand-aligned columns in the bit-flag
tables in actor.h and iterator.h. Only leading whitespace is touched here, and
lines beginning inside a string literal are skipped outright."
(goto-char (point-min))
(while (not (eobp))
(let ((bol (point)))
(unless (nth 3 (syntax-ppss bol))
(skip-chars-forward " \t" (line-end-position))
(let ((col (current-column)))
(unless (or (zerop col) (eolp))
(let ((want (concat (make-string (/ col tab-width) ?\t)
(make-string (% col tab-width) ?\s))))
(unless (string= want (buffer-substring bol (point)))
(delete-region bol (point))
(insert want)))))))
(forward-line 1)))
(defun akgl-reindent-file (path)
"Reindent PATH in place. Returns t if the file changed on disk."
(let ((before (with-temp-buffer
(insert-file-contents path)
(buffer-string))))
(with-current-buffer (find-file-noselect path t)
(c-mode)
(c-set-style "stroustrup")
(setq indent-tabs-mode t
tab-width 8
c-basic-offset 4
require-final-newline t)
;; 1. Put every line at its correct column.
(indent-region (point-min) (point-max))
;; 2. Convert leading whitespace to the canonical tab/space mix.
(akgl-retab-leading-whitespace)
;; 3. Trailing whitespace and a single final newline.
(delete-trailing-whitespace)
(goto-char (point-max))
(unless (bolp) (insert "\n"))
(let ((changed (not (string= before (buffer-string)))))
(when changed (save-buffer))
(kill-buffer)
changed))))
;; Emacs leaves the "--" separator in `command-line-args-left'; drop it, along
;; with any empty argument, so the remainder is exactly the file list.
(dolist (path (seq-remove (lambda (a) (or (string= a "--") (string= a "")))
command-line-args-left))
(when (akgl-reindent-file path)
(princ (format "reindented %s\n" path))))
;; Exit 0 on success whether or not anything changed, so that any non-zero
;; status from this script means a real failure. Callers detect "something
;; changed" from stdout, or by comparing files themselves.
(kill-emacs 0)
;;; reindent.el ends here