max / makeover-build

Strip historical narrative from documentation Remove what a doc used to say, when it changed, the incidents that justified a rule, finished migration narration, and counts and versions that rot. State the rules in the present tense instead. Keep every instruction, prohibition and threshold, and keep the measurements that make a rule actionable. Public-facing docs keep their explanatory voice.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MptwXZ8k65v19rFmdGAyki

Author: Max Johnson <me@maxj.phd> · 2026-08-31 01:51 UTC

Signed with PGP, not checked

Commit: de77ad5e5a679c839e8fbe37698b7877626bc023

Parent: 0b05e47

3 files changed, +10 insertions, -24 deletions

M README.md +4 -8

			@@ -2,11 +2,8 @@
2	2
3	3		Build-script support for the make-family design system.
4	4
5		-	Every consumer materialises the same generated files from a `build.rs`, and
6		-	until now every consumer wrote that code itself. GoingsOn and Balanced Breakfast
7		-	grew byte-identical copies of the theme materialiser during the makeover-geometry
8		-	adoption, and the layout stylesheet would have been the third and fourth copies.
9		-	This is that code, once.
	5	+	Every consumer materialises the same generated files from a `build.rs`. This is
	6	+	that code, once, so no consumer keeps its own copy.
10	7
11	8		```rust
12	9		// build.rs
			@@ -40,9 +37,8 @@
40	37		drifts. Every path written here is expected to be gitignored.
41	38
42	39		`themes` clears stale `.toml` files before writing, so a theme removed or
43		-	renamed upstream does not linger in the bundle from an earlier build. That is
44		-	the detail that makes this worth sharing rather than retyping: it is easy to
45		-	omit, and its absence shows up as a theme that will not go away.
	40	+	renamed upstream does not linger in the bundle from a previous build. Omitting
	41	+	that step shows up as a theme that will not go away.
46	42
47	43		## Status
48	44

M src/drift.rs +3 -10

			@@ -409,9 +409,7 @@
409	409		/// specificity, either by being unlayered or by sitting in a layer the app's
410	410		/// order statement puts after `makeover`, so an app declaration for a property
411	411		/// makeover already sets does not merge with it: it wins, silently, and the
412		-	/// design system's version of that component stops applying. Both sort-caret
413		-	/// defects found on 2026-08-11 were this, and both were live for months because
414		-	/// nothing looked.
	412	+	/// design system's version of that component stops applying, and nothing looks.
415	413		///
416	414		/// # Two passes, because a rule can carry no class
417	415		///
			@@ -460,13 +458,8 @@
460	458		/// that guesses wrong about specificity fails correct builds -- so the app
461	459		/// declares it instead, the same shape as quasi-webview's `RENDERER_OWN`.
462	460		///
463		-	/// The example this section used to give was the sort caret: `content` on
464		-	/// `.table-heading`'s unsorted arm reserved the gap, `content` on the sorted arm
465		-	/// was the generated glyph, and the two were a pairing rather than a collision.
466		-	/// makeover-webview 0.31.0 emits both arms itself, so that pair is now an app
467		-	/// overriding the caret and the check is right to fail it. A reviewed pairing is
468		-	/// a claim about who owns a property, and it expires when the design system
469		-	/// takes the property back.
	461	+	/// A reviewed pairing is a claim about who owns a property, and it expires when
	462	+	/// the design system takes the property back.
470	463		///
471	464		/// `allowed_elements` is the same thing one pass down: `(element, class,
472	465		/// property)` triples where a bare element rule reaching a generated class has

M src/lib.rs +3 -6

			@@ -71,9 +71,8 @@
71	71		/// Write the themes `makeover` ships into `dir`, as `<id>.toml`.
72	72		///
73	73		/// Clears stale `.toml` files first, so a theme removed or renamed upstream
74		-	/// does not linger in the bundle from an earlier build. That detail is the
75		-	/// reason this is worth sharing rather than retyping: it is easy to omit and
76		-	/// its absence shows up as a theme that will not go away.
	74	+	/// does not linger in the bundle from a previous build. Omitting that step
	75	+	/// shows up as a theme that will not go away.
77	76		///
78	77		/// # Panics
79	78		///
			@@ -123,9 +122,7 @@
123	122		/// [`makeover_geometry::size_class_css`] follows the density block because it
124	123		/// is the narrower claim: density says what is pointing at the screen, size
125	124		/// class says how much screen there is, and on a compact window the two shells
126		-	/// tighten regardless of which density selected them. Shipped in
127		-	/// makeover-geometry 0.7.0 and emitted by nobody until 2026-08-10, which meant
128		-	/// the axis existed in the crate and reached no stylesheet.
	125	+	/// tighten regardless of which density selected them.
129	126		///
130	127		/// # Panics
131	128		///