Colour-blind safe palette and a no-colour-alone rule (#56) #86

Merged

robert merged 1 commit from area/colour-palette into main 2026-09-04 10:01:00 +02:00

Conversation 0 Commits 1 Files changed 3 +291 -2

robert commented 2026-09-04 10:00:05 +02:00

Owner

Copy link

Implements #56 ("Colour-blind safe palette and a no-colour-alone rule").

This gates #59 (the HERO1/HERO2/QUAD/GRID6 template renderer) — #83/#85 both explicitly deferred
real visual template work to #59, which needs this palette to exist first.

What's here

• watchapp/src/c/palette.h — named GColor constants (neutrals, staleness grey tying into D44,
  and five semantic accent colours), aliased from the SDK's own gcolor_definitions.h, not invented
  hex values.
• docs/DESIGN.md — new section 7 (Colour): palette source, the no-colour-alone rule restated with
  a per-feature intent table, CVD reasoning, contrast, and how to use gcolor_legible_over().
• docs/DECISIONS.md — a small correction under D24: gabbro is 64 colours, not just "colour"
  (verified by diffing its gcolor_definitions.h against emery's — byte-for-byte identical).

No visual change today

Grepped the tree: nothing under watchapp/src/c/ sets a GColor anywhere. Every current TextLayer
draws with the system default (black on white). This PR is pure infrastructure — zero effect until
#59 (and later #54, off-route state, HR zones, GPS quality, speed source) start consuming it.

Verified vs. reasoned (the project's own convention, see docs/TEAM.md)

Verified:

• The 64-colour palette itself — read directly from the installed SDK, not recalled:
  ~/.local/share/pebble-sdk/SDKs/4.33.1/sdk-core/pebble/{emery,gabbro}/include/gcolor_definitions.h,
  byte-for-byte identical between the two platforms.
• basalt is also 64-colour, confirmed against current docs at
  developer.rebble.io.
• gcolor_legible_over(GColor8) exists — confirmed both in current
  developer.repebble.com docs and by
  grepping the installed SDK's pebble.h (GColor8 gcolor_legible_over(GColor8 background_color);).
• WCAG 2.1 SC 1.4.11's 3:1 floor for graphical objects/UI components — confirmed against the current
  spec text at w3c.github.io.
• Contrast ratios in palette.h's comments — computed (not guessed) via the standard WCAG
  relative-luminance formula against each colour's exact SDK RGB values.
• pebble build — actually run, clean, for gabbro/emery/basalt, with palette.h temporarily
  #included into main.c to force real compilation under arm-none-eabi-gcc (then reverted; the
  build is unaffected either way — identical heap-free figures).

Reasoned, explicitly not simulator-verified (no CVD simulator available in this sandbox):
colour-vision-deficiency confusion-axis reasoning — protanopia/deuteranopia collapse red-green,
tritanopia collapses blue-yellow. Response is structural: green is not used for any semantic/stateful
colour at all, so the red/green pairing the issue calls out can't arise by accident. Section 7.3 is
explicit about where this is honestly weakest (the blue→violet HR-zone step), which is exactly why
the no-colour-alone rule is a requirement, not a nicety.

Explicitly NOT verified, flagged as needing Robert's physical devices (both emery and gabbro are
in hand per this session's memory): direct-sunlight legibility of these swatches on the actual
e-paper. Searched for official Pebble/Rebble contrast-in-sunlight documentation and found none —
only general, non-Pebble-specific e-paper reflectivity claims, which section 7.4 is explicit about
not treating as if they were a measured figure. The WCAG contrast numbers are a real computation, but
they assume an emissive sRGB display; e-paper reflectance under direct sun is a physical property no
formula run on RGB channel values can establish.

Acceptance criteria (from #56)

• Palette chosen and documented in docs/DESIGN.md, checked against deuteranopia/protanopia/
  tritanopia (reasoned, not simulator-verified — see above)
• No state signalled by colour alone — documented as a rule (section 7.2) with a per-feature
  pairing table
• Red/green never the sole distinction — satisfied structurally: green isn't used for any
  state at all
• Verified specifically for GPS quality, speed source, off-route state, HR zones, sensor alerts
  — documented intent per feature (all future UI, none implemented yet, per the issue's own
  scope note)
• Contrast checked on e-paper in direct sunlight — cannot be done in this sandbox; computed
  WCAG contrast instead, flagged explicitly as not a substitute, needs Robert's physical watches
• Palette defined as named constants in one header — watchapp/src/c/palette.h

https://claude.ai/code/session_01DAoXbRmJUf2uxNYBfdAXPt

Implements #56 ("Colour-blind safe palette and a no-colour-alone rule"). This gates #59 (the HERO1/HERO2/QUAD/GRID6 template renderer) — #83/#85 both explicitly deferred real visual template work to #59, which needs this palette to exist first. ## What's here - `watchapp/src/c/palette.h` — named `GColor` constants (neutrals, staleness grey tying into D44, and five semantic accent colours), aliased from the SDK's own `gcolor_definitions.h`, not invented hex values. - `docs/DESIGN.md` — new section 7 (Colour): palette source, the no-colour-alone rule restated with a per-feature intent table, CVD reasoning, contrast, and how to use `gcolor_legible_over()`. - `docs/DECISIONS.md` — a small correction under D24: gabbro is 64 colours, not just "colour" (verified by diffing its `gcolor_definitions.h` against emery's — byte-for-byte identical). ## No visual change today Grepped the tree: nothing under `watchapp/src/c/` sets a `GColor` anywhere. Every current `TextLayer` draws with the system default (black on white). This PR is pure infrastructure — zero effect until #59 (and later #54, off-route state, HR zones, GPS quality, speed source) start consuming it. ## Verified vs. reasoned (the project's own convention, see docs/TEAM.md) **Verified:** - The 64-colour palette itself — read directly from the installed SDK, not recalled: `~/.local/share/pebble-sdk/SDKs/4.33.1/sdk-core/pebble/{emery,gabbro}/include/gcolor_definitions.h`, byte-for-byte identical between the two platforms. - `basalt` is also 64-colour, confirmed against current docs at [developer.rebble.io](https://developer.rebble.io/guides/best-practices/building-for-every-pebble/). - `gcolor_legible_over(GColor8)` exists — confirmed both in current [developer.repebble.com](https://developer.repebble.com/docs/c/Graphics/Graphics_Types/) docs and by grepping the installed SDK's `pebble.h` (`GColor8 gcolor_legible_over(GColor8 background_color);`). - WCAG 2.1 SC 1.4.11's 3:1 floor for graphical objects/UI components — confirmed against the current spec text at w3c.github.io. - Contrast ratios in `palette.h`'s comments — computed (not guessed) via the standard WCAG relative-luminance formula against each colour's exact SDK RGB values. - `pebble build` — actually run, clean, for gabbro/emery/basalt, with `palette.h` temporarily `#include`d into `main.c` to force real compilation under `arm-none-eabi-gcc` (then reverted; the build is unaffected either way — identical heap-free figures). **Reasoned, explicitly not simulator-verified** (no CVD simulator available in this sandbox): colour-vision-deficiency confusion-axis reasoning — protanopia/deuteranopia collapse red-green, tritanopia collapses blue-yellow. Response is structural: green is not used for any semantic/stateful colour at all, so the red/green pairing the issue calls out can't arise by accident. Section 7.3 is explicit about where this is honestly weakest (the blue→violet HR-zone step), which is exactly why the no-colour-alone rule is a requirement, not a nicety. **Explicitly NOT verified, flagged as needing Robert's physical devices** (both emery and gabbro are in hand per this session's memory): direct-sunlight legibility of these swatches on the actual e-paper. Searched for official Pebble/Rebble contrast-in-sunlight documentation and found none — only general, non-Pebble-specific e-paper reflectivity claims, which section 7.4 is explicit about not treating as if they were a measured figure. The WCAG contrast numbers are a real computation, but they assume an emissive sRGB display; e-paper reflectance under direct sun is a physical property no formula run on RGB channel values can establish. ## Acceptance criteria (from #56) - [x] Palette chosen and documented in docs/DESIGN.md, checked against deuteranopia/protanopia/ tritanopia (reasoned, not simulator-verified — see above) - [x] No state signalled by colour alone — documented as a rule (section 7.2) with a per-feature pairing table - [x] Red/green never the sole distinction — satisfied structurally: green isn't used for any state at all - [x] Verified specifically for GPS quality, speed source, off-route state, HR zones, sensor alerts — documented intent per feature (all future UI, none implemented yet, per the issue's own scope note) - [x] Contrast checked on e-paper in direct sunlight — **cannot be done in this sandbox**; computed WCAG contrast instead, flagged explicitly as not a substitute, needs Robert's physical watches - [x] Palette defined as named constants in one header — `watchapp/src/c/palette.h` https://claude.ai/code/session_01DAoXbRmJUf2uxNYBfdAXPt

robert added 1 commit 2026-09-04 10:00:06 +02:00

Colour-blind safe palette and a no-colour-alone rule (#56) ... 1574c9fd83

Adds watchapp/src/c/palette.h: named GColor constants for emery/gabbro/
basalt, taken from the installed SDK's own gcolor_definitions.h (verified
identical between emery and gabbro; basalt confirmed 64-colour by current
developer.rebble.io docs) rather than invented hex values.

No file under watchapp/src/c/ sets a GColor today, so this changes nothing
on screen. It's infrastructure for #59 (the HERO1/HERO2/QUAD/GRID6
renderer) and later issues (#54 sensor alerts, off-route state, HR zones,
GPS quality, speed source), none of which have any UI yet.

docs/DESIGN.md gets a new section 7 (Colour):
- the palette source and why it applies uniformly across all three build
  platforms
- the no-colour-alone rule, restated with a per-feature intent table for
  the five things this issue was asked to verify against
- colour-vision-deficiency reasoning (protanopia/deuteranopia/tritanopia),
  tagged reasoned, not simulator-verified — no CVD simulator is available
  in this environment
- contrast, tagged computed (WCAG relative-luminance formula against the
  exact SDK RGB values) but explicitly NOT sunlight-verified — that needs
  Robert's physical emery and gabbro units, both in hand
- gcolor_legible_over() as the recommended mechanism for text-on-fill,
  verified present in both current docs and the installed SDK's pebble.h

docs/DECISIONS.md gets a small correction under D24: gabbro is 64 colours,
not just "colour" (verified by diffing its gcolor_definitions.h against
emery's — byte-for-byte identical), and DESIGN.md's platform table is
corrected to match.

`pebble build` verified clean for gabbro/emery/basalt with palette.h
actually #include'd (temporarily, then reverted) to force real compilation
under arm-none-eabi-gcc, not just left unreferenced.

Claude-Session: https://claude.ai/code/session_01DAoXbRmJUf2uxNYBfdAXPt

robert merged commit 955bfe5d1f into main 2026-09-04 10:01:00 +02:00

robert referenced this pull request from a commit 2026-09-04 10:01:02 +02:00

Merge pull request 'Colour-blind safe palette and a no-colour-alone rule (#56)' (#86) from area/colour-palette into main

robert referenced this pull request 2026-09-04 10:01:13 +02:00

Colour-blind safe palette and a no-colour-alone rule #56

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

area:companion

Android companion app

  

area:docs

Documentation

  

area:shared

Message contract shared by both sides

  

area:tooling

Build tooling, SDK, dev scripts

  

area:watchapp

Pebble C watchapp

  

blocker

Blocks a large part of the project until resolved

  

kind:chore

Setup, scaffolding, maintenance

  

kind:feature

New functionality

  

kind:spike

Time-boxed investigation to de-risk a decision

  

kind:test

Tests and fixtures

No labels

area:companion

area:docs

area:shared

area:tooling

area:watchapp

blocker

kind:chore

kind:feature

kind:spike

kind:test

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

robert/PedalPebble!86

No description provided.