Beta to main

.github/workflows/code-quality.yml

@@ -170,7 +170,7 @@ jobs:
       # entirely on the version openregister cannot load.
       #
       # EVERY MAJOR IN THE DECLARED RANGE, NOT JUST ITS ENDS. appinfo/info.xml declares
-      # <nextcloud min-version="32" max-version="34"/>, and
+      # <nextcloud min-version="32" max-version="35"/>, and
       # tests/Unit/ClaimAccuracyTest::testDeclaredNextcloudRangeMatchesTheTestedMatrix
       # asserts that the lowest ref here EQUALS that floor. A matrix of stable34
       # alone therefore advertises 32 and 33 to the app store while exercising
@@ -187,14 +187,20 @@ jobs:
       # in the LOOSER direction, which this fleet does not take. The floor is still
       # 32, so ClaimAccuracyTest's lowest-ref-equals-floor assertion is unaffected.
       #
-      # stable34 is FIRST deliberately. Playwright is the job that reads [0], and
+      # stable35 is FIRST deliberately, and moved there when max-version rose to
+      # 35. Playwright is the job that reads [0], and
       # tests/e2e/spec-coverage/selector-liveness.spec.ts is a cross-version
       # stylesheet audit whose SINCE deferrals expire only once the survey
-      # reaches the newest supported major (MAX_SUPPORTED_NC). Surveying 32 would
-      # defer the NC33+ header selectors indefinitely; surveying 34 forces them
-      # to be proven live — which is how the eight dead NC 32/33-only selectors
-      # now recorded in that file's ALLOWED list were found.
-      nextcloud-test-refs: '["stable34", "stable32", "stable33"]'
+      # reaches the newest supported major (MAX_SUPPORTED_NC, a constant in that
+      # spec file that mirrors info.xml BY HAND — it still reads 34, so raising
+      # it is a separate decision: it expires that file's SINCE deferrals).
+      # Surveying 32 would defer the NC33+ header selectors
+      # indefinitely; surveying the newest major forces them to be proven live —
+      # which is how the eight dead NC 32/33-only selectors now recorded in that
+      # file's ALLOWED list were found. Leaving stable34 in front of a declared
+      # 35 would put the survey one major behind the claim and start that drift
+      # over again.
+      nextcloud-test-refs: '["stable35", "stable32", "stable33", "stable34"]'
       # These three were `false`, and that is worse than it sounds. The shared
       # workflow builds one matrix leg per tool unconditionally; a disabled leg
       # prints "<tool> is disabled — skipping." and `exit 0`. The job therefore
@@ -299,6 +305,26 @@ jobs:
       # recursively.
       enable-newman: true
 
+      # ── Licensing ────────────────────────────────────────────────────────
+      # REUSE findings fail the build. The shared workflow defaults this OFF
+      # because most fleet apps ship no REUSE.toml or LICENSES/ directory;
+      # thematiq now ships both and lints at 3680/3680 with 0 missing, 0
+      # unused and 0 invalid expressions, compliant with version 3.3 of the
+      # REUSE specification.
+      #
+      # It stayed off while the repo was NOT compliant, deliberately: turning
+      # it on then would have failed every open PR for a debt none of them
+      # introduced. Now that the tree is clean, off is the worse default —
+      # non-blocking surfaces a regression only as REUSE ❌ in the Quality
+      # Report comment, which is easy to miss.
+      #
+      # The blanket `path = "**"` annotation in REUSE.toml makes every new file
+      # compliant on arrival, whatever its type, so ordinary work cannot trip
+      # this. Two regression modes remain: vendoring a file whose own SPDX
+      # header names a licence with no text in LICENSES/, and writing prose
+      # that QUOTES an SPDX tag outside a REUSE-IgnoreStart/End pair.
+      reuse-blocking: true
+
       # ── Frontend Check legs ──────────────────────────────────────────────
       # `frontend-checks` defaults to `[]`, and an empty list means the shared
       # workflow emits NO "Frontend Check" job at all — so these two validators
@@ -397,3 +423,14 @@ jobs:
       # construction — <= v1.4.0 passed over it (permanently green) and v1.5.0
       # refused with exit 99 (permanently red). The scope is now
       # `github.event.before...HEAD`, what the push actually changed.
+      # ── Cost controls (see ConductionNL/.github#596, #599) ───────────────
+      # Run PHPUnit and Playwright only when the diff could change their
+      # verdict. Fails safe TOWARDS running: an unreadable diff, a force-push,
+      # a branch creation or a dispatch all run everything, and a filtered
+      # skip is a declared state in the Quality Report, not a missing one.
+      enable-path-filtering: true
+      # PR-time PHPUnit runs the primary PHP against the newest declared
+      # Nextcloud; the full matrix still runs on every push to a default
+      # branch and on the release PR into beta. Breadth moves from
+      # per-commit to per-merge, it is not dropped.
+      reduce-pr-matrix: true

.gitignore

@@ -23,6 +23,41 @@
 # boot, so it must never be committed (a stray test value here restyles the
 # whole instance the moment the app is activated).
 /css/custom-overrides.css
+# The same, per set, for every set on no design system: the stock Nextcloud
+# theme (custom-overrides-nextcloud.css) and the themes saved off it. Only that
+# set loads its file, so emptying the stock one (the reset) gives stock back.
+/css/custom-overrides-*.css
+
+# Runtime theme state — Nextcloud's own logos, favicon and background image,
+# copied into a theme when it is saved from the editor (BrandingCaptureService)
+# so applying the theme brings them back. They are an admin's uploads on one
+# instance, never part of the app.
+/img/logos/*-captured-*
+/img/backgrounds/*-captured-*
+
+# Runtime theme state — written by the freeform custom-CSS editor at runtime
+# (CustomCssService::write(), atomic temp-file + rename). Unlike
+# custom-overrides.css nothing recreates it on boot: CssInjectionService only
+# emits the stylesheet when the feature is enabled AND the file has content, so
+# its absence is the correct default state. It is admin-generated data, never
+# source — a stray token dump committed here restyles every instance that
+# enables the feature.
+/css/custom-css.css
+
+# Local presentation mock — a visual shell of screens that do not exist yet,
+# drawn over the admin panel and loaded only with `?mock=1`. It is scaffolding
+# for screenshots and demos, not product: nothing it draws persists, none of it
+# is translated, and it must never reach an instance. Kept out of the repo but
+# usable in a working tree, so js/admin-mock.js and css/admin-mock.css can sit
+# here without ever being committed (js/admin-mock.js is already covered by the
+# /js/* rule below — the point is that it is NOT un-ignored there).
+/css/admin-mock.css
+
+# Talking points for a slide deck that lives outside the repo, alongside the
+# local presentation mock above. It names a .pptx on one machine's disk and
+# nothing in the app refers to it, so it is personal working material rather
+# than project documentation.
+/PRESENTATION-SCRIPT.md
 
 # Scratch input for scripts/build-icons.js's DSFR pack — the pre-fetched
 # @gouvfr/dsfr dist/icons/**/*.svg tree (that package cannot be npm-installed:
@@ -40,6 +75,8 @@
 !/js/admin.js
 !/js/lib/
 !/js/preview-banner.js
+!/js/playground.js
+!/js/playground/
 /build/
 /dist/
 
@@ -91,3 +128,10 @@ tests/e2e/.auth/
 # facts about the LIVE docs.numerique.gouv.fr bundle, not app source. The
 # authoritative, committed output is css/systems/lasuite/brand-override.css.
 /.lasuite-src/
+
+
+# Working plans. These are deleted once the work they describe has landed, so
+# nothing that stays in the repository may cite one — not by filename and not by
+# its stage numbering. Write the condition instead ("until the converter has
+# regenerated them"), or cite the openspec change the plan produced.
+/MAKEOVER-PLAN.md

.npmrc

@@ -16,19 +16,12 @@
 min-release-age=2
 min-release-age-exclude[]=@conduction/*
 
-# Required to install at all under npm 11. `@gouvfr/dsfr-nexus` (transitive,
-# via the DSFR toolchain) declares peer dependencies on `@gouvfr/dsfr-token`
-# and `@gouvfr/dsfr-weave` — TWO PACKAGES THAT WERE NEVER PUBLISHED. Both
-# answer 404 on the registry, in every version of dsfr-nexus from 1.2.12 to
-# 1.2.17, so there is no upgrade out of it.
-#
-# npm 10 tolerated the dangling peers. npm 11's `npm ci` does not: it reports
-#   Missing: @gouvfr/dsfr-token@ from lock file
-# and refuses to install, which would have made this repo's CI red the moment
-# the fleet moved to npm 11. Verified: npm ci fails on the UNCHANGED committed
-# lock under npm 11 and succeeds with this line, 944 packages.
-#
-# This makes peers advisory rather than blocking (npm 6 behaviour), the same
-# remedy 8 other apps in the fleet already carry for the same class of
-# upstream defect. Remove it if dsfr-nexus ever drops the phantom peers.
-legacy-peer-deps=true
+# NO legacy-peer-deps. It was here because `@gouvfr/dsfr-nexus` declared peer
+# dependencies on `@gouvfr/dsfr-token` and `@gouvfr/dsfr-weave`, two packages
+# that were never published, which npm 11's `npm ci` refuses to install around.
+# That package is no longer in this tree — the lockfile resolves `@gouvfr/dsfr`
+# alone — so the flag no longer buys anything, and a flag that makes peer
+# conflicts advisory hides the next one instead of reporting it. Verified on
+# npm 11.19: `npm install --package-lock-only` and a clean `npm ci` both exit 0
+# without it. If a phantom peer returns, align the declared ranges until the
+# tree resolves; do not reinstate the flag.

.prettierignore

@@ -36,6 +36,12 @@ css/tokens/
 css/systems/lasuite/defaults.css
 css/systems/lasuite/brand-override.css
 
+# Same contract, different generator: emitted by scripts/generate-guest-css.mjs
+# from the vendored core/css/guest.css and byte-compared against the committed
+# copy by `npm run test:guest-css`. Its generator input lives under
+# scripts/sources/, which is already ignored above.
+css/playground-guest.css
+
 # VENDORED UPSTREAM, AND GENERATOR INPUT. This is the deployed Cunningham token
 # dump that generate-lasuite-tokens.mjs READS to produce brand-override.css.
 # Reformatting the input changes the output, so the drift check would fail on a
@@ -46,3 +52,10 @@ scripts/sources/
 # file — so the validation tests have something to reject. Formatting them
 # would remove the very defect under test.
 tests/integration/fixtures/
+
+# Same contract again: both emitted by scripts/generate-component-scopes.mjs
+# from scripts/mapping/component-tokens.json and byte-compared against the
+# committed copies by `npm run test:component-scopes`. Prettier would rewrap
+# the var() fallback chains and fail that check on a file nobody edited.
+css/component-scopes.css
+css/primary-lock.css

AGREEMENT-MARIANNE.md

@@ -42,7 +42,7 @@ agrees that:
    becomes inapplicable.
 4. **No warranty from the licence itself.** Per the Etalab Open Licence
    2.0's own liability clause (reproduced in
-   [`LICENSES/Etalab-2.0.txt`](LICENSES/Etalab-2.0.txt)), the font files are
+   [`LICENSES/etalab-2.0.txt`](LICENSES/etalab-2.0.txt)), the font files are
    provided as-is; the French State ("le Concédant") gives no guarantee
    beyond what that licence states.