Add V108 + drag-and-drop core + PR #506 review follow-up by mdon · Pull Request #512 · BeamLabEU/phoenix_kit

mdon · 2026-05-02T03:53:26Z

Summary

Four commits on top of `upstream/dev` (`d3b06b13`):

V108 migration — position integer DEFAULT 0 on three list
surfaces (phoenix_kit_entities, phoenix_kit_cat_catalogues,
phoenix_kit_cat_items) so admin lists can persist user-driven
order. Pure schema add; no backfill, no indexes (lists are small).
Drag-and-drop core infrastructure — <.draggable_list> gained
a :draggable boolean attr (default true) so callers can
disable DnD without duplicating card markup. <.table_default>
gained four sortable attrs (:on_reorder, :reorder_scope,
:reorder_group, :item_id) that wire the card-view container as
a SortableGrid hook target. phoenix_kit.js gained
cross-container drop detection + scope passing (via
data-sortable-scope-* attrs and data-sortable-group) plus an
updated() callback on TableCardView so the user's chosen
view-mode survives LV re-renders. Consumers in this batch:
phoenix_kit_entities (records DnD) and the in-flight
phoenix_kit_catalogue work (catalogues / categories / items DnD,
including cross-category item moves).
AGENTS.md TODOs section — surfaces the component-test
coverage gap C12 triage flagged: there are no tests under
test/phoenix_kit_web/components/core/. Inventoried for a future
coverage sweep (<.draggable_list>, <.table_default>, the form
primitives, possibly <.flash> / <.modal>).
PR Support arity-2 dynamic_children callback with locale #506 review follow-up — closes the two NITPICKs from
Claude's review of the merged arity-2 dynamic_children work.

PR follow-up(s)

PR Support arity-2 dynamic_children callback with locale #506 — `3d4bf2f1` "PR Support arity-2 dynamic_children callback with locale #506 review follow-up — close the
two NITPICKs". The merged arity-2 dynamic_children work
shipped with a half-tautological dispatch test (anonymous fn
invoked via .(), asserting Elixir's call semantics rather than
the sidebar's actual dispatcher) and an inline # comment on
dynamic_children_fn that ExDoc didn't pick up.
Fix: added a `@doc false` test-only delegate
`AdminSidebar.invoke_dynamic_children_for_test/3` that
calls the private `invoke_dynamic_children/3`. Rewrote the
describe block with four assertion-pinned tests covering arity-1,
arity-2, nil locale, and return-value propagation. Replaced the
inline comment on `dynamic_children_fn` with a proper
`@typedoc`. FOLLOW_UP.md filed under
`dev_docs/pull_requests/2026/506-dynamic-children-locale/`.

Quality sweep

The big work, structured by commit.

V108 migration (`38b696cd`)

`lib/phoenix_kit/migrations/postgres/v108.ex` adds nullable
`position integer` (default 0) to three tables — the
entity-definitions list, the catalogues index, and the catalogue
items list. Idempotent (`ADD COLUMN IF NOT EXISTS` matches the
project pattern from V107 and earlier). `down/1` drops the
columns and reverts the `COMMENT ON TABLE phoenix_kit IS` marker
from `'108'` back to `'107'`.

No indexes on the new `position` columns. The lists are small
(≤ a few hundred rows in practice) and have other indexed scope
filters; a position-only btree would burn write amplification for
no real read win. `phoenix_kit_entity_data` already got both
`position` and the `(entity_uuid, position)` composite index
in V81, so it's untouched here.

`@current_version` bumped from 107 → 108 in
`lib/phoenix_kit/migrations/postgres.ex`.

DnD core infra (`b27e647e`)

Three additive changes to the core DnD primitives so consumer
modules wire reorder UIs without forking the hook or the components.

`SortableGrid` hook (`priv/static/assets/phoenix_kit.js`):

Reads `data-sortable-group` and passes it to SortableJS as the
`group` config. Tables sharing a group name can exchange items
via cross-container drag (catalogue's "items move between
categories" use case). When unset, single-container reorder
only — behavior unchanged.
On `onEnd`, detects cross-container drops via
`evt.from !== evt.to`. Cross-container payloads carry
`moved_id` plus the source container's scope attrs prefixed
with `from*`, alongside the destination's regular scope attrs.
Push routes through the destination's `data-sortable-event` so
the LV handler is co-located with the table the item ended up
in.
A `readScope/1` helper translates `data-sortable-scope-*`
attrs (camelCase via DOMStringMap) into a `{key: value}` map.
The prefix-length check ensures a bare `data-sortable-scope`
attr can't masquerade as scope data.
The `onEnd` body is wrapped in try/catch (a single bad dataset
value, e.g. corrupt scope or missing source container after a
fast unmount, logs to `console.error` instead of leaving
SortableJS half-initialized).
Leading docstring spells out the "trust your own DOM"
assumption — if a third-party script injects sortable-item
nodes, the LV handler should reject unknown IDs server-side.

`TableCardView` hook (`priv/static/assets/phoenix_kit.js`):

Added `updated()` callback that re-applies the user's saved view
mode after every LV-driven render. Without it, morphdom resets the
runtime `md:hidden` class toggles back to template defaults on
each diff (e.g. a SortableJS drop fires a reorder event, the LV
re-renders, and the card view snaps back to the table view). The
hook now caches `currentMode` on mount + every toggle/event and
replays it in `updated()`.

`<.draggable_list>` (`components/core/draggable_list.ex`):

New `:draggable` boolean attr (default `true`). When false,
the container renders without `phx-hook="SortableGrid"` and
items skip the `cursor-grab` styling — useful when a list is too
short to reorder (length ≤ 1) so the affordance doesn't lie.
Default `true` keeps every existing caller unchanged.

`<.table_default>` (`components/core/table_default.ex`):

Four new attrs — `:on_reorder`, `:reorder_scope`,
`:reorder_group`, `:item_id` — wire the card-view container
as a `SortableGrid` hook target when set. Each rendered card
gets `class="sortable-item cursor-grab active:cursor-grabbing"`

`data-id={item_id_fn.(item)}` so a user toggled into card view
(mobile or desktop click) gets the same DnD experience as the
table view. `:item_id` defaults to `& Map.get(&1, :uuid)` so
the conventional schema shape works zero-config.

A footer row joins the card-actions slot when sortable: drag-handle
icon (`hero-bars-3` with cursor-grab styling,
`gettext("Drag to reorder")` tooltip) bottom-left, action
buttons bottom-right. `use Gettext, backend: PhoenixKitWeb.Gettext`
added to match the convention in `flash.ex` / `input.ex` /
`integration_picker.ex`.

A private `build_sortable_scope_attrs/1` helper translates a
`%{key => value}` scope map into
`[{"data-sortable-scope-key" => value}, ...]` tuples, spread
via `{@reorder_scope_attrs}` in HEEX (Phoenix's
`attributes_escape/1` applies, so untrusted-input callers
don't introduce an XSS path).

AGENTS.md TODOs section (`af35b067`)

C12 triage on the V108 + DnD work surfaced that there are no tests
under `test/phoenix_kit_web/components/core/` — the directory
doesn't exist. Several components there have grown to non-trivial
attr surfaces that warrant rendered-HTML coverage, but landing a
single fixture in an empty test dir alongside a feature PR muddles
the diff.

Recorded the gap as a TODOs section at the bottom of `AGENTS.md`
so a future component-coverage sweep has the inventory ready
(`<.draggable_list>` :draggable branches, `<.table_default>`
:on_reorder / :reorder_scope / :reorder_group / :item_id branches,
the form primitives, plus `<.flash>` / `<.modal>` if their
complexity has grown).

Triage outcome (C12 + C12.5 deep dive)

Three Explore agents ran against the diff. Most findings turned
out to be agent overreach:

Agent feat: Add user creation and editing functionality with LiveView #1 "CRITICAL — SQL injection in V108 prefix" — **false
alarm**. The pattern is consistent with all 107 prior
migrations; `prefix` is host-app-controlled at migration time,
not user input.
Agent feat: Add user creation and editing functionality with LiveView #1 "BUG-MEDIUM — data-id leakage when draggable=false" —
**false alarm**. HEEX omits `nil` attrs.
Agent feat: Add user creation and editing functionality with LiveView #1 "IMPROVEMENT-HIGH — XSS via reorder_scope" — **false
alarm**. Phoenix HEEX `{@list_of_tuples}` spread routes
through `attributes_escape/1` which escapes values.
Agent Updated the comments in users_live html file, and added Modules link in dashboard. #2 "BUG-MEDIUM — Missing V108 test" — **false alarm**.
Per V107's docstring, schema-only changes are implicitly verified
by `test_helper.exs` running all migrations on boot; only
backfill logic gets a dedicated test, and V108 has none.
Agent Updated the comments in users_live html file, and added Modules link in dashboard. #2 "BUG-MEDIUM — Missing component attr tests" — **out
of scope**; landing a single fixture in an empty test dir is
awkward. Surfaced via the new AGENTS.md TODOs section.

Real findings, all closed in this batch:

Agent feat: Add user creation and editing functionality with LiveView #1 BUG-HIGH "JS try/catch on onEnd" — wrapped onEnd, errors
now log to `console.error`.
Agent feat: Add user creation and editing functionality with LiveView #1 IMPROVEMENT-MEDIUM "DOM trust assumption" — added
docstring at the hook.
Agent Updated the comments in users_live html file, and added Modules link in dashboard. #2 NITPICK "gettext for 'Drag to reorder'" — added
`use Gettext` to `<.table_default>`, wrapped the title.

Companion PR

The downstream consumer is the in-flight
`mdon/phoenix_kit_entities`
work which adds DnD to the entity definitions list and the
DataNavigator records list. Until this core PR merges + a release
cuts, the entities suite fails on `column phoenix_kit_entities.
position does not exist` because the local schema has the
`:position` field but the Hex-pinned core's migration chain stops
at V107.

Verification

`mix format --check-formatted` clean
`mix compile --warnings-as-errors` clean
`mix credo --strict` clean (whole tree, 7564 mods/funs, 0
issues — 0 net change vs pre-PR; the one pre-existing format
violation in `integrations/providers.ex` is intentionally
left out of this diff)
`mix test test/phoenix_kit_web/components/admin_sidebar_dynamic_children_test.exs`:
6 tests, 0 failures (2 from the original `dynamic_children_fn
type` describe block + 4 new dispatch tests via the
`invoke_dynamic_children_for_test/3` delegate)
V108 round-tripped locally via
`phoenix_kit_parent`'s migration runner: up adds the three
columns, down drops them, marker comment flips both directions.

Test plan

`mix format --check-formatted` clean
`mix compile --warnings-as-errors` clean
`mix credo --strict` clean
`mix test` for the touched test file passes (6/6)
V108 up + down round-trip verified
PR Support arity-2 dynamic_children callback with locale #506 FOLLOW_UP.md filed and committed
Companion entities PR Handling fix for missing handle_info() for :user_session_connected broadcast #11 verifies the new attrs end-to-end
via a parent app running V108 locally

Three independent admin lists previously sat in insertion order (latest first or alphabetical) with no user-driven order: - `phoenix_kit_entities` — the entity-definitions list at `/admin/entities` - `phoenix_kit_cat_catalogues` — the catalogues index at `/admin/catalogue` - `phoenix_kit_cat_items` — items shown on a catalogue's detail page and inside categories V108 adds a nullable `position integer` to each with a default of `0` so legacy rows can keep insertion-order shape until somebody drags; the LV reorder handlers re-index the visible group to `1..N` on the first user drag, so the default is only ever observed transiently. Idempotent — `ADD COLUMN IF NOT EXISTS` matches the project pattern from V107 and earlier; safe to re-run mid-migration. Down/1 reverses column-by-column AND updates the `COMMENT ON TABLE phoenix_kit IS` marker from `'108'` back to `'107'` (the source of truth read by `migrated_version/1`). Categories and smart-catalogue rules already carry their own `position` columns (V87 and V102 respectively), so they're untouched here. The `phoenix_kit_entity_data` table already got both `position` and the `(entity_uuid, position)` composite index in V81. No indexes on the three new columns themselves — entity / catalogue / item-per-catalogue lists are small (≤ a few hundred rows) and have other indexed scope filters; a position-only btree would burn write amplification for no real read win.

…e, :draggable opt-out Three additive changes to the core DnD primitives so consumer modules (phoenix_kit_entities, phoenix_kit_catalogue) can wire reorder UIs without forking the hook or the components. ## SortableGrid hook (`priv/static/assets/phoenix_kit.js`) - Reads `data-sortable-group` and passes it to SortableJS as the `group` config. Tables sharing a group name can exchange items via cross-container drag (catalogue's "items move between categories" use case). When unset, behavior is unchanged — single-container reorder only. - On `onEnd`, detects cross-container drops via `evt.from !== evt.to`. Cross-container payloads carry `moved_id` (the dragged item's data-id) and the source container's scope attrs prefixed with `from*` alongside the destination's regular scope attrs. The push routes through the destination's `data-sortable-event` so the LV handler is co-located with the table the item ended up in. - Pulls a `readScope/1` helper that translates `data-sortable-scope-*` attrs (camelCase via DOMStringMap) into a `{key: value}` map. The prefix-length check (`key.length > "sortableScope".length`) ensures a bare `data-sortable-scope` attr can't masquerade as scope data. - The `onEnd` body is now wrapped in try/catch — a single bad dataset value (corrupt scope attr, missing source container after a fast unmount) logs to `console.error` instead of leaving SortableJS in a half-initialized state with the LV unable to reorder again. - A leading docstring spells out the "trust your own DOM" assumption for the `.sortable-item[data-id]` query — if a third-party script injects sortable-item nodes, IDs may be poisoned, so the LV handler should reject unknown IDs at the server side. ## TableCardView hook (`priv/static/assets/phoenix_kit.js`) Added `updated()` callback that re-applies the user's saved view mode after every LV-driven render. Without it, morphdom resets the runtime `md:hidden` class toggles back to template defaults on each diff (e.g. after a SortableJS drop fires a reorder event, the LV re-renders and the card view snaps back to the table view). The hook now caches `currentMode` on mount + every toggle/event and replays it in `updated()` so the chosen mode survives. ## `<.draggable_list>` (`components/core/draggable_list.ex`) New `:draggable` boolean attr (default `true`). When false, the container renders without the `phx-hook="SortableGrid"` and items skip the `cursor-grab` styling — useful when the list is too short to reorder (length ≤ 1) so the affordance doesn't lie. Default `true` keeps every existing caller unchanged. ## `<.table_default>` (`components/core/table_default.ex`) Four new attrs — `:on_reorder`, `:reorder_scope`, `:reorder_group`, `:item_id` — wire the card-view container as a SortableGrid hook target when set. Each rendered card gets `class="sortable-item cursor-grab active:cursor-grabbing"` + `data-id={item_id_fn.(item)}` so a user toggled into card view (mobile or desktop click) gets the same DnD experience as the table view. The `:item_id` defaults to `& Map.get(&1, :uuid)` so the conventional schema shape works with zero config; any non-uuid keying passes a custom function. A footer row joins the card-actions slot when sortable: drag-handle icon (`hero-bars-3` with cursor-grab styling, `gettext("Drag to reorder")` tooltip) bottom-left, action buttons bottom-right. `use Gettext, backend: PhoenixKitWeb.Gettext` added to match the convention in `flash.ex` / `input.ex` / `integration_picker.ex`. A private `build_sortable_scope_attrs/1` helper translates a `%{key => value}` scope map into `[{"data-sortable-scope-key" => value}, ...]` tuples, spread via `{@reorder_scope_attrs}` in HEEX (Phoenix's attribute-spread escape applies, so untrusted-input callers don't introduce an XSS path). `mix format` clean. `mix credo --strict` clean (7564 mods/funs, 0 issues).

C12 triage on the V108 / DnD core work surfaced that there are no tests under `test/phoenix_kit_web/components/core/` — the directory doesn't exist. Several components there have grown to non-trivial attr surfaces that warrant rendered-HTML coverage, but landing a single fixture in an empty test dir alongside a feature PR muddles the diff. Recording the gap as a TODOs section so a future component-coverage sweep has the inventory ready (`<.draggable_list>` :draggable branches, `<.table_default>` :on_reorder/:reorder_scope/ :reorder_group/:item_id branches, the form primitives, plus `<.flash>` / `<.modal>` if their complexity has grown).

PR BeamLabEU#506 (Support arity-2 dynamic_children callback with locale, merged 2026-04-24) was APPROVED with two NITPICKs. Both closed: ## Test coverage was half-tautological The previous `"invoke_dynamic_children/3 dispatch"` describe block defined two anonymous functions, invoked them with `.(%{})` and `.(%{}, "en-US")`, and asserted the counters incremented. As Claude's review pointed out, that tested Elixir's function-call semantics rather than the sidebar's actual dispatch logic. Added a `@doc false` test-only delegate `AdminSidebar.__invoke_dynamic_children_for_test__/3` that calls the private `invoke_dynamic_children/3`, then rewrote the describe block with four assertion-pinned tests: - arity-1 callback receives only the scope - arity-2 callback receives both scope and locale - arity-2 callback handles a nil locale gracefully - return value is propagated unchanged The delegate carries an explicit `@doc false` so it's not part of the runtime surface — it exists solely to let the unit suite reach the private dispatcher without coupling to LV rendering. ## @TypeDoc on dynamic_children_fn The inline `#` comment above `@type dynamic_children_fn` wasn't picked up by ExDoc / `h Tab`. Replaced with a `@typedoc` block that documents both arities, the explicit-locale rationale, and the `nil` semantic. `mix test` for the test file: 6 tests, 0 failures (2 original + 4 new). `mix format`, `mix credo --strict` clean, `mix compile --warnings-as-errors` clean. PR-folder FOLLOW_UP.md filed under `dev_docs/pull_requests/2026/506-dynamic-children-locale/`.

C12 triage on `b451d64`'s DnD work surfaced four real issues; this batch closes them. C12.5 deep-dive categories all clean (most N/A in this diff: no PubSub/translations/auth additions; existing Gettext wrapping verified; activity-log helpers re-used). ## Cross-container atomicity (BUG-MEDIUM) `CatalogueDetailLive`'s cross-category drop did `Catalogue.move_item_to_category/3` and `Catalogue.reorder_items/4` as separate context calls. If the move committed but the reorder rolled back (transaction error, scope mismatch, future cap-exceeded), the item ended up in the new category with stale position values — inconsistent with what the user just dragged. Closes by adding `Catalogue.move_item_and_reorder_destination/4` which wraps both context fns in a single `repo().transaction/1`. Either both land or both roll back; the LV's `with` chain calls this single fn instead of chaining the two. Activity-log fan-out for `item.moved` and `item.reordered` still happens through their normal paths so the audit trail keeps full attribution. ## Length cap @ 1000 + dedup (IMPROVEMENT-MEDIUM) Parity with the entities reorder paths. Each of `Catalogue.reorder_catalogues/2`, `reorder_categories/4`, `reorder_items/4`, and `Rules.reorder_catalogue_rules/3` now: - Returns `{:error, :too_many_uuids}` when the input list exceeds the shared `@reorder_max_uuids 1000` cap. Protects the transaction from N+1 unbounded write storms; even a workspace with hundreds of catalogues / categories / items per group never paints a thousand at once. - Dedups inputs (last-occurrence wins via `dedupe_keep_last/1`). A double-drop or stale DOM no longer issues redundant per-uuid writes inside the transaction. ## Audit-trail gaps on rejection + DB error (BUG-HIGH) The reorder paths logged only on `:ok`. Now they log on every branch via two new private helpers, `log_reorder_rejected/5` and `log_reorder_db_error/5`: - Early rejection (`:too_many_uuids`, `:not_siblings`, `:wrong_scope`) lands an audit row with `metadata.db_pending: true` and `metadata.rejected: <atom>`. The user-initiated action is still attributable even though no DB write happens. - DB transaction failure on success-shape input lands a row with `metadata.db_pending: true` (no `rejected` flag — the input was valid, the transaction just failed). Lets audit consumers tell rejected and failed apart. `Rules.reorder_catalogue_rules/3` mirrors the shape with its own `log_smart_rules_reorder_rejected/4` and `log_smart_rules_reorder_db_error/3` (smart-rule activity rows go to the `item` resource_type, not a generic shared helper). ## Cleanup (credo) - Refactored `reorder_catalogues/2` to extract a `write_catalogue_positions/1` private helper. Drops the credo "Function body too deep" finding. - The pre-existing `cond` with one non-`true` branch in `CatalogueDetailLive.handle_event("reorder_items", _)` (a credo refactoring opportunity) becomes a clean `if` / `else` plus a new private `apply_in_scope_item_reorder/4` helper. ## Verification - `mix format --check-formatted` clean - `mix credo --strict` clean (1220 mods/funs, 0 issues) - `mix compile` clean apart from the documented standalone-vs-parent warnings (the `<.table_default>` `:on_reorder` / `:reorder_scope` / `:reorder_group` / `:item_id` attrs and `<.draggable_list>` `:draggable` attr land via the in-flight `BeamLabEU/phoenix_kit#512` DnD core PR; until that merges + a release cuts, standalone catalogue tests will continue to fail per `feedback_run_tests_via_parent.md`)

PR #512 review verdict: APPROVE. Six findings, all NITPICK / LOW. Two are safe to close inline: ## reorder_scope camelCase round-trip — documented Elixir map key `:category_uuid` round-trips through `build_sortable_scope_attrs/1` (lowercase + dash) and the `SortableGrid` hook's `readScope/1` (strip "sortableScope" prefix + lowercase first char) as `"categoryUuid"` in the LV handler payload. The original `:reorder_scope` doc didn't mention this; consumers following Phoenix params instinct would write `%{"category_uuid" => uuid}` and silently miss the param. Added the round-trip example to the attr `:doc`. ## draggable_list data-id always emitted The PR introduced `data-id={if @Draggable, do: @item_id_fn.(item)}` which strips `data-id` on the non-DnD path. `data-id` is useful even without DnD — click-to-select handlers, integration test selectors, JS that reads source-of-truth IDs off the element. Restored the unconditional emission and updated the `:draggable` attr doc to state the guarantee explicitly. ## on_reorder asymmetry — clarified The doc already noted that `:on_reorder` only wires the card view and the consumer owns the table view's tbody. Tightened the wording to make the desktop/mobile asymmetry impossible to miss. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

PR #514 extracted Customer Service to phoenix_kit_customer_support and renamed the module surface; reviewed and closed two minor findings. Changes (review close-outs): - Backfill V108 docstring section in lib/phoenix_kit/migrations/postgres.ex (was missing — pre-existing gap from PR #512 that this PR was the natural place to close) - Drop unused _prefix arg from rename_role_permission/4 in v109.ex (table name is already prefix-qualified at the call site) Skipped on purpose: rewriting V109's DO \$\$ blocks as parameterized queries. V109 is unpublished but the values are migration-time constants, the existing implementation is tested and idempotent, and minimizing surface area before publish beats stylistic polish. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Triage of CLAUDE_REVIEW.md against current code: all six findings (5 NITPICKs + 1 LOW) addressed pre-existing — `:reorder_scope` camelCase round-trip doc, `:on_reorder` asymmetry doc, `data-id` unconditionally emitted, V108 sort stability covered by downstream catalogue's `[asc: :position, asc: :name]` ordering. FOLLOW_UP.md records the audit.

V111 PDF library tables + #511/#512/#515 follow-ups

mdon added 4 commits May 2, 2026 06:54

mdon force-pushed the dev branch from 3d4bf2f to cc9ed09Compare May 2, 2026 03:55

mdon mentioned this pull request May 2, 2026
DnD reorder + bucket-C migration cleanup + reorder hardening BeamLabEU/phoenix_kit_catalogue#22
Merged
5 tasks

ddon merged commit df8e908 into BeamLabEU:devMay 2, 2026

mdon mentioned this pull request May 5, 2026
V111 PDF library tables + #511/#512/#515 follow-ups #516
Merged
5 tasks

ddon added a commit that referenced this pull request May 6, 2026
Merge pull request#516from mdon/dev
0966478
V111 PDF library tables + #511/#512/#515 follow-ups

Navigation Menu

Uh oh!

Add V108 + drag-and-drop core + PR #506 review follow-up - #512

Add V108 + drag-and-drop core + PR #506 review follow-up#512
ddon merged 4 commits into
BeamLabEU:devfrom
mdon:dev

mdon commented May 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

mdon commented May 2, 2026

Summary

PR follow-up(s)

Quality sweep

V108 migration (`38b696cd`)

DnD core infra (`b27e647e`)

AGENTS.md TODOs section (`af35b067`)

Triage outcome (C12 + C12.5 deep dive)

Companion PR

Verification

Test plan

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants