draw is single-token (no verb), so `clide draw --file x` mis-parsed as cmd
"draw.--file" — the dispatcher registered draw but the argv parser never
learned it was umbrella (unlike icon/image which use subsystem.verb). Found
driving the live install; the handler unit tests bypassed the argv path so
they missed it. Added draw to _umbrellaCommands + a parse regression test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fruchterman-Reingold (clide-owned, no layout package): nodes repel, edges
attract, cooled over iterations into a readable layout. Deterministic — a
fixed circular seed, no RNG — so the graph view is stable across rebuilds
and the solver is unit-tested (6 cases). Flutter-free; the interactive graph
pane (render, pan/zoom, hover, filter, pql wiring) builds on this.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
graphTemplateHandler lowers a {nodes,edges} payload to SVG: a deterministic
circular layout of labelled <circle> nodes with <line> edges, self-contained
(own light backdrop + content colors, like a d2 card) and painted by the
shared renderer. Display-only (D-78); decoupled from the interactive graph
pane (T-323) since a static card only needs a layout->SVG. Honest error on
a duplicate id or an edge to an unknown node.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The C client slurps stdin when it sees --stdin, strips the flag, and ships
the payload alongside the argv; the Dart unwrap folds it into the request
as a `stdin` arg (undeclared keys pass the schema untouched). icon.show and
image.show now read that payload as the peer of --file (stdin wins). Bounded
slurp + the envelope's existing size guard keep a huge payload from
corrupting the wire. C client builds clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
compareTemplateHandler lowers an `images` array to an SVG of side-by-side
<image> cells, each with the per-object data-label/description caption and
data-lightbox; paths resolve to absolute up front (injected, honest error
on a miss). _paintImage now aspect-fits (contain, centered) so cells of
differing-shape images don't distort. Flutter-free handler + painter pixel
tests. Card image-loading + registration next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
icon.show resolves Phosphor glyphs by name (injected resolver) or a 0xNNNN
codepoint, reads a --file JSON array of {icon,label,description,color}
entries, validates colors via parseSvgColor (hex or CSS name), and
publishes on the `icon` bus channel. Honest userError on an unknown glyph,
a bad color, or a malformed payload. A trailing stringList positional is
now variadic so `icon show gear folder gauge` collects every token.
Flutter-free, dart-tested. Card widget + wiring next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`clide image show --file meta.json` reads a {path,label,description,caption}
payload so an image can carry a title and a longer description, not just a
one-line caption. ImageMessage + the image card render the richer metadata;
the bare `image show <path> [--caption]` form is unchanged. Honest userError
on a malformed/missing payload. Text annotation only (option a) — visual
marker overlays stay a follow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The rendered diagram leads; the d2 source carries through the draw bus and
folds into a collapsed ClideCollapserCard beneath the card (D-78 display-
only). en+nl catalogs. Also switches the d2 test to a null-aware element.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The d2 drawing template compiles a diagram's source to SVG through the d2
binary (resolved via the D-104 path layer), then paints it with the same
renderer the svg card uses. `clide draw --file x.d2` infers the type from
the extension; `.svg` files render directly. Template handlers now return
a DrawResult so a compile failure or an unresolved d2 surface as an honest
userError with an install hint, not a generic "no SVG". Real d2 0.7.1
verified end to end.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ToolsSettingsExtension adds a Tools settings category — a path field per
tool (app.tools.<name>) plus a Re-detect action (tools.detect) — and keeps
the live resolver in sync as paths are edited (supporterBinariesFrom). en
+ nl catalogs. Completes the D-104 UI surface.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Store each tool's override under its own key (app.tools.<name>) instead of
one map, so the settings panel can bind a plain text field per tool.
loadSupporterBinaries reads the per-tool keys plus a first-run marker and
keeps an explicit path; redetectSupporterBinaries backs the Re-detect
action. The resolver itself is unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
main resolves the override map (and first-run auto-detects, using the
primed login-shell PATH) right after bootSettings.load(), exposing the
process-wide activeSupporterBinaries that tool consumers read — the d2
template (T-494) will resolve its binary through it. Analyze clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
loadSupporterBinaries reads the persisted tool→path override map; on first
run (key absent) it probes the known tools once and persists the result —
pinned thereafter, not re-detected each launch (D-104). Injected read/write
so it stays Flutter-free and dart-test covered (3 new cases). Boot wiring
next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SupporterBinaries resolves claude/d2/etc. to absolute paths (D-104): an
explicit override map first, then the login-shell/process PATH, then the
well-known bin dirs incl. Homebrew-on-Linux — the dir the standard PATH
expansion omits and a .bashrc-only brew shellenv misses (the D-104 gap).
detect() probes those to seed the override map (pinned, not re-probed each
launch); isStalePin flags a moved tool. Pure Dart, dart test (11 cases).
Settings storage + first-run detect wire on next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The builder captures data-label / data-description / data-lightbox on SVG
elements into SvgDocument.annotations, each with the element's AABB in
viewBox coordinates (accumulated transform applied — rect/ellipse/image/
line/poly/path bounded; groups skipped, text degenerates to its anchor
point). Feeds the Flutter caption/lightbox overlay next. Flutter-free,
covered by dart test (6 new cases).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The last wire: a DrawingMessage conversation item + the Claude-extension
subscriber on the `draw` channel (buildSvgDocument → inject), the
conversation renderer (DrawingCard), the fold/summary switches, the i18n
`drawing` label, and registering `clide draw` at boot (empty template
registry for now — primitive SVG works; d2/icon/compare/image handlers
plug in as they land). `clide draw --file doc.json` now renders a card in
the conversation. Render widget-tested; analyze + format clean across the
repo. Templates + per-object overlay remain; the CHANGELOG entry lands
with the user-complete feature.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirrors image.show: a Flutter-free `draw` handler reads + JSON-parses the
document (injected reader), lowers it to SVG via the template registry,
and publishes {svg, label, description} on the `draw` MessageBus channel
for the Claude extension to inject. Honest errors — notFound (missing
file), userError (bad JSON / unknown template / no --file), toolError (no
live UI). Covered by dart test (7 cases). The extension subscription that
injects the card remains to wire.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
resolveDrawingSvg lowers a DrawingCardDoc to an SVG string: primitive docs
use inline svg or read svgPath (via an injected reader); template docs use
a registered DrawingRegistry handler, so the d2/icon/compare/image children
plug in. Honest DrawErr on no source / unknown template / unreadable path /
empty output. Flutter-free, covered by dart test (7 cases). Not yet wired.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DrawingCardDoc + parseDrawingCardDoc read the `clide draw` JSON: card
label/description, primitive svg/svgPath, or a template name plus its
fields (for a handler to lower to SVG). Tolerant — a non-object payload
is null, blank strings are absent, never throws. Flutter-free, covered by
dart test (10 cases). First slice of the drawing-card core (D-103); not
yet wired.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collect <marker> defs (refX/refY, orient=auto, markerUnits) and the path
marker-start/mid/end refs in the builder, then paint them at the path
endpoints rotated to the path tangent (orient=auto). The viewBox→viewport
scale is approximated 1:1 (holds for d2's markers). d2 edges now render
with arrowheads. Covered by dart test (builder) + a flutter-test pixel
probe (the green arrowhead draws). image painting remains the last T-320
follow-on.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The builder ties the five parsers together: parseXml → inlineStyles →
a typed SvgNode tree (group/rect/ellipse/line/poly/path/text/image) with
viewBox, per-node Affine transforms, and inheritance-flattened SvgStyle
(fill/stroke/font resolved to ARGB; opacity per-node). Tolerant — a
non-svg root yields an empty doc; defs/marker deferred. An end-to-end test
runs the whole pipeline against a real d2-rendered SVG fixture. Flutter-
free, covered by dart test (17 cases). No user-visible behaviour yet.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
parseSvgColor → packed ARGB (#rgb/#rgba/#rrggbb/#rrggbbaa, rgb()/rgba()
integer or %, common names, none/transparent → 0x0); null on unrecognised
so the caller can inherit. parseTransform → a composed 2-D Affine
(translate/scale/rotate[/about-point]/matrix/skew), applied left-to-right.
Both tolerant and Flutter-free, covered by dart test (22 cases). These feed
the typed node model next. No user-visible behaviour yet.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Zero-dependency, tolerant XML reader (elements/attrs/text/comments/prolog/
CDATA/entities; <style> read as raw text) producing a generic element tree.
The normalizer folds d2/graphviz's class-based <style> rules into inline
presentation attributes — cascade presentation-attr < tag < class < style —
then drops <style>/class/style, so the painter only ever sees inline attrs
(D-103). Flutter-free, covered by dart test (26 cases). No user-visible
behaviour yet.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Parse the `d` attribute into an absolute-coordinate SvgPathSeg list,
lowering SVG shorthands to a small command set the painter can replay:
relative→absolute, H/V→lineTo, S→cubic and T→quad with control
reflection, arcs kept as arcTo. Tolerant — malformed tail returns the
understood prefix and never throws, so a bad diagram can't crash the
conversation. Flutter-free, covered by dart test. First slice of the SVG
engine (D-103); no user-visible behaviour yet.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes the new-project story (T-486). The dead-end "not a git repo" dialog now
offers to initialize the folder: project.init runs git init + a non-clobbering
scaffold, then opens + announces on projectCreatedChannel so the account
roadblock fires — the same path a brand-new project takes. Adds initExistingProject
+ the `clide project init [--dir]` verb (default: the current workspace).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The UI half of the new-project flow (story T-486). A "New project…" welcome
action opens a dialog (location + name) that dispatches project.new (T-487),
opens the result, and announces it on projectCreatedChannel. The Claude
extension consumes that and shows the account roadblock — the embedded
per-workspace picker + accounts list, so a fresh project gets bound to an
account (or Default) right at birth.
The two halves stay decoupled: the welcome builtin only publishes the event
(no claude import); the claude builtin owns the account dialog. Only freshly-
created projects announce, so existing opens are never prompted.
Closes T-488. The non-repo "initialize as a project" path (T-489) is next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The backend half of the new-project flow (story T-486). createNewProject
validates the name, makes <parent>/<name>/, runs git init (injected from the
toolchain in main.dart so the handler stays Flutter-free), and writes a minimal
scaffold (.gitignore + a CLAUDE.md stub). The project.new verb wraps it; --dir
defaults to the current workspace's parent so a new project lands beside it.
Create-only by design — opening the new workspace and the account roadblock are
the UI flow's job (T-488). Closes T-487.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
clide treats a git repo as the workspace, so creating a new project starts with
`git init`. Add GitClient.init — `git init -b <main>` in workDir, deterministic
default branch, idempotent on an existing repo. The first primitive of the
new-project flow (T-486); the create-dir + scaffold service and the
`clide project new` verb build on it next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A `claude` started with CLAUDE_CONFIG_DIR=<dir> looks for clide's /ide
discovery lock under <dir>/ide, not ~/.claude/ide — so today a per-repo
account session (T-476) can't reach clide's IDE bridge.
McpServer now reconciles a SET of lock paths instead of one: the default
~/.claude/ide always, plus the bound account's <dir>/ide when this workspace
is bound (boundConfigDir, injected from main.dart via the AccountRegistry).
syncDiscoveryLocks() writes/removes locks to match the active set; it runs on
start, once post-boot when the registry becomes resolvable, and on every
accountActionChannel event, so binding/unbinding adds/removes the account lock
at runtime. stop() sweeps every path written — no orphans. Lock content +
0600 are identical across dirs.
Acceptance 1-4 and 6 are covered by mcp_server_test (multi-dir write, no-binding
regression, dynamic add/remove, full cleanup); #5 (a live claude resolving the
bridge end-to-end) is manual.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The extension consumer for the per-repo account verbs, making set/unset
fully functional. The Claude extension subscribes to accountActionChannel:
- set / unset → ClaudeSessionOrchestrator.respawnForWorkspace(cwd): closes the
workspace's solo sessions (awaiting real process death, T-437) and re-spawns
each on the same id with --resume, so the conversation continues under the
newly-bound CLAUDE_CONFIG_DIR (resolved at spawn by agentBootstrap). Team and
forked sessions are skipped — re-joining the broker / re-forking on an account
swap is out of scope; they adopt the account on their next natural spawn.
- remove --purge → deletes the config dir behind isPurgeableAccountDir, a strict
guard that only ever removes a ~/.claude-* directory that is a direct child of
$HOME. The purge payload now carries the dir (the account is gone from the
registry by publish time).
login still only publishes its action — spawning the `claude login` terminal
pane needs argv+env terminal-pane support and is split to T-485.
Covered: respawnForWorkspace (respawn solo, skip fork/other-repo) and the purge
guard's accept/reject matrix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The CLI half of the multi-account feature (epic T-476; D-6 parity). A new
`claude.account` dispatcher command multiplexes the six sub-verbs over an
injected, Flutter-free AccountStore port (runs under `dart test`):
- add <name> [--dir] register (default ~/.claude-<name>); idempotent, clear
conflict error
- list {accounts, boundAccount (this workspace), detected}
- set <name> bind this workspace (persists)
- unset clear this workspace's binding
- remove <name> [--purge] registry-remove; refuses while any workspace is
bound
- login <name> (publishes the login action)
Registry reads/writes go through the user-scope SettingsStore; side-effects
that only the UI layer can do — respawn on set/unset, the `claude login`
terminal pane, and the --purge rm — are published on accountActionChannel for
the Claude extension to consume (that consumer is T-480 part 2). main.dart
adapts the real AccountRegistry to the port and registers the command alongside
image.show / status.
Adds SettingsStore.keysAt (binding enumeration) and AccountRegistry.boundName /
boundAccountNames. No changelog yet — set/unset don't auto-respawn until part 2,
so the feature isn't user-complete. Verb behaviour + payloads + the in-use
guard are unit-tested against a fake store.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes the observability half of T-247: a way to find and identify running
clides. `clide instances` scans the runtime socket dir, probes each live
*.sock, and prints its identity (version/pid/workspace/socketPath) as jsonl;
dead sockets are skipped. `clide instance` reports the one you're connected
to. Combined with CLIDE_SOCK honoring (this same ticket), you can now list
instances and pin the CLI to a chosen one.
Server: a new `instance` dispatcher command (registered in buildDispatcher
with the live workspace/pid/socket) returns the identity map; added to the
argv translator's umbrella set so a bare `instance` token routes to it.
Client: a POSIX dir-scan in clide.c (Windows stub until it ships).
Tests: e2e `instances` lists the live test server with its identity; a
cc-free unit test covers the `instance` command shape.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The runtime socket dir accumulated orphaned *.sock nodes from crashed
instances — only the current workspace's own path was ever cleaned. Add a
best-effort startup sweep that probes every *.sock in the dir and unlinks
only the dead ones; live instances (something answers) and unresponsive
nodes (possibly hung) are left alone. Runs before bind, alongside the
existing per-workspace stale-unlink. Never blocks our own startup on a
sweep failure.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
EditorRegistry.close() guarded its active-changed emit on `_activeId !=
null`, so closing the LAST buffer (active clears to null) emitted only
editor.closed — never the active-changed(id:null) the editor extension
listens for to call closeEditor(). editorOpen stayed true and the top
split sat orphaned over the Claude pane. Always emit active-changed when
the active buffer is removed, including the cleared-to-null case; the
slot renderer already collapses correctly once editorOpen flips false.
The existing extension test fabricated the null active-changed event, so
it passed despite the registry never emitting it — that gap is why the
bug shipped. Add a registry test that drives the real close() path, plus
a slot_host widget test asserting the split (drag handle) drops out and
the primary pane fills the column.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Make the localization usable: an app.locale select (English / Nederlands) in
the Appearance category, applied live by root_shell — it parses app.locale and
calls i18n.setLocale on boot + on settings change (setLocale is a no-op when
unchanged). nl_NL registered in availableLocales. The ui-design skill now
documents the locale-dir config and the rule to design for string-length
variation (translations run ~20% longer; never hard-size to the English label).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Migrate ClideTheme.of(context) reads across the widget, feature, and shell
layers to the unified ClideSettings.theme.of(context) facade (D-101), so
theme/i18n/fonts/settings share one widget-facing entry. The facade delegates
straight to ClideTheme, so behaviour is unchanged — goldens are unmoved.
The low-level theme provider keeps its direct ClideTheme.of: the facade is
built on it, and the two kernel sites (ClideTheme's own definition + the
panels drag-resize widget) stay direct to avoid a widgets→kernel import cycle.
Dead controller.dart/kernel.dart imports left by the sweep removed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Migrate widget i18n reads from ClideKernel.of(context).i18n to the unified
ClideSettings.i18n.of(context) facade (D-101), so settings/theme/fonts/i18n
share one widget-facing entry point. The facade delegates to the same I18n
service, so behaviour is unchanged. Infra plumbing (KernelServices /
ExtensionManager fields, the slot-host Listenable.merge) keeps its direct
kernel reference.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Introduce ClideSettings, a single widget-facing facade for the app's live
user preferences, namespaced by concern: ClideSettings.fonts.monoOf(context)
/ .fonts.uiOf(context), ClideSettings.theme.of(context),
ClideSettings.i18n.of(context). "Plumb once, use many."
Fonts are carried by a new root-provided ClideSettingsScope (resolved from the
font settings in root_shell, rebuilt on change); theme and i18n delegate to
their existing live providers (ClideTheme / the I18n service) so there's one
source of truth and their many consumers migrate incrementally rather than in
a big-bang. Reads outside a scope fall back to the bundled font defaults, so a
widget renders without a provider (isolated tests).
Migrate ~93 monospace-font call sites across 33 files from the clideMonoFamily
const to ClideSettings.fonts.monoOf(context) — pure refactor, identical family
when no override is set. 11 context-less helper sites (markdown static spans,
a few top-level/static builders) keep the const for now and are tracked in
T-472. Records D-101; updates the ui-design skill's font-family rule.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Vendor Inter (variable + italic, OFL) under assets/fonts/inter/ and make it
the default application UI face, replacing Josefin Sans (which stays bundled
as a selectable option). pubspec font family + licenses.yaml entry per D-42.
Settings → Appearance gains a UI-font select (Inter / Josefin Sans). The root
DefaultTextStyle reads app.ui.font (kUiFontSettingKey) over the default and
re-applies live on settings change, so a pick takes effect immediately. Bump
the default UI weight to w400 — Inter reads better at Regular than Josefin's
Light.
The monospace picker is deferred to T-471: clideMonoFamily is a const at ~38
call sites (not inherited), so a live mono picker needs the family routed
through context first.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Under the Vim preset, `:` opens a transient one-line ex overlay running a
fixed v1 table; ZZ runs :wq directly. Completes the last built child of the
T-403 cross-pane vim layer (T-405 part 2 gt/gT still open).
- ExLineController + parseExCommand grammar + editor-targeted executors
(lib/kernel/src/ex_line.dart); the overlay (lib/widgets/src/ex_line_overlay
.dart) reuses the quick-open chrome, mounts in the root_shell Stack, and
publishes the exline.open scope flag. Unknown commands flash + stay open;
with no active buffer every command no-ops (2026-06-13 decision).
- :q closes the active tab via editor.close on its id — the registry promotes
the next buffer and the split self-collapses on the last (2026-06-12
decision); :w/:wq/:x/ZZ save (+close) the active buffer.
- :e <path> seeds quick-open (new QuickOpenController.open(seed:)); :N adds the
editor.goto-line IPC/CLI verb (reuses _offsetForLine). Goto needs caret sync:
EditorController now handles editor.selection-changed and the editor view
moves the caret on a selection-only change.
- `:` and ZZ are typed intents; the editor matcher and PaneKeyNav now bubble
unhandled typed intents to the app-root Actions, so they fire from any focus.
vim.yaml binds `:`, ZZ (shift+z shift+z), and Esc-dismiss.
Tests: parser/controller/executors, editor.goto-line daemon tests,
selection-changed (controller + view), full overlay widget test. make test
green; analyze + format clean.
Also files T-441 (drop bold from the ticket-id card label) and T-442
(sub-agent renders as 3 cards instead of one bundle) under the T-276 UI epic.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Desktop/dock-launched clide inherits a sparse PATH (no ~/.local/bin, brew,
nvm, …), so pql/git/claude and PTY tools went missing. T-347 fixed only the
toolchain/pql path on Linux; env.dart's expander was still macOS-only and
claude/PTY/git used the raw PATH — the breakage recurred per spawn site
because there were three divergent expanders.
Consolidate into one resolver (lib/src/env/shell_env.dart):
- primeLoginShellPath(): probe the user's real login shell once at startup
(`$SHELL -l -c`, sentinel-framed, bounded timeout, graceful fallback to the
process PATH). Captures the user's actual PATH, not a hardcoded guess.
- expandToolPath(): the canonical merge (moved from toolchain_paths, which
re-exports it for its tests) — unions the well-known user/local bin dirs.
- resolvedToolPath(): currentSearchPath() + expandToolPath, the single call
every spawn site uses.
Routed through it: PTY children (registry.dart now overrides PATH), git
(env.dart → operations.dart), the toolchain probe (toolchain_paths), and
hosted claude (agent_bootstrap). Primed in main.dart's !kIsWeb boot. Deleted
the macOS-only env.dart copy and the cli_install copy.
Tests: new shell_env_test (probe + every fallback + merge); env_test and
cli_install_test updated to the consolidated surface. analyze clean, web wasm
build still green, make test green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`flutter build web --wasm` had been broken since the tree-sitter/PTY dart:ffi
pivot. Per D-100 (resolving Q-50: keep the web "happy accident" alive), every
native binding now sits behind a `dart.library.ffi` conditional import with a
graceful web stub. Desktop builds are unchanged — no fidelity loss; the web
target degrades (no terminal, native git, or syntax highlighting).
Discriminator is `dart.library.ffi`, not `dart.library.io` — dart2wasm provides
dart:io, so FFI is the only blocker.
Fences:
- PTY: pty_session → pty_backend_io / pty_backend_web (stub throws).
- tree-sitter: pure types → syntax_result.dart; tree_sitter_service is now a
facade over _ffi/_stub; tree_sitter_boot_io/stub fences TreeSitterLib.init().
- watchdog: watchdog_windows_stub (all -1 sampler).
- claude ABI probe: native_abi_io/stub (was `dart:ffi show Abi`).
- testmode fd-check: fd_check_io/stub.
Also dart2js-safe: the 64-bit FNV literals in session_naming.dart + paths.dart
(the dual JS fallback rejected them) — split into 32-bit halves, dropped a
no-op 64-bit mask. Desktop/wasm hash values unchanged.
CI: added a `web-wasm` job (flutter build web --wasm) so the fence can't rot.
Two FFI-constructing tree-sitter tests import _ffi.dart directly (the analyzer
resolves the conditional facade to the stub branch).
Verified: `flutter build web --wasm` → built; `flutter analyze` clean;
`make test` green. Full Playwright e2e harness wiring is the tracked follow-on.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The boot-time toggle (CLIDE_LOG / app.log.level) existed; this makes it
adjustable at runtime, from the two surfaces that fit — D-6 parity — and NOT a
Claude-composer slash command (log verbosity isn't a Claude-session concept).
- Output dock Level chip: was a view-only filter; now also drives the kernel
Logger.minLevel and persists app.log.level, so the choice is real (changes
what's captured) and survives restart. Initialized from the logger's current
level so it reflects a CLI change. A null callback keeps it a pure view
filter (tests / no kernel).
- `clide log level [<level>]` (lib/src/daemon/log_commands.dart): no arg reports
the level + vocabulary; a valid level sets the live logger and persists;
unknown → bad_arg (code 64), logger untouched. Wired in buildDispatcher with
the kernel logger + settings (captured post-boot).
Tested: the command (get/set/case-insensitive/bad-arg, live + persist) and the
controller (initial level, callback fires, same-level no-op, filter-only when
unwired). Coverage gate 95.05%.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The T-436 soak run came back green but produced no conpty-kill-crumbs artifact:
IsolateCrumbFile opened its append handle without creating the parent dir, and
the standalone orphan probe points it at a fresh CLIDE_LOG_DIR that nothing
else had made — so openSync failed, the crumb file was disabled, and no crumbs
were written. In the app this was masked because FileLogSink already creates
logDirectory() at boot. Create the parent ourselves (no-op when it exists).
Verified by the bundle-smoke artifact, which DID capture a real watchdog sample
(threads=31, handles=25, rssMB=175) because the release app makes the dir.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Closes the observability loop: the log + breadcrumb + watchdog files are now
collected by CI so a wedged run leaves downloadable evidence instead of
nothing.
- logDirectory(): CLIDE_LOG_DIR overrides the per-platform default, so CI can
point the logs at an uploadable workspace dir (and tests at a temp dir).
Now takes an injectable env map; tested.
- test_app.dart: when CLIDE_LOG_DIR is set, the testmode harness tees its
logger to a FileLogSink + spawns the watchdog (off by default — normal
run-testmode keeps the stderr-only path, no isolate). _say breadcrumbs each
test into the file.
- conpty_orphan_probe.dart: with CLIDE_LOG_DIR set it passes a verbose PtyLog,
so when soak-conpty-kill.ps1 force-kills the parent, the reader/waiter
isolates' LAST crumb is fsynced to disk — naming what the wedged isolate was
doing at the instant of death.
- bundle-smoke job: runs the real release app with CLIDE_LOG=debug +
CLIDE_LOG_DIR, uploads clide-logs (watchdog heartbeat/sample + FileLogSink)
in an always() step.
- windows-soak kill-probe job: sets CLIDE_LOG_DIR, uploads the FFI crumbs.
Coverage gate 95.08%.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The freeze hypothesis is a wedged FFI call — a reader isolate blocked forever
in ReadFile, a waiter in WaitForSingleObject, Isolate.kill unable to interrupt
either. To NAME the wedge after a power-cycle, each backend now drops a
breadcrumb before/after every risky syscall.
- pty_log.dart (new, Flutter-free, tested): PtyLog — an injectable, no-op-by-
default breadcrumb hook for the MAIN isolate (wired to the kernel Logger,
source 'conpty'/'pty' = an eager FileLogSink source) — and IsolateCrumbFile,
which the SPAWNED reader/waiter isolates use to open their OWN append handle
and flushSync per line, so a wedged isolate's last crumb survives even a
frozen main isolate (the whole point). Bounded by a truncating size cap.
- native_pty.dart + windows_pty.dart: crumbs around posix_spawn/read and
CreatePseudoConsole/CreateProcessW/ReadFile/WaitForSingleObject; the reader/
waiter isolates carry a sendable crumb path + verbose flag. Per-syscall crumbs
only at debug/trace; lifecycle crumbs always.
- Wiring: startPtySession → PaneRegistry → buildDispatcher build the PtyLog from
the kernel Logger + a crumb file under logDirectory(); verbose follows the log
level. Default everywhere is PtyLog.none — zero behaviour change off the wire.
Tested: PtyLog/IsolateCrumbFile units (cap-truncation, append, no-op) + an
end-to-end real-PTY test asserting the reader isolate writes its own crumbs
('reader started' / 'read -> n=' / 'reader exiting'), which validates the
identical Windows structure that can't run here. Coverage gate 95.10%.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First increment of the observability epic (T-425), the productive pivot after
the ConPTY freeze refused to reproduce on CI: if we can't reproduce it, make
the next occurrence leave evidence.
- FileLogSink (lib/kernel/src/file_log_sink.dart): synchronous, crash-survivable
LogSink. Appends each record as one JSON line to a size-rotated file; fsyncs
warn/error + risky-source (pty/ffi/conpty/watchdog) records immediately so the
last breadcrumb is on disk before a hard death, batches the rest on a timer.
Never throws. Flutter-free → unit-tested under dart test against a temp dir.
- logDirectory() (paths.dart): persistent per-platform log dir (LOCALAPPDATA /
~/Library/Logs / $XDG_STATE_HOME) — durable across reboot, unlike the
ephemeral socketDirectory.
- resolveLogLevel() (log.dart): the requested dev/prod toggle. CLIDE_LOG
dart-define → CLIDE_LOG env → app.log.level setting → warn(release)/info(debug).
Lenient parse; an invalid source falls through.
- Boot wiring (facade.boot + main.dart): FileLogSink leads the sink chain (so a
crash records before the volatile stderr/ring sinks) and the resolved level
sets Logger.minLevel.
Tests: FileLogSink (JSON shape, error/stack, rotation cap, append-across-restart,
timer-cancel), resolveLogLevel precedence + fall-through, logDirectory per-OS.
Coverage gate 95.11%.
Follow-ups under T-425: live toggle CLI/command/chip (T-433), FFI breadcrumbs
(T-434), watchdog isolate (T-435), CI artifact wiring (T-436).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A testability audit of the ignore span found the original comment overclaimed
("none of it can execute on Linux") and credited the wrong test file. Reality,
verified fragment by fragment:
- The span is excluded at FILE granularity but is not 100% syscall — _Coord /
_StartupInfoExW struct packing and write()'s empty-guard are pure transforms
that could be unit-tested on Linux if extracted from the binding-touching
methods. Tracked in T-431 (also covers the mirror gap in native_pty.dart's
POSIX marshalling).
- The pure helpers are tested by windows_pty_args_test.dart (not _test.dart).
- The FFI path's BEHAVIOUR is validated on windows-latest (real ConPTY spawn),
but windows.yml collects no coverage — so there is intentionally no line-
coverage metric for this span anywhere; correctness rests on that functional
suite + the VM soak, not on coverage.
Comment-only; no code or coverage change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
windows_pty.dart's Win32 structs, kernel32 bindings, and the WindowsPty
session (CreatePseudoConsole / CreateProcessW / WaitForSingleObject) cannot
execute on the ubuntu-latest runner that produces the coverage report —
there is no kernel32 to bind. They were being counted as 214 uncovered
lines (the file sat at 13.7%), dragging total line coverage to 94.17% and
failing the 95% gate for code the gate's platform structurally cannot reach.
Wrap exactly that FFI span in `// coverage:ignore-start/end` (flutter test
--coverage honours the markers, verified). The pure, platform-agnostic spawn
helpers — resolveExecutable / quoteArg / composeEnvironmentBlock — sit after
the ignore-end and stay measured (covered by windows_pty_args_test.dart on
every platform). FFI-path behaviour is validated on the Windows runner
(windows.yml) and end-to-end in the Windows VM soak (tools/windows-verify/).
Gate back to green: 95.16% (floor 95%). Floor unchanged — the exclusion
restores the denominator to Linux-runnable code rather than lowering the bar.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Pre-Windows-VM hardening — the parts validatable on Linux, leaving the
unrunnable FFI (Job Object, T-424) for the VM session:
- Clamp PTY cols/rows to >= 2 in both backends' spawn + resize (new
pty_size.dart). A 1-column ConPTY makes conhost spin emitting CRLF
(microsoft/terminal#19922); 0 is invalid on both platforms.
- ci/test.sh: --timeout 60s on the dart-test pty line (matches the flutter
lines) so a wedged ConPTY reader fails fast instead of hanging the run.
- Make windows_pty.dart's pure helpers public + testable off-Windows:
quoteArg (MSVCRT quoting), composeEnvironmentBlock, and resolveExecutable
(now takes an injectable existence probe). New windows_pty_args_test.dart
+ pty_size_test.dart give 15 cross-platform assertions over the trickiest
Windows logic the on-Windows smoke suite can't reach off-platform.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
