Overlaps master on the preset e2e, the web scaffold, and the generated
config-catalog triplet; all merged textually, so the catalog is regenerated and
its pairing hashes re-recorded rather than trusted.
Adapt to two contract changes master introduced:
- The generated Remote face now wraps every business result in
RemoteResult, folding carrier failures into an ok:false branch instead
of rejecting. The controller reads that envelope at its three call
sites and maps a carrier failure onto the same settled shape the
controls already render; three specs cover the new branch.
- Client packages split their tsconfig into host and client halves, and
the host aggregate now compiles any test not named *.client.spec.*.
Rename this package's specs to the client convention and drop the
../connection project reference, which pointed at a solution file that
no longer carries the client sources.
Keep master's mount loop with its rollback-on-failure in api-remotes and
add messageFeedbackRemote to it.
The source-through-tsx fallback in windowsAclRunnerInvocation was only
reachable while sandbox-windows-acl/lib/runner.js was absent, so the
per-file 100% gate failed wherever a gate had already built it. In
ci-primary and ci-windows-complete the typert-contracts gate runs
build:lib:host ahead of test:coverage in one workspace, leaving both
fallback statements uncovered; the sharded coverage job never builds
first and stayed green.
Cover the fallback through an injected entry path that is guaranteed not
to exist, and keep the uninjected case asserting only the verdict that
holds under either arm. Verified at 100% statements/branches/functions/
lines with lib/runner.js both present and absent.
The locale plugin's inject list gained 'remote' and 'settingsScope' on
2026-08-10, but the built-bundle harness mounted it without providing
those services, so the locale fiber never started and the trajectory
plugin's apply never ran — the ring assertion saw no 'trajectory' entry.
Provide the same stubs ui-layout's apply bench uses.
master cut 0.0.1-rc.2 while this branch was open. The version bump touched
every existing package but not this new one, so check-workspace-constraints
rejected the mismatch and took the required "all checks passed" job down
with it.
A test file under packages/client now says which face it covers:
`*.client.spec.{ts,tsx}` and its `*.client.{ts,tsx}` helpers belong to the
Client aggregate, `*.host.spec.ts` to the host aggregate. The carrier's four
node-half specs take the Host suffix.
The two suffixes are mutually exclusive, so each aggregate excludes the
other's and both keep one broad test glob: `exclude` wins over `include`, and
`packages/client/**` no longer has to be excluded wholesale from the host
program with per-file `files` entries carved back out of it. A Host-face spec
that reaches only Host source therefore needs no cross-face project
reference, which the split-project rule rejects.
vitest still discovers every file through `**/*.spec.{ts,tsx}`.
Re-record the sidebar regions that the redesign intentionally changed:
group session counts are gone, search is a collapsed header action, and
grouping lives in the View options menu. Session rows also follow the new
recency ordering, so their tree order updates with the counts.
The sidebar redesign removed group session counts, collapsed search into a
header action, and moved grouping into the View options menu. Update the
browser scenarios that waited on the old surfaces:
- Wait on the Ungrouped bucket row instead of the removed session-count text.
- Expand the collapsed search control before filling the renamed input.
- Open the View options menu instead of the removed Group by button.
- Assert the persisted grouping under the new dsh.workspace.view.v4 key.
- built-boot: assert the fixture workspace group row instead of its count.
Resolve the note-erasure race the review found: a control that rendered
before the first list read held no item and passed note=undefined, so
switching a rating silently dropped the stored note. The controller now
owns note resolution and toggle-vs-retract, deciding from the committed
item inside the serialized mutation, and clearNote expresses deletion.
Also from review:
- serialize the reconnect re-read behind queued mutations (resync), so a
list reply cannot resurrect a version a newer mutation replaced
- re-check disposal after the seeding read, so an unloaded fiber never
reaches the wire
- drop Object.freeze on Maps, which does not prevent set/delete
- declare the @deepseek-ai/dsh-client-connection dependency it imports
- surface a failed list load in the controls
- drop the /client value exports that had no consumer
- state the per-turn render scope in the README and subsystem pages
- align the package version with the root
Local run of check:ci:windows-complete (the windows-native gate) on
latest master surfaced five Windows-only failures, all unreachable by
current CI because the native windows job is disabled and the wine gate
only covers build+site.
- install-lefthook/translation-pairing-merge specs junctioned the real
scripts/ and tsx package into fixtures; Windows recursive deletion
(Node rmSync and git worktree remove) follows MOUNT_POINT junctions and
deleted the repository's own directories mid-run. Fixtures now unlink
their reparse points before any recursive removal (shared helper in
scripts/test-fixture-cleanup.ts).
- workflow-workerthread spawned its worker with an empty env; on Windows
os.tmpdir() then degrades to the literal relative path undefined\temp,
so tsx wrote its transform cache into a cwd-relative undefined/
directory inside the repo. The worker env now injects the host temp
path on win32 (workerSpawnEnv, platform-parameterized and unit-tested
on both arms).
- workspace-context spec did not stub USERPROFILE (win32 homedir) or a
set DSH_HOME, leaking the developer machine's real ~/.dsh/AGENTS.md
into discovery.
- ui-trajectory client-bundle spec mounted the built artifact without the
remote/settingsScope provides the locale plugin needs, so the plugin
never activated and no view registered.
- subagent temp-fixture cleanup lacked the maxRetries Windows handle
release needs under load (EPERM); added retries to the three affected
specs and the fixture-cleanup helper.
The preset-roster loop added by this PR names the \cordis\ preset id in
apps/cli/tests/windows-shell.spec.ts; the bare token is product data like
the existing agent-preset allowlist entries, not a package reference, so
the rescope-vendor hygiene check must not rewrite it.
Routes koffi allocations through a branded NativePtr helper (koffi's TS
types are any), binds the creationTime callback instead of passing the
unbound method, and braces the no-op signal assertions.
Adds the tool-pwsh-persistent README trio, documents the pty-local
shellDialect and the subprocess-local Windows inspector (console-wide
signalling, pseudo foreground groups, taskkill teardown) in both
languages, updates the tool-pwsh and persistent-pty notes in place, and
moves the pwsh-persistent-pty design note to implemented with the
shipped Decision and Consequences.
Adds @deepseek-ai/dsh-tool-pwsh-persistent, the mirror of
tool-bash-persistent for PowerShell: one owner-scoped persistent pwsh
per agent, an Invoke-Expression wrapper with backtick-escaped bodies and
exact native exit codes ( reset, \True fallback, catch to 1),
PSReadLine-echo tolerance (the echoed wrapper is stripped from captured
output and can never fabricate completion), and the same
timeout/cancel/exit reset semantics with pwsh-flavored diagnostics.
The minimal preset now gates its persistent shell stack by platform with
the #2234 disabled interpolation: the bash rows mount on POSIX and the
pwsh rows (pty-local shellDialect pwsh + the new tool) on win32, keeping
exactly one persistent shell per host. windows-shell.spec pins the
per-platform roster; the real Loader composition proves cwd/env
persistence, multiline and here-string commands, large-output clipping,
and exit/reset over a real ConPTY pwsh.
Adds shellDialect ('bash' | 'pwsh') to the local PTY backend. The
effective shellPath/shellArgs resolve per dialect (pwsh through the
shared dsh-pwsh-local resolver, bash defaults unchanged), the child
environment drops bash-only PS1/PROMPT_COMMAND markers and adds
NO_COLOR for pwsh, and pwsh startup bootstraps the prompt function that
emits the shared OSC 133;D + BEL marker, waiting (across follow-up
sends) until the controlled prompt is actually visible so the
banner-to-prompt gap cannot settle startup early. Bash behavior is
byte-identical; the real-pwsh suite exercises persistent state and
secret scrubbing on Windows.
createProcessInspector now returns a WindowsProcessInspector on win32
instead of throwing: Toolhelp32 tree enumeration with GetProcessTimes
start identities, the shell pid as a pseudo foreground group, taskkill
tree signalling, and inspector-verified Windows teardown (node-pty
signal kills throw on Windows, and externally taskkilled shells may
never fire its exit notification, so the handle settles \done\ from the
verified absence). subprocess-local and pty-local suites now run on
Windows with platform gates; the koffi-backed inspector joins the
windows-only coverage exclusions on Linux and is fully covered by the
windows-native lane.
Also flips vitest.config so subprocess-local and pty-local sources are
coverage-required on win32, and adapts the spawn/terminal suites to run
natively there (node-translated shell commands, injected POSIX group
paths, taskkill signal semantics).
Every generated method resolves to RemoteResult, so the plain-Node script
must read the business value through `.value`: the CAS ref it passed on to
goals/edit was undefined, which the client codec rejected before the request
left. The invalid-payload case keeps its try/catch — a codec-rejected
argument still throws at the Client Remote face rather than folding into the
error branch.
Review follow-ups that are logic rather than documentation:
- rpc.schema.ts had lost the credential-rejected branch while api-proxy.ts
still returns that code, so a legitimate business error failed the
client's response parse. Restore the branch and assert it.
- rpc-schemas.spec.ts had lost the workspace list, archiveSession and
insertSessionBefore cases along with the command schemas; those routes
still ship, so restore their coverage.
- An omitted SRC field is now recognized by an absent key instead of an
undefined value, which makes the allowance assertExactArguments already
granted reachable; an explicitly undefined field stays invalid input. A
weak descriptor's undefined result rides the wire as an absent value,
matching the envelope removal.
- The chooser unmounts an already-created backend when the surface entry
fails to load, and no longer reverses the captured id array in place.
`command.execute` no longer exists on the API Proxy, and the fake's own
handler went with it, so the case only reached a 404 body. The same carrier
behaviour is asserted on live routes by its `session.search`,
`subagent.prompt` and `host.pickDirectory` siblings.
Matching the full `RemoteFailure` keeps the assertion inside the typed result
and drops the `expect.stringContaining` placeholders, whose `any` return the
lint rule rejects on assignment.
`@deepseek-ai/cordis` returns to the peer and dev dependencies of
`dsh-client-ui-model` and `dsh-client-ui-skill`; every harness package declares
it, and the client type-assembly rename dropped it from both. The commands
manifest ships `src` alongside its generated typert faces, matching the
`./src/*` export it already declares.
knip gains the two directory-picker surface workspaces, whose specs are `.tsx`
and matched no default pattern, and ignores `zod` in the commands workspace:
that dependency belongs to the generated Remote and Host faces in `lib/`, which
knip never scans. The unused `dsh-client-test-runtime` and
`dsh-client-connection` dev dependencies are gone; the picker surface's own spec
never imported the former, and ui-settings reads the carrier's types through the
Remote assembly now.
imccyu