The ::warning:: wrapper kept the lane job green on failure — honest
about not gating merges, but a Windows regression was visible only as
an annotation buried in the run summary. GitHub has no yellow job
state, so the choice is green+annotation (too hidden) or a red X on a
non-required job (visible, still non-blocking). Take the red X:
job-level continue-on-error, plain 'Run gates' step, one less wrapper.
The step stays on the runner's native pwsh — never shell: bash — per
the MSYS-parent leak recorded in the windows-support RFC.
Also remove the temporary Windows-only failing demo test that
exercised the wrapper's annotation path (REVERT ME commit a496b9ae).
The lane-split merge moved the Run gates step to `shell: bash`, which
broke it twice over. First, GHA's bash shell runs with -e, so a failing
gate aborted the step before the ::warning::/exit 0 lines — the lane
went red X instead of the intended yellow warning. Second, and worse,
Git Bash as the PARENT of the gate run leaks MSYS environment into the
suite itself, producing 8 real test failures the pwsh-launched runs
(and the DSec VM runs) never saw:
- bash exports PWD; the MSYS runtime rewrites it to Windows form for
native children, dsh-bash-local's adaptEnv passes it through, and the
executor's MSYS bash adopts it — `pwd` prints `D:/a/...` where the
tests (and the executor's MSYS dialect) expect `/d/a/...` (7 tests).
- cygwin enables SeBackupPrivilege on the runner's admin token; children
inherit the enabled state, and libuv's FILE_FLAG_BACKUP_SEMANTICS read
opens then pierce the dwShareMode=0 lock the jsonl EBUSY test holds —
loadLive resolves instead of rejecting (1 test).
Evidence: run 28918325498 (pwsh step, pre-merge) failed only the two
hooks dispose tests since fixed by f8fd8c00; run 28921741006 (bash
step) fixed those and failed exactly the 8 above, with zero relevant
source diff between them.
Fix: run the wrapper in pwsh — a native command's failure doesn't abort
pwsh, so $LASTEXITCODE capture + ::warning:: + exit 0 works without an
errexit dance, and the gates start from a native Windows shell as they
do everywhere else Windows CI has been validated.
Docs: the windows-support RFC drops the stale continue-on-error wording
(replaced by the warning wrapper) and records the launch-environment
limitation — native shell required today; making an MSYS parent a
supported launch environment (PWD scrub in adaptEnv, privilege-explicit
tests) is a future improvement direction.
The Windows CI was a single job running the full ci-windows inventory
(check:ci:windows), while Linux splits into 5 lanes (static/lint/coverage/
snapshot/artifacts) per the parallel-gates RFC. The single-job shape was a
transitional artifact from when Windows CI was added as a non-required
observation lane; its rationale ('keep gate parallelism modest so coverage
is not starved') conflated run-gates intra-job concurrency
(DSH_GATE_CONCURRENCY) with GitHub job fan-out — orthogonal concerns.
Split Windows into 4 lanes mirroring Linux (snapshot absent: its goldens
are Linux-recorded and self-skip on Windows). Each lane is a separate
GitHub job so a Windows regression is attributable to one lane, not
buried in one job's log. Concurrency is NOT throttled versus Linux: the
lane is non-blocking (continue-on-error), and the observational stance
is to actively expose Windows-arm issues rather than hide them behind
reduced parallelism.
- scripts/run-gates.ts: add ci-windows:static/lint/coverage/artifacts
modes; ci-windows (full inventory) is retained as the local one-process
entry, symmetric with Linux's ci-primary.
- .github/workflows/ci.yml: windows job becomes a matrix over the 4 lanes.
- package.json: check:ci:windows:{static,lint,coverage,artifacts} scripts.
- AGENTS.md + windows-support RFC: document the per-lane, non-blocking,
unthrottled stance.
Verified: scripts/caohuanqi-private/run-ci.py --windows (full
check:ci:windows) — all gates green except the known hooks-claude
bridge.spec waitFor timeout (pre-existing Windows subprocess-timing
flake, unrelated).
Add a Windows CI job alongside the existing Linux checks. Runs the full
test suite (without the Linux-only coverage gate) plus typecheck, lint,
doc-sync, build, hygiene, and demo smoke under PowerShell. Developer
Mode is enabled via registry for symlink support (fs-local tests,
verify-node-next-types).
Per the windows-support RFC transition plan: step (2) — non-required
Windows CI job to observe stability.
Runs `pnpm install` + `pnpm run build` (tsc -b + tsdown) on windows-2025,
and is listed in all-checks-passed `needs` so a Windows build regression
cannot land silently. Windows path/shell support is still partial, so this
lane covers the build surface only; tests and gates are not run here yet.
The existing assertion-only executable smoke proved selected outputs but could not detect drift across the integrated Python SDK, JSON-RPC notification stream, and persisted session shape.
Keep this separate from ACP snapshots because it must launch the actual platform-native packaged executable through the Python SDK. The deterministic model drives Cordis dynamic tool mounting, a Code Mode worker dispatch, direct spawn delegation, workflow-worker delegation, plugin disposal, and the parent/child persistence lineage.
Commit four portable goldens for the SDK result and three JSONL logs. Normalize timestamps, temporary paths, opaque session and agent identifiers, and bulky request headers while retaining ordering, tool names and arguments, header deltas, results, lineage, and final responses so all native build legs compare the same behavior.
Run the comparison in the label-gated executable build workflow and document the current coverage in the paired implemented RFC.
Address the three ds-review-bot warnings on #253:
- An empty DSH_CORDIS_CONFIG now counts as absent when deciding whether
to inject the bundled default config, matching the runtime bin's
config-discovery semantics.
- The injection moves from DeepSeekHarness into HarnessClient.start(),
so the low-level client's default bundled launch also boots without
callers duplicating the env setup.
- The bare single-file exe artifact ships inside a tar.gz like the
Python bundle: upload-artifact's zip transport drops the executable
bit.
sandbox-acp-agent's whole surface moves into examples/acp-agent following
the example's existing multi-variant shape (the code-mode/both-mode
overlays): sandbox.cordis.yml + its replay overlay, the four snapshot
scenarios (config-options / mode-switching / escalation-approved /
escalation-rejected) as a 'sandbox' header class over per-scenario
configPath — goldens byte-identical, zero re-recording — and
escalation.e2e.ts unchanged apart from the config path. One ACP example
remains; demo:sandbox-acp keeps its name and boots the variant. References
(both RFCs, group/package READMEs, the examples table and test map, the
e2e workflow comment) now point at the variant.
The three-entry cordis.yml (dsh-sandbox-local + dsh-bash-sandbox at a
read-only default + dsh-approval) served over ACP: the first live approval
composition. Recorded snapshot scenarios pin the wire end to end —
config-options advertisement, the mode-switching arc as the suite pinned
header (both switches, the prompt-section delta, one changed-by-the-user
notice per knob, a confined write landing under the switched mode), and
both escalation branches over scripted permissionAnswers (a grant runs
confined under workspace-write; a rejection executes nothing and pins the
fail-closed text). The with-key escalation e2e drives a real model +
real runner + the real bridge answerer, world-verified; ci.yml snapshot
lane and e2e.yml install bubblewrap so the confined replays actually
execute.
Both RFCs move to implemented/ (Decision/Consequences form, deferred
phases tracked in their own sections), with every cross-reference flipped.
ctx.sandbox (dsh-sandbox): confine(argv, policy) returns the argv to spawn
instead — wrapped so the process and its children run confined — plus the
enforcement completeness and the backend denial/runner-failure dialects;
no usable backend throws the fail-closed SANDBOX_UNAVAILABLE. Policy rides
per call. dsh-sandbox-local selects by platform and caches the verdict:
multi-candidate chains probe FUNCTIONALLY in preference order (Linux:
bwrap → the registry-installed node-addon-landlock-run launcher), a sole
candidate is selected unprobed (darwin: sandbox-exec/Seatbelt) and fails
closed at execution via runnerFailureSignatures; win32 is a reserved empty
chain. Profile parity is honest per backend (documented temp-area and ABI
differences; enforcement full|partial is a structured result fact).
CI: the sandbox-e2e matrix proves real-kernel confinement per rung (bwrap,
Landlock per architecture through the registry-installed launcher,
Seatbelt), failing on a silent all-skip; the packed-install rehearsal
installs the launcher family from the registry and asserts the binary
executable apart from kernel enforcement.
A single stable required check that needs every other job in ci.yml, so
branch protection no longer enumerates matrix leg names that change as
lanes and node versions evolve. if: always() keeps the job running when
a dependency fails (a skipped required check would count as passing);
any non-success result — failure, cancelled, or skipped — fails it.
Type-aware ESLint loads every package tsconfig through the project
service and peaks at ~3.4GB RSS. The default V8 old-space ceiling
(~2GB) OOMs it (FATAL ERROR: Ineffective mark-compacts near heap
limit, exit 134) on both node 24 and 26. Set NODE_OPTIONS with an
8GB ceiling for the Lint step, comfortably above the peak.
BLOCKER — the published lib/bin.js (stdio + acp) was exercised only via tsx
(demo:* / the src/bin.ts smokes); the built artifact under plain `node` was
unguarded. Root-cause on the BUILT bin:
1. Settle race: boot() returned once loader.create() registered the include
ENTRY, but the include loads its child plugins asynchronously — so boot()
(and main()) resolved while the app plugins (stdin reader, agent loop, ACP
bridge) were still mounting. A CLI with no attached handles yet exits 0
silently, and a load error surfaces as an unhandled rejection AFTER boot.
Fix: `await ctx.loader.await()` after create() — settle the whole tree.
2. Config-path robustness: hand the include the config's ABSOLUTE file:// URL
so resolution never depends on ctx.baseUrl / can never fall back to cwd.
Both bins fixed identically. NOTE: the cordis Loader resolves a config's bare
plugin specifiers via its internal module loader, active only under
`node --expose-internals`; the bin cannot add a node flag itself, so this is
documented in the bin JSDoc + both package READMEs (the demos already comply).
The repo `examples/*/cordis.yml` are tsx-only artifacts (workspace plugins
resolve through the tsconfig paths map, not node_modules), so they are not a
valid plain-node bin target — the smokes use a real-install-shaped temp dir.
Fail loud on a load failure: boot() previously exited 0 SILENTLY when a config
path's directory does not exist — the include plugin fails to IMPORT, the cordis
Loader catches+LOGS it and leaves the entry with no fiber (no rejection), and
`loader.await()` does not rethrow (EntryTree.await uses Promise.allSettled). Fix:
boot() now calls assertEntriesLoaded(ctx) after the tree settles and throws on
any entry with no fiber, so a typo'd config dir exits non-zero with a clear
message. main() also installs an unhandledRejection guard (installFailLoud) that
replaces Node's stack dump with a single labelled stderr line for the
companion case (a missing config FILE in a real dir, whose include-init throw
surfaces as a rejection Node already exits non-zero on). Regression tests added
to both built-bin smokes (missing dir + missing file → non-zero exit + stderr);
verified the missing-dir test fails on the pre-fix bin.
Built-bin smokes (the reviewer's ask): packages/ui/{stdio,acp}-agent/tests/
built-bin.e2e.ts run the REAL lib/bin.js under `node` (NOT tsx) in a temp
consumer dir, asserting the stdio echo round-trip / the acp initialize response
+ stdout purity, plus the fail-loud cases above. They build-gate (skip if lib/
absent) and run in a new ci.yml step after the build.
Issue 2 — packages/README.md + docs/architecture.md said "plugins depend on
interfaces, never on the concrete loop", but dsh-agent-core imports the concrete
dsh-agent-loop. Scope the rule to EXTENSION plugins and carve out the sanctioned
COMPOSITION/bundle exception (dsh-agent-core composes the concrete spine); note
it in the implemented RFC too.
Issue 3 — examples/acp-agent/tests/acp.snapshot.ts fixture-guard claimed
no-model scenarios need no session.jsonl, but runScenario() always boots
llm-replay with the session.jsonl path and loadReplayScript() throws when it is
absent. Require session.jsonl for ALL scenarios (no-model ones ship a
header-only fixture) and rewrite the comment to match reality.
The CI "Demo smoke test" step booted the deleted examples/echo-agent/start.ts
(removed when the boot glue moved into the dsh-stdio-agent bin), so CI failed on
node 24 + 26 while the local gates and e2e passed. Invoke `pnpm run demo:echo`
instead of hardcoding the boot path — that routes through the canonical demo
script, so the smoke can never drift from it again. The output assertions and
the .sessions/_no-cwd/main-session-*.jsonl artifact check are unchanged
(verified the new bin produces identical output).
Add a second axis to every RFC — its class (feature, bug-fix,
simplification, architecture, process, testing) — encoded in the path
as docs/rfc/{lifecycle}/{class}/file.md. The folder is the label, so
the closed set is enforced by structure rather than a parsed field.
Two new doc-sync gates back it:
- verify-rfc-classification: every RFC sits in a valid class folder and
the README index lists it under the matching lifecycle→class heading.
- verify-doc-refs: every docs/*.md path cited in a packages|examples TS
comment resolves — closes a drift class verify-md-links can't see, and
catches the four comment refs this reorg moved.
The README gains a Classification section explaining the taxonomy and
per-class index sub-sections. A self-referential process RFC records why
the scheme is path-encoded and gated.
Add scripts/gen-cordis-catalog.ts: a fully-generated docs/cordis-catalog/
events-and-services.md cataloging every cordis event (exact signature + @mode)
and ctx.<key> service (exact interface), modeled on gen-module-graph's
--write/--check freshness gate. The harness tier renders in full from the
interface Events / interface Context declarations and their JSDoc; the inherited
cordis-core/loader/hmr/timer surface renders tersely from a curated table.
The generator hard-errors on a missing @mode tag and on a tag that contradicts
a conclusive signature shape (a trailing next param is structurally a
waterfall). Signature blocks use a ts cordis-catalog fence that doc-typecheck
skips. Type tokens cross-link to the core-data-structures catalog.
This supersedes the hand-maintained event-taxonomy table: verify-event-taxonomy
is deleted and verify-cordis-catalog joins doc-sync. architecture.md keeps the
Event taxonomy heading (TOC anchor) but points at the catalog; the Service-map
role table stays. RFC, AGENTS.md @mode authoring rule, and dependent doc/skill
references updated. Negative gate tests cover the missing-tag and
tag/shape-contradiction paths.
Adds docs/rfc/implemented/2026-06-19-real-api-e2e-ci.md covering the rationale
for running the real-API e2e suite in a separate secret-consuming workflow, the
fork/Dependabot/secret threat model, the residual exposure of the pull_request
trigger, and what changes when the repo goes public. Indexes it in the RFC
README.
Also adds a SECURITY comment on the pull_request trigger forbidding a switch to
pull_request_target (an untrusted-code-with-secrets leak vector, especially once
public), pointing at the RFC.
imccyu
Huanqi Cao
Tianyi Cui
Yichen Jiang
kingwl
imccyu