dsh-plugin-wsl-env 0.8.0 → 0.10.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (54) hide show
  1. package/CHANGELOG.md +488 -12
  2. package/CONTRIBUTING.md +125 -62
  3. package/README.md +99 -68
  4. package/SECURITY.md +18 -2
  5. package/SUPPORT.md +26 -10
  6. package/agent/native/linux-aarch64/dsh-wsl-agent +0 -0
  7. package/agent/native/linux-x86_64/dsh-wsl-agent +0 -0
  8. package/agent/wsl-agent.sh +83 -16
  9. package/cordis.patch.yml +50 -35
  10. package/docs/ARCHITECTURE.md +203 -55
  11. package/docs/CONFIGURATION.md +185 -18
  12. package/docs/LIMITATIONS.md +125 -45
  13. package/docs/PARITY.md +68 -38
  14. package/docs/PEER-PARITY.md +103 -33
  15. package/docs/README.zh.md +105 -64
  16. package/docs/UPSTREAM-FSIO-EXPORT.md +159 -50
  17. package/docs/UPSTREAM-TERMINAL-TITLE.md +56 -14
  18. package/examples/preset-wsl-session-sandbox.cordis.patch.yml +178 -0
  19. package/lib/agent-protocol.d.ts +10 -4
  20. package/lib/agent-protocol.js +10 -3
  21. package/lib/agent.d.ts +84 -0
  22. package/lib/agent.js +167 -15
  23. package/lib/bwrap.d.ts +8 -0
  24. package/lib/bwrap.js +4 -1
  25. package/lib/fs-routing.d.ts +40 -4
  26. package/lib/fs-routing.js +80 -12
  27. package/lib/fsio-agent.js +1 -1
  28. package/lib/index.d.ts +134 -0
  29. package/lib/index.js +202 -9
  30. package/lib/native-agent.d.ts +124 -0
  31. package/lib/native-agent.js +198 -0
  32. package/lib/ports.d.ts +13 -2
  33. package/lib/ports.js +18 -5
  34. package/lib/preflight.d.ts +91 -0
  35. package/lib/preflight.js +117 -0
  36. package/lib/probe-cache.d.ts +12 -1
  37. package/lib/probe-cache.js +11 -3
  38. package/lib/sandbox-core.d.ts +70 -0
  39. package/lib/sandbox-core.js +187 -18
  40. package/lib/session-runtime.d.ts +161 -0
  41. package/lib/session-runtime.js +220 -0
  42. package/lib/session-scope.d.ts +124 -0
  43. package/lib/session-scope.js +191 -0
  44. package/lib/shell-env.js +10 -3
  45. package/lib/watcher.d.ts +25 -12
  46. package/lib/watcher.js +120 -32
  47. package/lib/workspace-files-route.d.ts +8 -4
  48. package/lib/workspace-files-route.js +42 -14
  49. package/lib/workspace-files-wsl.js +25 -8
  50. package/lib/wsl.js +10 -4
  51. package/locale/preflight/en.json +6 -0
  52. package/locale/preflight/zh.json +6 -0
  53. package/package.json +44 -27
  54. package/scripts/diagnose.sh +92 -0
package/CONTRIBUTING.md CHANGED
@@ -14,14 +14,22 @@ the conventions this repository uses.
14
14
 
15
15
  ## Toolchain
16
16
 
17
- The runtime package ships an empty `dependencies` object on purpose: it is
17
+ The runtime package ships no `dependencies` field at all, on purpose: it is
18
18
  loaded into the DSH process and everything it needs arrives as a peer package.
19
19
  Running the plugin from a checkout has no install step.
20
20
 
21
- Developing and running the gates do have dependencies. The 0.7.x tests import
22
- pinned `@deepseek-ai/*` packages — the composition and boot tests run the
23
- loader's real patch algorithm — so a fresh checkout needs one `pnpm install`
24
- before the gates pass.
21
+ Developing and running the gates do have dependencies. The tests import pinned
22
+ `@deepseek-ai/*` packages — the composition test runs the loader's real patch
23
+ algorithm, and the boot test starts a row through the real Loader — so a fresh
24
+ checkout needs one `pnpm install` before the gates pass.
25
+
26
+ Those pins are deliberate, and they move with the harness rather than on their
27
+ own: the `@deepseek-ai/dsh-*` peers sit at `0.2.1-rc.2`, `@deepseek-ai/cordis`
28
+ at `~4.0.5` and `@deepseek-ai/schemastery` at `~3.18.5`. The plugin subclasses
29
+ peer classes and mirrors peer protocol, so raising one pin alone is how the
30
+ seams drift: bump them in the commit that matches a harness release, re-run the
31
+ comparison `docs/PEER-PARITY.md` describes, and let the gate chain and the
32
+ probes confirm the result.
25
33
 
26
34
  pnpm runs the scripts in this repository, pinned through the `packageManager`
27
35
  field (corepack resolves it). Treat `pnpm run <script>` as the reference
@@ -40,24 +48,24 @@ Exactly three dependencies run install scripts (`@deepseek-ai/dsh-subprocess-loc
40
48
 
41
49
  ## The type gate
42
50
 
43
- `lint:types` runs `tsc --noEmit` under `strict` with `checkJs`: the JavaScript
44
- in `lib/` is fully type-checked, and the types come from the peer packages'
45
- own `.d.ts` — nothing is invented twice. `types/dsh-services.d.ts` pulls in the
46
- `declare module '@deepseek-ai/cordis'` augmentations the service packages ship,
47
- so `ctx.fs`, `ctx.shell`, `ctx.sandboxPolicy` and friends are typed; it declares
48
- nothing itself and ships nowhere.
49
-
50
- The gate's scope is what is NOT generated: `test/**` (the probe scripts
51
- included) and the ambient service types the build shares. The sources under
52
- `src/` are checked by the build program itself (`strict`, `noEmitOnError`),
53
- which is strictly stronger than checking their emitted JavaScript. The shipped
54
- artifact was checked first because it is where protocol drift with the pinned
55
- peers hurts; the tests and probes came next, and the errors they surfaced were
56
- real — a test that pinned a raw number into an argv, a couple of hand-written
57
- JSDoc contracts wider than the code (`wslErrorCode`, `gitSpawnRewrite`) and
58
- narrower than their seams, and one accepted-but-never-enforced byte-window cap
59
- in the distro-routed file service, found because a fake hands the confinement
60
- seam an `"full"` enforcement the core itself never reports.
51
+ `lint:types` runs `tsc --noEmit` under `strict` with `checkJs` over what is NOT
52
+ generated — `test/**` and the ambient types — and the types come from the peer
53
+ packages' own `.d.ts`: nothing is invented twice. `types/dsh-services.d.ts`
54
+ pulls in the `declare module '@deepseek-ai/cordis'` augmentations the service
55
+ packages ship, so `ctx.fs`, `ctx.shell`, `ctx.sandboxPolicy` and friends are
56
+ typed; it declares nothing itself and ships nowhere.
57
+
58
+ The probe scripts under `test/probe/` are inside the gate as well. The sources
59
+ under `src/` are checked by the build program itself (`strict`,
60
+ `noEmitOnError`), which is strictly stronger than checking their emitted
61
+ JavaScript. The shipped artifact was checked first because it is where protocol
62
+ drift with the pinned peers hurts; the tests and probes came next, and the
63
+ errors they surfaced were real — a test that pinned a raw number into an argv,
64
+ a couple of hand-written JSDoc contracts wider than the code (`wslErrorCode`,
65
+ `gitSpawnRewrite`) and narrower than their seams, and one
66
+ accepted-but-never-enforced byte-window cap in the distro-routed file service,
67
+ found because a fake hands the confinement seam an `"full"` enforcement the core
68
+ itself never reports.
61
69
 
62
70
  Three patterns keep the check honest:
63
71
 
@@ -69,9 +77,12 @@ Three patterns keep the check honest:
69
77
  at the injection point — the cast states which contract the double stands in
70
78
  for, instead of silencing the mismatch with `any`.
71
79
  - `@ts-expect-error` is allowed exactly when the suppressed error names a real
72
- upstream seam gap, carries the reason inline, and the gap is written up in
73
- `docs/UPSTREAM-*.md`. It is a published finding with a tracking doc, not a
74
- silencer.
80
+ upstream seam gap, carries the reason inline, and the gap is written up — in a
81
+ `docs/UPSTREAM-*.md` tracking doc when the fix is an upstream ask (the
82
+ `LocalBashExecutor.spawnSpec` seam has one), or in the source comment itself
83
+ when the honest fix needs a protocol decision this package owns (the exit-4
84
+ relay's unknown entry kind is the other site, deliberately left visible). It
85
+ is a published finding, not a silencer.
75
86
 
76
87
  ## The build
77
88
 
@@ -119,15 +130,25 @@ Rules that keep this honest:
119
130
  2. `pnpm run build`, then `pnpm test`. There is no exclude list to maintain and
120
131
  no registration step: the build and `lint:build` derive everything from
121
132
  `src/`, and `lib/` is generated output — never edit it by hand.
133
+ 3. If the module gets a row of its own on the Plugins page, add its display
134
+ metadata as `locale/<name>/en.json` and `locale/<name>/zh.json` — a
135
+ `meta.title` / `meta.description` pair each — and expose both through
136
+ `exports` (`./<name>/locale/en.json`, `./<name>/locale/zh.json`), which is
137
+ what that page reads before falling back to the package `description`. The
138
+ bundle's own pair is `locale/en.json` / `locale/zh.json`, and `subprocess`,
139
+ `picker`, `shell-env` and `auto-preset` carry theirs; `files` already ships
140
+ `locale/**/*.json`, and `lint:style` fails an `exports` target that does not
141
+ exist.
122
142
 
123
143
  ## The development loop
124
144
 
125
145
  1. Clone the repository inside the distro. Most contributors develop there.
126
146
  2. Make the change.
127
147
  3. Run the gates below.
128
- 4. Restart the DSH app before testing by hand, if you changed anything under
129
- `lib/`. A running process caches ES modules, so it keeps the old code
130
- otherwise.
148
+ 4. If you changed anything under `lib/`, run `pnpm run sync:windows` first and
149
+ then restart the DSH app. The app loads the Windows mirror (see below), so a
150
+ restart without the sync just reloads the old code; a running process also
151
+ caches ES modules.
131
152
 
132
153
  ### The runtime mirror
133
154
 
@@ -151,26 +172,28 @@ terminal if you would rather not see the prompt.
151
172
 
152
173
  | Command | Covers | Needs a harness |
153
174
  |---|---|---|
154
- | `pnpm test` | style checks, syntax pass, type check, and unit tests | no |
175
+ | `pnpm test` | style checks, syntax pass, build consistency, type check, and unit tests | no |
155
176
  | `pnpm run test:coverage` | unit tests with coverage | no |
156
177
  | `pnpm run lint:style` | style and lint rules | no |
157
- | `pnpm run lint:types` | the TypeScript check over `lib/` (`tsc --noEmit`) | no |
178
+ | `pnpm run lint:build` | the build-freshness check: every source has its artifacts and the committed bytes match a real build | no |
179
+ | `pnpm run lint:types` | the TypeScript check over `test/**` and the ambient types (`tsc --noEmit`) | no |
158
180
  | `pnpm run probe` | filesystem behaviour against a real distro | yes |
159
181
  | `pnpm run probe:sandbox` | what bubblewrap confines and what it does not | no |
160
182
  | `pnpm run probe:sandbox-shell` | the confined executor through a real boot | yes |
161
183
  | `pnpm run probe:terminal` | the terminal provider | yes |
162
- | `pnpm run probe:substrate` | the agent filesystem substrate over a real `wsl.exe` transport | yes; inside the distro |
163
- | `pnpm run probe:watch` | the in-distro watcher over a real directory | yes; inside the distro |
164
- | `pnpm run probe:agent` | the resident-vs-one-shot fallback parity legs | yes; inside the distro |
165
- | `pnpm run probe:exec` | the agent-backed execution handle: timeout, kill, cwd failure | yes; inside the distro |
184
+ | `pnpm run probe:substrate` | the agent filesystem substrate over a real `wsl.exe` transport | no; inside the distro |
185
+ | `pnpm run probe:watch` | the in-distro watcher over a real directory | no; inside the distro |
186
+ | `pnpm run probe:agent` | the resident-vs-one-shot fallback parity legs | no; inside the distro |
187
+ | `pnpm run probe:exec` | the agent-backed execution handle: timeout, kill, cwd failure | no; inside the distro |
166
188
  | `pnpm run probe:missing-wsl` | how a `wslPath` that cannot start is reported | yes |
167
189
  | `pnpm run probe:picker` | the picker's root level, its refusals and its cap | yes |
168
190
  | `pnpm run probe:mode` | which POSIX-mode facts survive the share | no; Windows Node |
169
191
  | `pnpm run probe:sandbox-off` | the documented opt-out on both providers | yes |
170
192
 
171
193
  Every script above is wired in `package.json`. `pnpm test` runs `lint:style`,
172
- `test:syntax`, `lint:types` and `test:unit` in that order. `prepublishOnly` runs `pnpm test` again
173
- at publish time, so a broken gate stops a release before the upload.
194
+ `test:syntax`, `lint:build`, `lint:types` and `test:unit` in that order.
195
+ `prepublishOnly` runs `pnpm test` again at publish time, so a broken gate stops a
196
+ release before the upload.
174
197
 
175
198
  CI runs `pnpm test` on Node 22 and 24, on both `ubuntu-latest` and
176
199
  `windows-latest`, and `pnpm run test:coverage` on Node 24. The Windows runners are
@@ -181,6 +204,19 @@ The probes marked "yes" need Windows, WSL2, a mounted harness profile and a link
181
204
  checkout, which a hosted runner does not have. Run them on your own machine and
182
205
  paste the output into the pull request.
183
206
 
207
+ The rows marked "no; inside the distro" (`substrate`, `watch`, `agent`, `exec`)
208
+ still need Windows and WSL2, but no harness boot and no profile: run them from a
209
+ distro terminal with Node available, and they reach `wsl.exe` through interop.
210
+
211
+ Two conveniences for the real-machine set. `test/probe/run-all-when-closed.sh
212
+ --include-fs` runs all twelve probes once the app has exited. And `ci.yml` has a
213
+ `probes` job you dispatch by hand: `sandbox` runs whenever WSL is present, then
214
+ `mode`, `terminal`, `picker` and `missing-wsl` where the harness app is
215
+ installed. It is named for a self-hosted Windows runner although its `runs-on`
216
+ carries the hosted `windows-latest` label, and it is not a green no-op on a
217
+ hosted runner — the `sandbox` leg fails, not skips, where the default distro has
218
+ no bubblewrap.
219
+
184
220
  ## Adding a test
185
221
 
186
222
  Choose the layer by what the test needs.
@@ -190,44 +226,70 @@ Choose the layer by what the test needs.
190
226
  | Unit test | `test/*.test.mjs` | pure logic: path translation, listing, shell argument building, terminal helpers |
191
227
  | Probe | `test/probe/` | behaviour that needs a real distro, a real profile, or bubblewrap |
192
228
 
193
- Unit tests run under `node --test`. A test that imports only this repository goes in
194
- both the `test:unit` and the `test:coverage` lists in `package.json`, which name
195
- their files explicitly. A test that needs a DSH peer — `test/sandbox.test.mjs`
196
- imports `lib/sandbox.js`, which imports `@deepseek-ai/dsh-sandbox` — goes in
197
- `test:unit` only and must skip itself when the peer is missing, because CI installs
198
- nothing: try the import, print a `SKIP` line, exit 0.
199
-
200
- Probes are shell scripts that mount a throwaway profile, plus the small plugin
201
- files they load. Those plugin files are the `*-probe.mjs` and `*-probe.yml` files
202
- in the same directory. The setup steps for the throwaway profile are in the header
203
- of `test/probe/run.sh`. `terminal.sh` and `sandbox-shell.sh` reuse it.
229
+ Unit tests run under `node --test`. The `test:unit` and the `test:coverage` lists
230
+ in `package.json` name their files explicitly; a test that only touches this
231
+ repository belongs in both. Two lags in those lists today are hygiene to fix in
232
+ passing, not a pattern to copy: each names one file twice
233
+ (`test/fsio-text.test.mjs` in `test:unit`, `test/fs.boot.test.mjs` in
234
+ `test:coverage`), and `test/fs-routing.test.mjs` is in neither
235
+ (`docs/REQUIREMENTS.zh.md` tracks that as OI-7).
236
+
237
+ A test that imports a pinned `@deepseek-ai/*` package goes in the same lists and
238
+ needs no guarding — CI runs `pnpm install --frozen-lockfile` over the locked tree
239
+ before `pnpm test`. Five files import such packages at runtime today —
240
+ `test/provider.test.mjs` is what covers the assembled service rows, and
241
+ `test/composition.test.mjs` runs the loader's real patch algorithm — and a sixth
242
+ (`test/terminal-activity.test.mjs`) reaches a peer only through a JSDoc type
243
+ import, which `checkJs` still resolves.
244
+
245
+ Where a peer-free seam exists, prefer it. `test/sandbox.test.mjs` runs against
246
+ `lib/sandbox-core.js`, with a stand-in for the peer's `SandboxUnavailableError`
247
+ contract, which leaves only the one-line peer binding in `lib/sandbox.js`. That
248
+ line is covered by the harness probes, which boot the plugin and so reach
249
+ `WslSandbox` through `lib/index.js` — not by `probe:sandbox`, which reads only
250
+ the argument builder out of `lib/bwrap.js`. A test that cannot run everywhere
251
+ must skip itself — print a `SKIP` line and exit 0 — the way that file does on
252
+ `win32`, where its fake `wsl.exe` (a POSIX script) cannot execute.
253
+
254
+ Probes are shell scripts, plus the small plugin files they load: the `*-probe.mjs`
255
+ and `*-probe.yml` files in the same directory. Two shapes. The harness probes
256
+ (`run.sh`, `terminal.sh`, `sandbox-shell.sh`, `picker.sh`, `missing-wsl.sh`,
257
+ `sandbox-off.sh`) boot a throwaway profile whose setup steps are in the header of
258
+ `test/probe/run.sh`; the five others reuse the profile `run.sh` creates rather
259
+ than making their own. The distro probes (`substrate.sh`, `watch.sh`, `agent.sh`,
260
+ `exec.sh`) boot nothing: they run from a distro terminal and drive the resident
261
+ agent through the `wsl.exe` interop transport. `sandbox.sh` and `mode.sh` sit
262
+ outside both groups — no harness boot, no profile.
204
263
 
205
264
  Keep probe output idempotent. Several probes delete their scratch files at the
206
265
  start, because a leftover file from an earlier run changes the result.
207
266
 
208
267
  ## What has been verified
209
268
 
210
- `pnpm test` covers the pure modules and runs a `--check` pass over every shipped
211
- module. It cannot import the service modules, because those need DSH peer packages
212
- that a bare checkout does not have. Only booting the harness closes that gap, which
213
- is why the probes exist; the archived record's §15.4 documents the five rounds of
214
- misdiagnosis the gap once caused.
269
+ `pnpm test` runs the whole unit suite once the one `pnpm install` above has
270
+ fetched the pinned packages: the files that import them statically exercise the
271
+ services through the loader and their declared types, and `test:syntax` keeps a
272
+ `node --check` pass over every shipped module. What no test on a Linux leg can
273
+ reach is the Windows-plus-WSL2 topology itself: a real `wsl.exe`, a real distro,
274
+ a real harness boot. That is the gap the probes close; the archived record's
275
+ §15.4 documents the five rounds of misdiagnosis it once caused.
215
276
 
216
277
  `test/probe/sandbox.sh` needs no harness: it uses the same arguments
217
278
  `lib/sandbox.js` builds and asserts what bubblewrap does and does not confine,
218
279
  recording the interop escape as `INFO` because a Linux sandbox cannot govern a
219
- Windows process. The other probes boot a throwaway profile bound to the distro:
280
+ Windows process. The harness probes boot a throwaway profile bound to the distro:
220
281
 
221
282
  - the filesystem probe asserts the write path and the write checks — a write outside
222
283
  the policy root and a write in `read-only` mode both return `FS_SANDBOX_DENIED`,
223
- while `danger-full-access` is not checked;
284
+ while `danger-full-access` is the approved escalation and its write is asserted
285
+ to land;
224
286
  - the shell probe drives all three modes through `ctx.shell` and checks the refusal
225
287
  classification the tool layer returns;
226
288
  - the terminal probe asserts the distro, the starting directory, and the `DSH_*`
227
289
  variables forwarded through `WSLENV`.
228
290
 
229
- The setup steps for the throwaway profile are in the header of `test/probe/run.sh`;
230
- `terminal.sh` and `sandbox-shell.sh` reuse it.
291
+ The setup steps for that throwaway profile are in the header of
292
+ `test/probe/run.sh`; the harness probes share the one profile it creates.
231
293
 
232
294
  On Windows 11 + WSL2 (Ubuntu 26.04) the following has been verified:
233
295
 
@@ -242,11 +304,12 @@ On Windows 11 + WSL2 (Ubuntu 26.04) the following has been verified:
242
304
  outside the session workspace is refused inside the distro and the following
243
305
  wider-permission request succeeds.
244
306
 
245
- Counts: 57 unit assertions on a bare checkout, plus 5 in `test/sandbox.test.mjs` which
246
- need the `dsh-sandbox` peer and skip themselves without it; 19 filesystem-probe
247
- assertions; 15 shell-probe checks; 9 picker checks; 6 missing-executable checks; 5 mode
248
- checks; 5 opt-out checks and 10 sandbox expectations plus the recorded escape; and 3
249
- terminal assertions.
307
+ Counts: 28 unit test files under `test/` — 27 of them registered in the
308
+ `test:unit` list, see the OI-7 note above — which need the pinned packages
309
+ installed, and over 1000 `assert` call sites across them; 19 filesystem-probe
310
+ assertions; 15 shell-probe checks; 9 picker checks; 6 missing-executable checks;
311
+ 5 mode checks; 5 opt-out checks and 10 sandbox expectations plus the recorded
312
+ escape; and 3 terminal assertions.
250
313
 
251
314
  ## Commit and pull request conventions
252
315