golem15/msd-core
3
0
Fork 0
Code Issues Pull Requests 1 Actions 1 Packages Projects Releases Wiki Activity
3,203 Commits 4 Branches 1 Tag
42f8c192a8eafd96a00216e263bed83dc034f708
Commit Graph

4 Commits

Author SHA1 Message Date
Tom Boucher
6f2520786d feat(#498): single Package Identity seam for /gsd:update + fix runtime undefined-name bug (#499) 
* feat(#498): generated package-identity seam derived from package.json

Introduce a single source for GSD's published-package coordinates:
scripts/generate-package-identity.cjs (pure deriveIdentity + formatManualInstall
+ render) emits the generated get-shit-done/bin/lib/package-identity.cjs with
values baked from package.json at build time. Baking is required because the
installed tree carries only a synthetic {"type":"commonjs"} package.json, so a
runtime require('package.json').name resolves to undefined (#378). Reconciles

Wired into npm run build; a parity test fails CI if the committed file drifts
from package.json.

Refs #498

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(#498): repoint update worker + check-latest-version at the seam

- check-latest-version.cjs sources PACKAGE_NAME from the package-identity seam
  instead of a re-typed literal (single source; #2992's constant guarantee is
  preserved since the seam bakes from package.json).
- gsd-check-update-worker.js no longer does require('../package.json').name
  (resolved to undefined in the installed tree → background update check
  silently broken, #378). It now delegates the latest-version lookup to
  checkLatestVersion(), collapsing the duplicated npm-view call onto the single
  deterministic adapter and inheriting its typed {ok,version,reason} surface.
- Move the PR #3102 Windows shell-gate contract test onto execNpm (where the
  spawn now lives) and assert the worker no longer spawns npm directly.
- Rewrite the #378 contract: worker must NOT use require(package.json).name and
  must delegate; check-latest-version PACKAGE_NAME is single-sourced from the seam.

Fixes #378-class runtime breakage. Refs #498

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* chore(#498): changeset for package-identity seam + update-check fix

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* feat(#498): drift-guard lint — value-check GSD coordinate literals against the seam

scripts/lint-package-identity-drift.cjs scans the runtime/code surface
(bin/, hooks/, scripts/, get-shit-done/) and asserts every GSD package name
and GitHub repo slug literal equals the Package Identity seam's current value.
Passes today; fails the moment a repoint isn't propagated (rename package.json,
regenerate the seam, and stale literals are reported until updated). This is
the second adapter that makes the seam real and a repoint mechanically safe.

Enforced via tests/issue-498-identity-drift-lint.test.cjs (scanRepo === [])
under npm test; also exposed as `npm run check:identity-drift`.

Refs #498

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* feat(#498): update-context projection — port update.md resolution to a tested seam

Add get-shit-done/bin/lib/update-context.cjs: a pure, injected-fs port of
update.md's ~280-line get_installed_version bash. resolveUpdateContext()
reproduces the full precedence cascade (preferred fast-path -> local probe ->
global probe via env overrides then $HOME -> LOCAL-if-distinct -> scope
cascade -> UNKNOWN) and returns the 4-field contract { installedVersion,
scope, runtime, gsdDir }. The fs is injected so every branch is finally
testable without a live multi-runtime install.

Expose it as `gsd-tools update-context [--config-dir <d>] [--runtime <r>] --json`.
Purely additive — update.md is unchanged in this commit; the workflow swap
follows separately.

Refs #498

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* feat(#498): swap update.md resolution to the update-context projection

Replace ~280 lines of inline runtime/scope/config-dir bash in update.md's
get_installed_version step with a call to `gsd-tools update-context --json`
(60 lines: derive PREFERRED_* from execution_context, resolve gsd-tools.cjs,
parse the 4-field JSON). Behavior is unchanged — the projection reproduces the
same cascade — but the logic is now tested in update-context.cjs instead of
untestable bash-in-markdown.

Relocate the #3608 antigravity-first-class contract onto the projection
(RUNTIME_DIRS order, inferPreferredRuntime, envRuntimeDirs) plus a behavioral
test; keep the execution_context path-classification assertion on update.md.
Re-point install.test's custom-config-dir assertion (kilo.jsonc/KILO_CONFIG)
to update-context.cjs where that detection now lives.

Full root suite: 2022 pass / 0 fail.

Refs #498

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* docs(#498): record Update Context Module in CONTEXT.md

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(#498): CI — avoid bare gsd-tools in update.md; register new CLI modules

- update.md update-context invocation: resolve the PATH gsd-tools shim into a
  variable and call "$GSD_TOOLS" (never a bare `gsd-tools` command) — satisfies
  the #2851 workflow-bare-gsd-tools guard.
- Register package-identity.cjs and update-context.cjs in docs/INVENTORY.md
  (CLI Modules 76 -> 78 + rows) and regenerate docs/INVENTORY-MANIFEST.json,
  fixing inventory-counts and inventory-manifest-sync.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* test(#498): make update-context + parity tests OS-agnostic (Windows CI)

Two Windows-only test failures, both test-portability (production code is fine —
the real-fs CLI integration test passed on Windows):

- update-context resolver tests + bug-3608 behavioral test used POSIX path-string
  keys in their fake fs, but the resolver builds lookups via path.join/resolve
  (backslash + drive letter on Windows) → keys never matched → everything
  resolved to UNKNOWN/claude. Normalize fake-fs keys and gsdDir comparisons
  through path.resolve so they match on both platforms.
- package-identity parity test compared render() (LF) to the committed file,
  which Windows git checks out as CRLF (no .gitattributes eol rule). Normalize
  line endings before comparing, matching the repo convention
  (autonomous-decomposition, bug-3707).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(#498): update.md backup must use GSD_DIR (adversarial-review finding)

The get_installed_version rewrite emits GSD_DIR but dropped the probe-loop
variables LOCAL_DIR/GLOBAL_DIR. The backup_custom_files step still read those,
so RUNTIME_DIR went empty for every LOCAL/GLOBAL install and detect-custom-files
was skipped — and since the update then runs a clean install that wipes managed
dirs (commands/gsd, get-shit-done), user-added files could be deleted without
the intended backup.

Set RUNTIME_DIR="$GSD_DIR" directly (the resolved config dir; empty for
UNKNOWN scope, which still skips the backup). Add a structural regression
(tests/issue-498-update-backup-runtime-dir.test.cjs).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* test(#503): re-point Antigravity .agent detection at the #498 projection

#499 moves the runtime/scope detection cascade out of update.md inline bash
into get-shit-done/bin/lib/update-context.cjs. The #503 regression test asserted
on the inline RUNTIME_DIRS array, which no longer exists, so it would fail
against the projected update.md even though the .agent guarantee is preserved.

Rewrite it to verify the surviving surfaces:
 - behavioral: resolveUpdateContext resolves a LOCAL ./.agent install to the
   antigravity runtime (the original root cause, now covered by adding
   ['antigravity', '.agent'] to the projection RUNTIME_DIRS table)
 - update.md prose classifier still maps /.agent/ -> antigravity
 - the post-update cache-clear for-dir loop still includes .agent

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(#498): finish de-hardcoding consumers + close adversarial-review parity gaps

Restore the consumer de-hardcoding that is the point of the seam, and close the
parity gaps an adversarial review (codex) found in the update-context projection.

De-hardcode the repo slug + install command in the changeset tooling — #516
only single-sourced the package NAME, leaving 'open-gsd/get-shit-done-redux'
hardcoded in scripts/changeset/cli.cjs and github-release-notes.cjs. Route both
through the seam's repoSlug/packageName so a rename is a regenerate, not a hand
edit. The drift-lint real scan now reports zero divergent coordinate literals.

Projection parity vs the old inline bash, as ONE consistent rule
(trustedVersionAt) applied on every path:
 - expand a leading ~/ in preferredConfigDir before the fast path (the bash ran
   expand_home first; a custom --config-dir ~/foo otherwise fell to UNKNOWN)
 - trust a version only when BOTH VERSION and the update.md marker exist — fast
   path AND LOCAL/GLOBAL cascade; a partial dir falls to 0.0.0 keeping scope
 - apply the same same-path dedup to the 0.0.0 fallback so a partial install
   probed from cwd===home is not misdetected as LOCAL

Adds regression tests for tilde expansion, VERSION-only (cascade + fast path),
and the cwd===home partial-install dedup.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
 2026-05-30 16:53:44 -04:00
Tom Boucher
5b3646d6c8 fix(#213): support antigravity 2.x config directory split (#217) 
* fix(#213): support antigravity 2.x config directory split

* fix(#213): harden antigravity tests for windows parity
 2026-05-24 14:17:03 -04:00
Tom Boucher
ca3be82f71 fix(3597): clear residual Windows test failures + add ratchet lint guard 
Two more diagnostic passes (clusters J: residuals in already-touched files,
K: 12 untouched files) plus a production-code path fix and a new
ratchet-style lint guard.

## Test-only fixes (14 files)

bug-1736, bug-2248, bug-2698 — replace inline 1s-budget rmSync with the
shared cleanup() helper (5s budget, 20×250ms retries). The earlier
inline maxRetries:10 / retryDelay:100 wasn't enough to absorb Windows
Defender's deferred-scan handle hold on cold runners.

bug-2256, skill-manifest — also override USERPROFILE alongside HOME in
beforeEach/runGsdTools calls. os.homedir() reads USERPROFILE on win32,
so HOME-only stubs leak the runner's real home into the SUT.

bug-2784, bug-3608, enh-2500, enh-2790, few-shot-calibration,
gsd-settings-advanced — CRLF tolerance: literal \n in regexes against
file content (frontmatter anchors, bash-fence regex, multi-line
numbered-list captures, awk-block extractors) becomes \r?\n; split('\n')
becomes split(/\r?\n/). Windows checkout with autocrlf=true puts \r
before every \n; .+ doesn't match \r in JS regex by default.

bug-2966 — three-part fix to extractStepRun (CRLF split), awk regex
(\r?\n), and conflict-marker parser (rawLine + \r$ strip).

bug-2969, config — normalize separators on test assertions where the
SUT correctly emits \ on win32 but the test compares against /.

prompt-injection-scan — normalize relPath via replace(/\\/g, '/') before
ALLOWLIST.has() lookup. ALLOWLIST keys are POSIX; path.relative returns
backslashes on win32 → falsely scans allowlisted security module → trips
the boundary-tag detector on its own legitimate detection code.

prune-orphaned-worktrees — use the existing canonicalPath +
listedWorktreePaths(repoDir).has(...) helpers instead of substring
matching the raw path. git stores long-form canonical paths
(runneradmin), but mkdtempSync returns 8.3 short-form (RUNNER~1) on
Windows runners; plain string compare misses every entry.

## Production-code fix (1 file)

get-shit-done/bin/lib/init.cjs — bug-3491 in_nested_subdir computation
canonicalizes both worktreeRoot and cwd via fs.realpathSync.native +
path.relative before declaring "nested." Windows runner cwd (8.3 short
name) vs git's --show-toplevel (long form, forward slashes) made the
raw string compare always say true even at the worktree root, breaking
the "init new-project at worktree root" subtest.

## New ratchet lint guard

tests/windows-test-parity-guard.test.cjs — scans tests/ for 7
anti-patterns that drove the Windows failure clusters. Each rule has a
baseline count snapshot from this PR; the test fails when a new
occurrence appears (count grows above baseline), ratcheting down as
existing offenders are fixed. Patterns covered:

  G1 split('\n') after readFileSync (use /\r?\n/)
  G2 ```bash\n fence regex (use ```bash\r?\n)
  G3 ^---\n frontmatter anchor (use ^---\r?\n)
  G4 hardcoded "/tmp/..." literal passed to fs.* (use os.tmpdir())
  G5 bare 'npm' to execFileSync without {shell:true} on win32
  G6 process.env.HOME stub with no USERPROFILE
  G7 fs.rmSync({recursive,force}) without maxRetries

Future Windows-parity regressions get caught at PR time rather than
five iterations into a CI loop.

Validated: holodeck (ubuntu docker) 11232/0 pass (count +8 = the 7
new ratchet tests + parent describe).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-16 12:39:02 -04:00
Tom Boucher
f27a20a388 fix(3608): model Antigravity as a first-class runtime in update.md 
bin/install.js and the SDK already treat Antigravity as a distinct runtime
with config dir ~/.gemini/antigravity, env var ANTIGRAVITY_CONFIG_DIR, and
CLI flag --antigravity. get-shit-done/workflows/update.md did not — so
/gsd-update invoked from an Antigravity install classified the runtime as
base Gemini, because:

- RUNTIME_DIRS listed "gemini:.gemini" with no antigravity entry, so the
  scan matched ~/.gemini before ever looking for ~/.gemini/antigravity.
- The PREFERRED_RUNTIME env-var ladder checked GEMINI_CONFIG_DIR but not
  ANTIGRAVITY_CONFIG_DIR.
- The local-scope scan loops at lines 101 and 590 listed .gemini with no
  .gemini/antigravity sibling.
- The ENV_RUNTIME_DIRS append block ignored ANTIGRAVITY_CONFIG_DIR.
- The path-to-runtime classification bullets only mapped /.gemini/ ->
  gemini, with no /.gemini/antigravity/ -> antigravity branch.

Every list is now updated so the more-specific antigravity entry precedes
the base gemini entry, matching the installer at bin/install.js (lines
396-404, 1745-1749, 6175, 6475).

tests/bug-3608-antigravity-update-runtime-classification.test.cjs is the
structural regression guard. It parses the RUNTIME_DIRS bash array out of
update.md and asserts the antigravity entry is present and ordered before
gemini, asserts the env-var ladder checks ANTIGRAVITY_CONFIG_DIR before
GEMINI_CONFIG_DIR, asserts every `for dir in ...` scan loop that mentions
.gemini also lists .gemini/antigravity ordered before it, and asserts the
path-classification prose bullet lists antigravity before gemini.

Per CONTRIBUTING.md: the test uses readFileSync on a .md file annotated
`// allow-test-rule: source-text-is-the-product` because the bash blocks
inside update.md ARE the deployed program — the agent loads update.md and
runs them as-written.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 21:58:51 -04:00