Merge pull request #175 from NDDev-OpenNetwork/fix/discovery-hidden-dirs

fix: skip hidden directories in discovery; read-only remote observation for quarantine

Author: rldyourmnd

.gds/bundle.lock.yaml

@@ -5,14 +5,14 @@ bundle:
   version: "0.9.6-dev"
   release_sequence: 0
   channel: "development"
-  source_tree_digest: "sha256:6799d517ad06faba9cc1edc3830dd54d0a890f3573bc1f60a26232f5d1926524"
-  digest: "sha256:db7dae4282ef3d69f0375c9e54feada9dc11f219d64e591a76559883ea598f0e"
+  source_tree_digest: "sha256:b9d572eb73bb14ae8ede20dfc1e9317046dcb544270eff95153746f93f094820"
+  digest: "sha256:a45453995b649d082b39753c1d7eff869d2dcad5180532cb14f85435fae2a7fa"
 
 projection:
-  input_digest: "sha256:37d63f73feba8549ec22aa97b81b7cc0d53b701d73b36661625c0bb61da53998"
-  output_digest: "sha256:508542eaec283ca3e14bb78c495c0c52a45ebb85b24cce2e104f8655ad1b0bb1"
+  input_digest: "sha256:90af33f4abbb808f5567e074347f17822cc610029e11a835fb5a8918c376ec9b"
+  output_digest: "sha256:080add7de51702df727bb8f1b770dcd5fd8343000d8b249af94a7b342a0a240b"
   files:
     - path: ".gds/compiled-policy.json"
       digest: "sha256:211668dd85e9d4bd3f547a1ea0cc9b1a5f486c64bde02a328b53dc3ef665c9fd"
     - path: ".github/workflows/gds-ci.yml"
-      digest: "sha256:2e18aafadceafafab54dbf21fd0f68b50c9618aed05d8f2eb39e161ef999c287"
+      digest: "sha256:eef6659ecb3cd40914b73fde9856ad2bfa96f106ceac52f8c2fda0ef4fc00a79"

.github/workflows/gds-ci.yml

@@ -1,8 +1,8 @@
 # GENERATED FILE - DO NOT EDIT DIRECTLY
 # generator: gds
 # bundle: 0.9.6-dev
-# source-tree-digest: sha256:6799d517ad06faba9cc1edc3830dd54d0a890f3573bc1f60a26232f5d1926524
-# input-digest: sha256:37d63f73feba8549ec22aa97b81b7cc0d53b701d73b36661625c0bb61da53998
+# source-tree-digest: sha256:b9d572eb73bb14ae8ede20dfc1e9317046dcb544270eff95153746f93f094820
+# input-digest: sha256:90af33f4abbb808f5567e074347f17822cc610029e11a835fb5a8918c376ec9b
 # output-digest: sha256:b9bf3d0c64c0fb371596e7d090e82e62aebbfde91929115fc15fb28644e4fd38
 # edit-source:
 #   - .gds/repository.yaml

core/discovery/local.go

@@ -306,9 +306,15 @@ func pathDepth(relative string) int {
 }
 
 func excludedDirectory(name string) bool {
+	// Hidden directories other than a repository's own .git boundary are tool
+	// state, caches, or generated fixtures (".tmp", ".idea", ".venv", ...) —
+	// never portfolio checkouts. Scanning them manufactures anchor findings
+	// for synthetic trees and burns time on ignored content.
+	if name != ".git" && strings.HasPrefix(name, ".") {
+		return true
+	}
 	switch name {
-	case ".cache", ".idea", ".pytest_cache", ".ruff_cache", ".tox", ".venv",
-		"__pycache__", "node_modules", "target", "vendor":
+	case "__pycache__", "node_modules", "target", "vendor":
 		return true
 	default:
 		return false

core/discovery/local_test.go

@@ -190,3 +190,33 @@ func TestDiscoverAcceptsIdentityPinnedBySuperprojectAndCheckedOutStandalone(t *t
 		}
 	}
 }
+
+func TestDiscoverSkipsHiddenDirectoryTrees(t *testing.T) {
+	t.Parallel()
+	// Boundaries are reported canonicalized: on macOS t.TempDir() lives under
+	// /var, which is a symlink to /private/var.
+	root, err := filepath.EvalSymlinks(t.TempDir())
+	if err != nil {
+		t.Fatal(err)
+	}
+	plain := filepath.Join(root, "plain")
+	fixture := filepath.Join(root, ".tmp", "native-linux", "fixture-repo")
+	for _, path := range []string{plain, fixture} {
+		if err := os.MkdirAll(path, 0o755); err != nil {
+			t.Fatal(err)
+		}
+		if output, err := exec.Command("git", "init", "-q", path).CombinedOutput(); err != nil {
+			t.Fatalf("git init %s: %v\n%s", path, err, output)
+		}
+	}
+
+	result, err := newTestDiscovery(t).Discover(
+		context.Background(), root, Options{MaxDepth: 8, MaxRepositories: 10, Concurrency: 2},
+	)
+	if err != nil {
+		t.Fatalf("Discover() error = %v", err)
+	}
+	if len(result.Boundaries) != 1 || result.Boundaries[0].Path != plain {
+		t.Fatalf("boundaries = %#v, want only %s", result.Boundaries, plain)
+	}
+}

core/providers/git/handoff.go

@@ -362,7 +362,10 @@ func (runner *MutationRunner) observeRemoteRef(
 ) (string, error) {
 	result, err := runner.runWithEnvironment(
 		ctx, root, map[int]struct{}{0: {}}, nil,
-		"-c", "protocol.allow=never", "-c", "protocol.file.allow=always",
+		"-c", "protocol.allow=never",
+		"-c", "protocol.file.allow=always",
+		"-c", "protocol.https.allow=always",
+		"-c", "protocol.ssh.allow=always",
 		"ls-remote", "--refs", remoteURL, remoteRef,
 	)
 	if err != nil {

core/providers/git/quarantine.go

@@ -168,7 +168,11 @@ func (runner *MutationRunner) observeCheckoutForQuarantine(
 		status.Worktrees[0].Locked || status.Worktrees[0].Prunable {
 		return CheckoutQuarantineEvidence{}, errors.New("checkout is not clean, current, and exclusive")
 	}
-	remoteURL, err := runner.validatedPushURL(ctx, physical, "origin")
+	// Quarantine only reads the remote (ls-remote below) to prove the checkout's
+	// commits are published; it never pushes. The fetch-side URL is therefore
+	// the right one to observe and network transports are allowed — the
+	// mutation-only gate (validatedPushURL) does not apply to a read.
+	_, remoteURL, err := runner.validatedRemoteURL(ctx, physical, "origin")
 	if err != nil {
 		return CheckoutQuarantineEvidence{}, err
 	}

core/providers/git/quarantine_test.go

@@ -3,6 +3,7 @@ package git
 import (
 	"context"
 	"crypto/sha256"
+	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -75,3 +76,45 @@ func TestQuarantineCheckoutPreservesDirtyRepository(t *testing.T) {
 		t.Fatalf("dirty work changed: %q %v", content, err)
 	}
 }
+
+func TestQuarantineRemoteValidationAdmitsNetworkTransportsForReadOnlyObservation(t *testing.T) {
+	client, _ := mutationRepository(t)
+	runFetchGit(t, client, "remote", "add", "origin", "git@github.com:example/repository.git")
+	runner, err := NewMutationRunner()
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, url, err := runner.validatedRemoteURL(context.Background(), client, "origin")
+	if err != nil {
+		t.Fatalf("read-only observation must admit the fetch URL: %v", err)
+	}
+	if url != "git@github.com:example/repository.git" {
+		t.Fatalf("url=%q", url)
+	}
+	if _, err := runner.validatedPushURL(context.Background(), client, "origin"); !errors.Is(err, ErrNetworkMutationDisabled) {
+		t.Fatalf("push URL must remain mutation-gated: %v", err)
+	}
+}
+
+func TestQuarantineCheckoutNetworkRemoteFailsAtObservationNotAtTheMutationGate(t *testing.T) {
+	fixture := fastForwardFixture(t)
+	runFetchGit(t, fixture.client, "remote", "set-url", "origin", "ssh://127.0.0.1:1/repository.git")
+	head := stringsTrim(runFetchGit(t, fixture.client, "rev-parse", "HEAD"))
+	workspaceRoot := filepath.Dir(fixture.client)
+	stateRoot := t.TempDir()
+	quarantine := filepath.Join(stateRoot, "quarantine", "checkouts", "repo_fixture", head)
+	runner, err := NewMutationRunner()
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, err = runner.QuarantineCheckout(
+		context.Background(), workspaceRoot, fixture.client, stateRoot, quarantine,
+		head, "refs/heads/main", fmt.Sprintf("sha256:%x", sha256.Sum256([]byte("x"))),
+	)
+	if err == nil {
+		t.Fatal("an unreachable remote must fail the observation")
+	}
+	if errors.Is(err, ErrNetworkMutationDisabled) {
+		t.Fatalf("read-only remote observation must not hit the mutation gate: %v", err)
+	}
+}