refactor: migrate from go-git to go-getter (#2512)

taskfile/node_git.go

@@ -3,16 +3,14 @@ package taskfile
 import (
 	"context"
 	"fmt"
-	"io"
 	"net/url"
+	"os"
 	"path/filepath"
 	"strings"
+	"sync"
 
 	giturls "github.com/chainguard-dev/git-urls"
-	"github.com/go-git/go-billy/v5/memfs"
-	"github.com/go-git/go-git/v5"
-	"github.com/go-git/go-git/v5/plumbing"
-	"github.com/go-git/go-git/v5/storage/memory"
+	"github.com/hashicorp/go-getter"
 
 	"github.com/go-task/task/v3/errors"
 	"github.com/go-task/task/v3/internal/execext"
@@ -28,6 +26,36 @@ type GitNode struct {
 	path   string
 }
 
+type gitRepoCache struct {
+	mu    sync.Mutex             // Protects the locks map
+	locks map[string]*sync.Mutex // One mutex per repo cache key
+}
+
+func (c *gitRepoCache) getLockForRepo(cacheKey string) *sync.Mutex {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	if _, exists := c.locks[cacheKey]; !exists {
+		c.locks[cacheKey] = &sync.Mutex{}
+	}
+
+	return c.locks[cacheKey]
+}
+
+var globalGitRepoCache = &gitRepoCache{
+	locks: make(map[string]*sync.Mutex),
+}
+
+func CleanGitCache() error {
+	// Clear the in-memory locks map to prevent memory leak
+	globalGitRepoCache.mu.Lock()
+	globalGitRepoCache.locks = make(map[string]*sync.Mutex)
+	globalGitRepoCache.mu.Unlock()
+
+	cacheDir := filepath.Join(os.TempDir(), "task-git-repos")
+	return os.RemoveAll(cacheDir)
+}
+
 func NewGitNode(
 	entrypoint string,
 	dir string,
@@ -72,24 +100,78 @@ func (node *GitNode) Read() ([]byte, error) {
 	return node.ReadContext(context.Background())
 }
 
-func (node *GitNode) ReadContext(_ context.Context) ([]byte, error) {
-	fs := memfs.New()
-	storer := memory.NewStorage()
-	_, err := git.Clone(storer, fs, &git.CloneOptions{
-		URL:           node.url.String(),
-		ReferenceName: plumbing.ReferenceName(node.ref),
-		SingleBranch:  true,
-		Depth:         1,
-	})
+func (node *GitNode) buildURL() string {
+	// Get the base URL
+	baseURL := node.url.String()
+
+	ref := node.ref
+	if ref == "" {
+		ref = "HEAD"
+	}
+	// Always use git:: prefix for git URLs (following Terraform's pattern)
+	// This forces go-getter to use git protocol
+	return fmt.Sprintf("git::%s?ref=%s&depth=1", baseURL, ref)
+}
+
+// getOrCloneRepo returns the path to a cached git repository.
+// If the repository is not cached, it clones it first.
+// This function is thread-safe: multiple goroutines cloning the same repo+ref
+// will synchronize, and only one clone operation will occur.
+//
+// The cache directory is /tmp/task-git-repos/{cache_key}/
+func (node *GitNode) getOrCloneRepo(ctx context.Context) (string, error) {
+	cacheKey := node.repoCacheKey()
+
+	repoMutex := globalGitRepoCache.getLockForRepo(cacheKey)
+	repoMutex.Lock()
+	defer repoMutex.Unlock()
+
+	// Check if context was cancelled while waiting for lock
+	if err := ctx.Err(); err != nil {
+		return "", fmt.Errorf("context cancelled while waiting for repository lock: %w", err)
+	}
+
+	cacheDir := filepath.Join(os.TempDir(), "task-git-repos", cacheKey)
+
+	// check if repo is already cached (under the lock)
+	gitDir := filepath.Join(cacheDir, ".git")
+	if _, err := os.Stat(gitDir); err == nil {
+		return cacheDir, nil
+	}
+
+	getterURL := node.buildURL()
+
+	client := &getter.Client{
+		Ctx:  ctx,
+		Src:  getterURL,
+		Dst:  cacheDir,
+		Mode: getter.ClientModeDir,
+	}
+
+	if err := client.Get(); err != nil {
+		_ = os.RemoveAll(cacheDir)
+		return "", fmt.Errorf("failed to clone repository: %w", err)
+	}
+
+	return cacheDir, nil
+}
+
+func (node *GitNode) ReadContext(ctx context.Context) ([]byte, error) {
+	// Get or clone the repository into cache
+	repoDir, err := node.getOrCloneRepo(ctx)
 	if err != nil {
 		return nil, err
 	}
-	file, err := fs.Open(node.path)
-	if err != nil {
-		return nil, err
+
+	// Build path to Taskfile in the cached repo
+	taskfilePath := node.path
+	if taskfilePath == "" {
+		taskfilePath = "Taskfile.yml"
 	}
-	// Read the entire response body
-	b, err := io.ReadAll(file)
+	filePath := filepath.Join(repoDir, taskfilePath)
+
+	// Read file from cached repo
+	b, err := os.ReadFile(filePath)
 	if err != nil {
 		return nil, err
 	}
@@ -138,6 +220,22 @@ func (node *GitNode) CacheKey() string {
 	return fmt.Sprintf("git.%s.%s.%s", node.url.Host, prefix, checksum)
 }
 
+// repoCacheKey generates a unique cache key for the repository+ref combination.
+// Unlike CacheKey() which includes the file path, this identifies the repository itself.
+// Two GitNodes with the same repo+ref but different file paths will share the same cache.
+//
+// Returns a path like: github.com/user/repo.git/main
+func (node *GitNode) repoCacheKey() string {
+	repoPath := strings.Trim(node.url.Path, "/")
+
+	ref := node.ref
+	if ref == "" {
+		ref = "HEAD"
+	}
+
+	return filepath.Join(node.url.Host, repoPath, ref)
+}
+
 func splitURLOnDoubleSlash(u *url.URL) (string, string) {
 	x := strings.Split(u.Path, "//")
 	switch len(x) {

taskfile/node_git_test.go

@@ -102,3 +102,146 @@ func TestGitNode_CacheKey(t *testing.T) {
 		assert.Equal(t, tt.expectedKey, key)
 	}
 }
+
+func TestGitNode_buildURL(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name        string
+		entrypoint  string
+		expectedURL string
+	}{
+		{
+			name:        "HTTPS with ref",
+			entrypoint:  "https://github.com/foo/bar.git//Taskfile.yml?ref=main",
+			expectedURL: "git::https://github.com/foo/bar.git?ref=main&depth=1",
+		},
+		{
+			name:        "SSH with ref",
+			entrypoint:  "git@github.com:foo/bar.git//Taskfile.yml?ref=main",
+			expectedURL: "git::ssh://git@github.com/foo/bar.git?ref=main&depth=1",
+		},
+		{
+			name:        "HTTPS with tag ref",
+			entrypoint:  "https://github.com/foo/bar.git//Taskfile.yml?ref=v1.0.0",
+			expectedURL: "git::https://github.com/foo/bar.git?ref=v1.0.0&depth=1",
+		},
+		{
+			name:        "HTTPS without ref (uses remote HEAD)",
+			entrypoint:  "https://github.com/foo/bar.git//Taskfile.yml",
+			expectedURL: "git::https://github.com/foo/bar.git?ref=HEAD&depth=1",
+		},
+		{
+			name:        "SSH with directory path",
+			entrypoint:  "git@github.com:foo/bar.git//directory/Taskfile.yml?ref=dev",
+			expectedURL: "git::ssh://git@github.com/foo/bar.git?ref=dev&depth=1",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			node, err := NewGitNode(tt.entrypoint, "", false)
+			require.NoError(t, err)
+			gotURL := node.buildURL()
+			assert.Equal(t, tt.expectedURL, gotURL)
+		})
+	}
+}
+
+func TestRepoCacheKey_SameRepoSameRef(t *testing.T) {
+	t.Parallel()
+
+	// Same repo, same ref, different files should have SAME cache key
+	node1, err := NewGitNode("https://github.com/foo/bar.git//file1.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	node2, err := NewGitNode("https://github.com/foo/bar.git//dir/file2.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	key1 := node1.repoCacheKey()
+	key2 := node2.repoCacheKey()
+
+	assert.Equal(t, key1, key2, "Same repo+ref should generate same cache key regardless of file path")
+}
+
+func TestRepoCacheKey_SameRepoDifferentRef(t *testing.T) {
+	t.Parallel()
+
+	// Same repo, different ref should have DIFFERENT cache keys
+	node1, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	node2, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=dev", "", false)
+	require.NoError(t, err)
+
+	key1 := node1.repoCacheKey()
+	key2 := node2.repoCacheKey()
+
+	assert.NotEqual(t, key1, key2, "Different refs should generate different cache keys")
+}
+
+func TestRepoCacheKey_DifferentRepos(t *testing.T) {
+	t.Parallel()
+
+	// Different repos should have DIFFERENT cache keys
+	node1, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	node2, err := NewGitNode("https://github.com/foo/other.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	key1 := node1.repoCacheKey()
+	key2 := node2.repoCacheKey()
+
+	assert.NotEqual(t, key1, key2, "Different repos should generate different cache keys")
+}
+
+func TestRepoCacheKey_NoRefVsHead(t *testing.T) {
+	t.Parallel()
+
+	// No ref (defaults to HEAD) vs explicit HEAD should have SAME cache key
+	node1, err := NewGitNode("https://github.com/foo/bar.git//file.yml", "", false)
+	require.NoError(t, err)
+
+	node2, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=HEAD", "", false)
+	require.NoError(t, err)
+
+	key1 := node1.repoCacheKey()
+	key2 := node2.repoCacheKey()
+
+	assert.Equal(t, key1, key2, "No ref and explicit HEAD should generate same cache key")
+}
+
+func TestRepoCacheKey_SSHvsHTTPS(t *testing.T) {
+	t.Parallel()
+
+	// SSH vs HTTPS pointing to same repo should have SAME cache key
+	// They clone the same repo, so we want to share the cache
+	node1, err := NewGitNode("git@github.com:foo/bar.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	node2, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	key1 := node1.repoCacheKey()
+	key2 := node2.repoCacheKey()
+
+	assert.Equal(t, key1, key2, "SSH and HTTPS for same repo should share cache")
+}
+
+func TestRepoCacheKey_Consistency(t *testing.T) {
+	t.Parallel()
+
+	// Calling repoCacheKey multiple times on same node should return same key
+	node, err := NewGitNode("https://github.com/foo/bar.git//file.yml?ref=main", "", false)
+	require.NoError(t, err)
+
+	key1 := node.repoCacheKey()
+	key2 := node.repoCacheKey()
+	key3 := node.repoCacheKey()
+
+	assert.Equal(t, key1, key2)
+	assert.Equal(t, key2, key3)
+}

taskfile/reader.go

@@ -204,9 +204,15 @@ func (o *promptFuncOption) ApplyToReader(r *Reader) {
 // building an [ast.TaskfileGraph] as it goes. If any errors occur, they will be
 // returned immediately.
 func (r *Reader) Read(ctx context.Context, node Node) (*ast.TaskfileGraph, error) {
+	// Clean up git cache after reading all taskfiles
+	defer func() {
+		_ = CleanGitCache()
+	}()
+
 	if err := r.include(ctx, node); err != nil {
 		return nil, err
 	}
+
 	return r.graph, nil
 }