Fix submodule parsing (#32571)

Fix #32568, parse `.gitmodules` correctly

---------

Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>

modules/git/commit.go

@@ -9,7 +9,6 @@ import (
 	"bytes"
 	"context"
 	"errors"
-	"fmt"
 	"io"
 	"os/exec"
 	"strconv"
@@ -29,7 +28,7 @@ type Commit struct {
 	Signature     *CommitSignature
 
 	Parents        []ObjectID // ID strings
-	submoduleCache *ObjectCache
+	submoduleCache *ObjectCache[*SubModule]
 }
 
 // CommitSignature represents a git commit signature part.
@@ -357,69 +356,6 @@ func (c *Commit) GetFileContent(filename string, limit int) (string, error) {
 	return string(bytes), nil
 }
 
-// GetSubModules get all the sub modules of current revision git tree
-func (c *Commit) GetSubModules() (*ObjectCache, error) {
-	if c.submoduleCache != nil {
-		return c.submoduleCache, nil
-	}
-
-	entry, err := c.GetTreeEntryByPath(".gitmodules")
-	if err != nil {
-		if _, ok := err.(ErrNotExist); ok {
-			return nil, nil
-		}
-		return nil, err
-	}
-
-	rd, err := entry.Blob().DataAsync()
-	if err != nil {
-		return nil, err
-	}
-
-	defer rd.Close()
-	scanner := bufio.NewScanner(rd)
-	c.submoduleCache = newObjectCache()
-	var ismodule bool
-	var path string
-	for scanner.Scan() {
-		if strings.HasPrefix(scanner.Text(), "[submodule") {
-			ismodule = true
-			continue
-		}
-		if ismodule {
-			fields := strings.Split(scanner.Text(), "=")
-			k := strings.TrimSpace(fields[0])
-			if k == "path" {
-				path = strings.TrimSpace(fields[1])
-			} else if k == "url" {
-				c.submoduleCache.Set(path, &SubModule{path, strings.TrimSpace(fields[1])})
-				ismodule = false
-			}
-		}
-	}
-	if err = scanner.Err(); err != nil {
-		return nil, fmt.Errorf("GetSubModules scan: %w", err)
-	}
-
-	return c.submoduleCache, nil
-}
-
-// GetSubModule get the sub module according entryname
-func (c *Commit) GetSubModule(entryname string) (*SubModule, error) {
-	modules, err := c.GetSubModules()
-	if err != nil {
-		return nil, err
-	}
-
-	if modules != nil {
-		module, has := modules.Get(entryname)
-		if has {
-			return module.(*SubModule), nil
-		}
-	}
-	return nil, nil
-}
-
 // GetBranchName gets the closest branch name (as returned by 'git name-rev --name-only')
 func (c *Commit) GetBranchName() (string, error) {
 	cmd := NewCommand(c.repo.Ctx, "name-rev")

modules/git/commit_info.go

@@ -7,5 +7,5 @@ package git
 type CommitInfo struct {
 	Entry         *TreeEntry
 	Commit        *Commit
-	SubModuleFile *SubModuleFile
+	SubModuleFile *CommitSubModuleFile
 }

modules/git/commit_info_gogit.go

@@ -71,7 +71,7 @@ func (tes Entries) GetCommitsInfo(ctx context.Context, commit *Commit, treePath
 			commitsInfo[i].Commit = entryCommit
 		}
 
-		// If the entry if a submodule add a submodule file for this
+		// If the entry is a submodule add a submodule file for this
 		if entry.IsSubModule() {
 			subModuleURL := ""
 			var fullPath string
@@ -85,7 +85,7 @@ func (tes Entries) GetCommitsInfo(ctx context.Context, commit *Commit, treePath
 			} else if subModule != nil {
 				subModuleURL = subModule.URL
 			}
-			subModuleFile := NewSubModuleFile(commitsInfo[i].Commit, subModuleURL, entry.ID.String())
+			subModuleFile := NewCommitSubModuleFile(subModuleURL, entry.ID.String())
 			commitsInfo[i].SubModuleFile = subModuleFile
 		}
 	}

modules/git/commit_info_nogogit.go

@@ -79,7 +79,7 @@ func (tes Entries) GetCommitsInfo(ctx context.Context, commit *Commit, treePath
 			} else if subModule != nil {
 				subModuleURL = subModule.URL
 			}
-			subModuleFile := NewSubModuleFile(commitsInfo[i].Commit, subModuleURL, entry.ID.String())
+			subModuleFile := NewCommitSubModuleFile(subModuleURL, entry.ID.String())
 			commitsInfo[i].SubModuleFile = subModuleFile
 		}
 	}

modules/git/commit_submodule.go

@@ -0,0 +1,47 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+// GetSubModules get all the submodules of current revision git tree
+func (c *Commit) GetSubModules() (*ObjectCache[*SubModule], error) {
+	if c.submoduleCache != nil {
+		return c.submoduleCache, nil
+	}
+
+	entry, err := c.GetTreeEntryByPath(".gitmodules")
+	if err != nil {
+		if _, ok := err.(ErrNotExist); ok {
+			return nil, nil
+		}
+		return nil, err
+	}
+
+	rd, err := entry.Blob().DataAsync()
+	if err != nil {
+		return nil, err
+	}
+	defer rd.Close()
+
+	// at the moment we do not strictly limit the size of the .gitmodules file because some users would have huge .gitmodules files (>1MB)
+	c.submoduleCache, err = configParseSubModules(rd)
+	if err != nil {
+		return nil, err
+	}
+	return c.submoduleCache, nil
+}
+
+// GetSubModule get the submodule according entry name
+func (c *Commit) GetSubModule(entryName string) (*SubModule, error) {
+	modules, err := c.GetSubModules()
+	if err != nil {
+		return nil, err
+	}
+
+	if modules != nil {
+		if module, has := modules.Get(entryName); has {
+			return module, nil
+		}
+	}
+	return nil, nil
+}

modules/git/commit_submodule_file.go

@@ -15,24 +15,15 @@ import (
 
 var scpSyntax = regexp.MustCompile(`^([a-zA-Z0-9_]+@)?([a-zA-Z0-9._-]+):(.*)$`)
 
-// SubModule submodule is a reference on git repository
-type SubModule struct {
-	Name string
-	URL  string
-}
-
-// SubModuleFile represents a file with submodule type.
-type SubModuleFile struct {
-	*Commit
-
+// CommitSubModuleFile represents a file with submodule type.
+type CommitSubModuleFile struct {
 	refURL string
 	refID  string
 }
 
-// NewSubModuleFile create a new submodule file
-func NewSubModuleFile(c *Commit, refURL, refID string) *SubModuleFile {
-	return &SubModuleFile{
-		Commit: c,
+// NewCommitSubModuleFile create a new submodule file
+func NewCommitSubModuleFile(refURL, refID string) *CommitSubModuleFile {
+	return &CommitSubModuleFile{
 		refURL: refURL,
 		refID:  refID,
 	}
@@ -109,11 +100,12 @@ func getRefURL(refURL, urlPrefix, repoFullName, sshDomain string) string {
 }
 
 // RefURL guesses and returns reference URL.
-func (sf *SubModuleFile) RefURL(urlPrefix, repoFullName, sshDomain string) string {
+// FIXME: template passes AppURL as urlPrefix, it needs to figure out the correct approach (no hard-coded AppURL anymore)
+func (sf *CommitSubModuleFile) RefURL(urlPrefix, repoFullName, sshDomain string) string {
 	return getRefURL(sf.refURL, urlPrefix, repoFullName, sshDomain)
 }
 
 // RefID returns reference ID.
-func (sf *SubModuleFile) RefID() string {
+func (sf *CommitSubModuleFile) RefID() string {
 	return sf.refID
 }

modules/git/commit_submodule_file_test.go

@@ -9,7 +9,7 @@ import (
 	"github.com/stretchr/testify/assert"
 )
 
-func TestGetRefURL(t *testing.T) {
+func TestCommitSubModuleFileGetRefURL(t *testing.T) {
 	kases := []struct {
 		refURL     string
 		prefixURL  string

modules/git/commit_test.go

@@ -135,7 +135,7 @@ author KN4CK3R <admin@oldschoolhack.me> 1711702962 +0100
 committer KN4CK3R <admin@oldschoolhack.me> 1711702962 +0100
 encoding ISO-8859-1
 gpgsig -----BEGIN PGP SIGNATURE-----
- 
+<SPACE>
  iQGzBAABCgAdFiEE9HRrbqvYxPT8PXbefPSEkrowAa8FAmYGg7IACgkQfPSEkrow
  Aa9olwv+P0HhtCM6CRvlUmPaqswRsDPNR4i66xyXGiSxdI9V5oJL7HLiQIM7KrFR
  gizKa2COiGtugv8fE+TKqXKaJx6uJUJEjaBd8E9Af9PrAzjWj+A84lU6/PgPS8hq
@@ -150,7 +150,7 @@ gpgsig -----BEGIN PGP SIGNATURE-----
  -----END PGP SIGNATURE-----
 
 ISO-8859-1`
-
+	commitString = strings.ReplaceAll(commitString, "<SPACE>", " ")
 	sha := &Sha1Hash{0xfe, 0xaf, 0x4b, 0xa6, 0xbc, 0x63, 0x5f, 0xec, 0x44, 0x2f, 0x46, 0xdd, 0xd4, 0x51, 0x24, 0x16, 0xec, 0x43, 0xc2, 0xc2}
 	gitRepo, err := openRepositoryWithDefaultContext(filepath.Join(testReposDir, "repo1_bare"))
 	assert.NoError(t, err)

modules/git/config.go

@@ -0,0 +1,187 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+import (
+	"fmt"
+	"os"
+	"regexp"
+	"runtime"
+	"strings"
+
+	"code.gitea.io/gitea/modules/setting"
+)
+
+// syncGitConfig only modifies gitconfig, won't change global variables (otherwise there will be data-race problem)
+func syncGitConfig() (err error) {
+	if err = os.MkdirAll(HomeDir(), os.ModePerm); err != nil {
+		return fmt.Errorf("unable to prepare git home directory %s, err: %w", HomeDir(), err)
+	}
+
+	// first, write user's git config options to git config file
+	// user config options could be overwritten by builtin values later, because if a value is builtin, it must have some special purposes
+	for k, v := range setting.GitConfig.Options {
+		if err = configSet(strings.ToLower(k), v); err != nil {
+			return err
+		}
+	}
+
+	// Git requires setting user.name and user.email in order to commit changes - old comment: "if they're not set just add some defaults"
+	// TODO: need to confirm whether users really need to change these values manually. It seems that these values are dummy only and not really used.
+	// If these values are not really used, then they can be set (overwritten) directly without considering about existence.
+	for configKey, defaultValue := range map[string]string{
+		"user.name":  "Gitea",
+		"user.email": "gitea@fake.local",
+	} {
+		if err := configSetNonExist(configKey, defaultValue); err != nil {
+			return err
+		}
+	}
+
+	// Set git some configurations - these must be set to these values for gitea to work correctly
+	if err := configSet("core.quotePath", "false"); err != nil {
+		return err
+	}
+
+	if DefaultFeatures().CheckVersionAtLeast("2.10") {
+		if err := configSet("receive.advertisePushOptions", "true"); err != nil {
+			return err
+		}
+	}
+
+	if DefaultFeatures().CheckVersionAtLeast("2.18") {
+		if err := configSet("core.commitGraph", "true"); err != nil {
+			return err
+		}
+		if err := configSet("gc.writeCommitGraph", "true"); err != nil {
+			return err
+		}
+		if err := configSet("fetch.writeCommitGraph", "true"); err != nil {
+			return err
+		}
+	}
+
+	if DefaultFeatures().SupportProcReceive {
+		// set support for AGit flow
+		if err := configAddNonExist("receive.procReceiveRefs", "refs/for"); err != nil {
+			return err
+		}
+	} else {
+		if err := configUnsetAll("receive.procReceiveRefs", "refs/for"); err != nil {
+			return err
+		}
+	}
+
+	// Due to CVE-2022-24765, git now denies access to git directories which are not owned by current user.
+	// However, some docker users and samba users find it difficult to configure their systems correctly,
+	// so that Gitea's git repositories are owned by the Gitea user.
+	// (Possibly Windows Service users - but ownership in this case should really be set correctly on the filesystem.)
+	// See issue: https://github.com/go-gitea/gitea/issues/19455
+	// As Gitea now always use its internal git config file, and access to the git repositories is managed through Gitea,
+	// it is now safe to set "safe.directory=*" for internal usage only.
+	// Although this setting is only supported by some new git versions, it is also tolerated by earlier versions
+	if err := configAddNonExist("safe.directory", "*"); err != nil {
+		return err
+	}
+
+	if runtime.GOOS == "windows" {
+		if err := configSet("core.longpaths", "true"); err != nil {
+			return err
+		}
+		if setting.Git.DisableCoreProtectNTFS {
+			err = configSet("core.protectNTFS", "false")
+		} else {
+			err = configUnsetAll("core.protectNTFS", "false")
+		}
+		if err != nil {
+			return err
+		}
+	}
+
+	// By default partial clones are disabled, enable them from git v2.22
+	if !setting.Git.DisablePartialClone && DefaultFeatures().CheckVersionAtLeast("2.22") {
+		if err = configSet("uploadpack.allowfilter", "true"); err != nil {
+			return err
+		}
+		err = configSet("uploadpack.allowAnySHA1InWant", "true")
+	} else {
+		if err = configUnsetAll("uploadpack.allowfilter", "true"); err != nil {
+			return err
+		}
+		err = configUnsetAll("uploadpack.allowAnySHA1InWant", "true")
+	}
+
+	return err
+}
+
+func configSet(key, value string) error {
+	stdout, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
+	if err != nil && !IsErrorExitCode(err, 1) {
+		return fmt.Errorf("failed to get git config %s, err: %w", key, err)
+	}
+
+	currValue := strings.TrimSpace(stdout)
+	if currValue == value {
+		return nil
+	}
+
+	_, _, err = NewCommand(DefaultContext, "config", "--global").AddDynamicArguments(key, value).RunStdString(nil)
+	if err != nil {
+		return fmt.Errorf("failed to set git global config %s, err: %w", key, err)
+	}
+
+	return nil
+}
+
+func configSetNonExist(key, value string) error {
+	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
+	if err == nil {
+		// already exist
+		return nil
+	}
+	if IsErrorExitCode(err, 1) {
+		// not exist, set new config
+		_, _, err = NewCommand(DefaultContext, "config", "--global").AddDynamicArguments(key, value).RunStdString(nil)
+		if err != nil {
+			return fmt.Errorf("failed to set git global config %s, err: %w", key, err)
+		}
+		return nil
+	}
+
+	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
+}
+
+func configAddNonExist(key, value string) error {
+	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key, regexp.QuoteMeta(value)).RunStdString(nil)
+	if err == nil {
+		// already exist
+		return nil
+	}
+	if IsErrorExitCode(err, 1) {
+		// not exist, add new config
+		_, _, err = NewCommand(DefaultContext, "config", "--global", "--add").AddDynamicArguments(key, value).RunStdString(nil)
+		if err != nil {
+			return fmt.Errorf("failed to add git global config %s, err: %w", key, err)
+		}
+		return nil
+	}
+	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
+}
+
+func configUnsetAll(key, value string) error {
+	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
+	if err == nil {
+		// exist, need to remove
+		_, _, err = NewCommand(DefaultContext, "config", "--global", "--unset-all").AddDynamicArguments(key, regexp.QuoteMeta(value)).RunStdString(nil)
+		if err != nil {
+			return fmt.Errorf("failed to unset git global config %s, err: %w", key, err)
+		}
+		return nil
+	}
+	if IsErrorExitCode(err, 1) {
+		// not exist
+		return nil
+	}
+	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
+}

modules/git/config_submodule.go

@@ -0,0 +1,75 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+import (
+	"bufio"
+	"fmt"
+	"io"
+	"strings"
+)
+
+// SubModule is a reference on git repository
+type SubModule struct {
+	Path   string
+	URL    string
+	Branch string // this field is newly added but not really used
+}
+
+// configParseSubModules this is not a complete parse for gitmodules file, it only
+// parses the url and path of submodules. At the moment it only parses well-formed gitmodules files.
+// In the future, there should be a complete implementation of https://git-scm.com/docs/git-config#_syntax
+func configParseSubModules(r io.Reader) (*ObjectCache[*SubModule], error) {
+	var subModule *SubModule
+	subModules := newObjectCache[*SubModule]()
+	scanner := bufio.NewScanner(r)
+	for scanner.Scan() {
+		line := strings.TrimSpace(scanner.Text())
+
+		// Skip empty lines and comments
+		if line == "" || strings.HasPrefix(line, "#") || strings.HasPrefix(line, ";") {
+			continue
+		}
+
+		// Section header [section]
+		if strings.HasPrefix(line, "[") && strings.HasSuffix(line, "]") {
+			if subModule != nil {
+				subModules.Set(subModule.Path, subModule)
+			}
+			if strings.HasPrefix(line, "[submodule") {
+				subModule = &SubModule{}
+			} else {
+				subModule = nil
+			}
+			continue
+		}
+
+		if subModule == nil {
+			continue
+		}
+
+		parts := strings.SplitN(line, "=", 2)
+		if len(parts) != 2 {
+			continue
+		}
+		key := strings.TrimSpace(parts[0])
+		value := strings.TrimSpace(parts[1])
+		switch key {
+		case "path":
+			subModule.Path = value
+		case "url":
+			subModule.URL = value
+		case "branch":
+			subModule.Branch = value
+		}
+	}
+
+	if err := scanner.Err(); err != nil {
+		return nil, fmt.Errorf("error reading file: %w", err)
+	}
+	if subModule != nil {
+		subModules.Set(subModule.Path, subModule)
+	}
+	return subModules, nil
+}

modules/git/config_submodule_test.go

@@ -0,0 +1,49 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestConfigSubmodule(t *testing.T) {
+	input := `
+[core]
+path = test
+
+[submodule "submodule1"]
+	path = path1
+	url =	https://gitea.io/foo/foo
+	#branch = b1
+
+[other1]
+branch = master
+
+[submodule "submodule2"]
+	path = path2
+	url =	https://gitea.io/bar/bar
+	branch = b2
+
+[other2]
+branch = main
+
+[submodule "submodule3"]
+	path = path3
+	url =	https://gitea.io/xxx/xxx
+`
+
+	subModules, err := configParseSubModules(strings.NewReader(input))
+	assert.NoError(t, err)
+	assert.Len(t, subModules.cache, 3)
+
+	sm1, _ := subModules.Get("path1")
+	assert.Equal(t, &SubModule{Path: "path1", URL: "https://gitea.io/foo/foo", Branch: ""}, sm1)
+	sm2, _ := subModules.Get("path2")
+	assert.Equal(t, &SubModule{Path: "path2", URL: "https://gitea.io/bar/bar", Branch: "b2"}, sm2)
+	sm3, _ := subModules.Get("path3")
+	assert.Equal(t, &SubModule{Path: "path3", URL: "https://gitea.io/xxx/xxx", Branch: ""}, sm3)
+}

modules/git/config_test.go

@@ -0,0 +1,66 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+import (
+	"os"
+	"strings"
+	"testing"
+
+	"code.gitea.io/gitea/modules/setting"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func gitConfigContains(sub string) bool {
+	if b, err := os.ReadFile(HomeDir() + "/.gitconfig"); err == nil {
+		return strings.Contains(string(b), sub)
+	}
+	return false
+}
+
+func TestGitConfig(t *testing.T) {
+	assert.False(t, gitConfigContains("key-a"))
+
+	assert.NoError(t, configSetNonExist("test.key-a", "val-a"))
+	assert.True(t, gitConfigContains("key-a = val-a"))
+
+	assert.NoError(t, configSetNonExist("test.key-a", "val-a-changed"))
+	assert.False(t, gitConfigContains("key-a = val-a-changed"))
+
+	assert.NoError(t, configSet("test.key-a", "val-a-changed"))
+	assert.True(t, gitConfigContains("key-a = val-a-changed"))
+
+	assert.NoError(t, configAddNonExist("test.key-b", "val-b"))
+	assert.True(t, gitConfigContains("key-b = val-b"))
+
+	assert.NoError(t, configAddNonExist("test.key-b", "val-2b"))
+	assert.True(t, gitConfigContains("key-b = val-b"))
+	assert.True(t, gitConfigContains("key-b = val-2b"))
+
+	assert.NoError(t, configUnsetAll("test.key-b", "val-b"))
+	assert.False(t, gitConfigContains("key-b = val-b"))
+	assert.True(t, gitConfigContains("key-b = val-2b"))
+
+	assert.NoError(t, configUnsetAll("test.key-b", "val-2b"))
+	assert.False(t, gitConfigContains("key-b = val-2b"))
+
+	assert.NoError(t, configSet("test.key-x", "*"))
+	assert.True(t, gitConfigContains("key-x = *"))
+	assert.NoError(t, configSetNonExist("test.key-x", "*"))
+	assert.NoError(t, configUnsetAll("test.key-x", "*"))
+	assert.False(t, gitConfigContains("key-x = *"))
+}
+
+func TestSyncConfig(t *testing.T) {
+	oldGitConfig := setting.GitConfig
+	defer func() {
+		setting.GitConfig = oldGitConfig
+	}()
+
+	setting.GitConfig.Options["sync-test.cfg-key-a"] = "CfgValA"
+	assert.NoError(t, syncGitConfig())
+	assert.True(t, gitConfigContains("[sync-test]"))
+	assert.True(t, gitConfigContains("cfg-key-a = CfgValA"))
+}

modules/git/fsck.go

@@ -0,0 +1,14 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package git
+
+import (
+	"context"
+	"time"
+)
+
+// Fsck verifies the connectivity and validity of the objects in the database
+func Fsck(ctx context.Context, repoPath string, timeout time.Duration, args TrustedCmdArgs) error {
+	return NewCommand(ctx, "fsck").AddArguments(args...).Run(&RunOpts{Timeout: timeout, Dir: repoPath})
+}

modules/git/git.go

@@ -11,7 +11,6 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
-	"regexp"
 	"runtime"
 	"strings"
 	"time"
@@ -95,17 +94,18 @@ func parseGitVersionLine(s string) (*version.Version, error) {
 	return version.NewVersion(versionString)
 }
 
-// SetExecutablePath changes the path of git executable and checks the file permission and version.
-func SetExecutablePath(path string) error {
-	// If path is empty, we use the default value of GitExecutable "git" to search for the location of git.
-	if path != "" {
-		GitExecutable = path
+func checkGitVersionCompatibility(gitVer *version.Version) error {
+	badVersions := []struct {
+		Version *version.Version
+		Reason  string
+	}{
+		{version.Must(version.NewVersion("2.43.1")), "regression bug of GIT_FLUSH"},
 	}
-	absPath, err := exec.LookPath(GitExecutable)
-	if err != nil {
-		return fmt.Errorf("git not found: %w", err)
+	for _, bad := range badVersions {
+		if gitVer.Equal(bad.Version) {
+			return errors.New(bad.Reason)
+		}
 	}
-	GitExecutable = absPath
 	return nil
 }
 
@@ -128,6 +128,20 @@ func ensureGitVersion() error {
 	return nil
 }
 
+// SetExecutablePath changes the path of git executable and checks the file permission and version.
+func SetExecutablePath(path string) error {
+	// If path is empty, we use the default value of GitExecutable "git" to search for the location of git.
+	if path != "" {
+		GitExecutable = path
+	}
+	absPath, err := exec.LookPath(GitExecutable)
+	if err != nil {
+		return fmt.Errorf("git not found: %w", err)
+	}
+	GitExecutable = absPath
+	return nil
+}
+
 // HomeDir is the home dir for git to store the global config file used by Gitea internally
 func HomeDir() string {
 	if setting.Git.HomePath == "" {
@@ -204,196 +218,3 @@ func InitFull(ctx context.Context) (err error) {
 
 	return syncGitConfig()
 }
-
-// syncGitConfig only modifies gitconfig, won't change global variables (otherwise there will be data-race problem)
-func syncGitConfig() (err error) {
-	if err = os.MkdirAll(HomeDir(), os.ModePerm); err != nil {
-		return fmt.Errorf("unable to prepare git home directory %s, err: %w", HomeDir(), err)
-	}
-
-	// first, write user's git config options to git config file
-	// user config options could be overwritten by builtin values later, because if a value is builtin, it must have some special purposes
-	for k, v := range setting.GitConfig.Options {
-		if err = configSet(strings.ToLower(k), v); err != nil {
-			return err
-		}
-	}
-
-	// Git requires setting user.name and user.email in order to commit changes - old comment: "if they're not set just add some defaults"
-	// TODO: need to confirm whether users really need to change these values manually. It seems that these values are dummy only and not really used.
-	// If these values are not really used, then they can be set (overwritten) directly without considering about existence.
-	for configKey, defaultValue := range map[string]string{
-		"user.name":  "Gitea",
-		"user.email": "gitea@fake.local",
-	} {
-		if err := configSetNonExist(configKey, defaultValue); err != nil {
-			return err
-		}
-	}
-
-	// Set git some configurations - these must be set to these values for gitea to work correctly
-	if err := configSet("core.quotePath", "false"); err != nil {
-		return err
-	}
-
-	if DefaultFeatures().CheckVersionAtLeast("2.10") {
-		if err := configSet("receive.advertisePushOptions", "true"); err != nil {
-			return err
-		}
-	}
-
-	if DefaultFeatures().CheckVersionAtLeast("2.18") {
-		if err := configSet("core.commitGraph", "true"); err != nil {
-			return err
-		}
-		if err := configSet("gc.writeCommitGraph", "true"); err != nil {
-			return err
-		}
-		if err := configSet("fetch.writeCommitGraph", "true"); err != nil {
-			return err
-		}
-	}
-
-	if DefaultFeatures().SupportProcReceive {
-		// set support for AGit flow
-		if err := configAddNonExist("receive.procReceiveRefs", "refs/for"); err != nil {
-			return err
-		}
-	} else {
-		if err := configUnsetAll("receive.procReceiveRefs", "refs/for"); err != nil {
-			return err
-		}
-	}
-
-	// Due to CVE-2022-24765, git now denies access to git directories which are not owned by current user.
-	// However, some docker users and samba users find it difficult to configure their systems correctly,
-	// so that Gitea's git repositories are owned by the Gitea user.
-	// (Possibly Windows Service users - but ownership in this case should really be set correctly on the filesystem.)
-	// See issue: https://github.com/go-gitea/gitea/issues/19455
-	// As Gitea now always use its internal git config file, and access to the git repositories is managed through Gitea,
-	// it is now safe to set "safe.directory=*" for internal usage only.
-	// Although this setting is only supported by some new git versions, it is also tolerated by earlier versions
-	if err := configAddNonExist("safe.directory", "*"); err != nil {
-		return err
-	}
-
-	if runtime.GOOS == "windows" {
-		if err := configSet("core.longpaths", "true"); err != nil {
-			return err
-		}
-		if setting.Git.DisableCoreProtectNTFS {
-			err = configSet("core.protectNTFS", "false")
-		} else {
-			err = configUnsetAll("core.protectNTFS", "false")
-		}
-		if err != nil {
-			return err
-		}
-	}
-
-	// By default partial clones are disabled, enable them from git v2.22
-	if !setting.Git.DisablePartialClone && DefaultFeatures().CheckVersionAtLeast("2.22") {
-		if err = configSet("uploadpack.allowfilter", "true"); err != nil {
-			return err
-		}
-		err = configSet("uploadpack.allowAnySHA1InWant", "true")
-	} else {
-		if err = configUnsetAll("uploadpack.allowfilter", "true"); err != nil {
-			return err
-		}
-		err = configUnsetAll("uploadpack.allowAnySHA1InWant", "true")
-	}
-
-	return err
-}
-
-func checkGitVersionCompatibility(gitVer *version.Version) error {
-	badVersions := []struct {
-		Version *version.Version
-		Reason  string
-	}{
-		{version.Must(version.NewVersion("2.43.1")), "regression bug of GIT_FLUSH"},
-	}
-	for _, bad := range badVersions {
-		if gitVer.Equal(bad.Version) {
-			return errors.New(bad.Reason)
-		}
-	}
-	return nil
-}
-
-func configSet(key, value string) error {
-	stdout, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
-	if err != nil && !IsErrorExitCode(err, 1) {
-		return fmt.Errorf("failed to get git config %s, err: %w", key, err)
-	}
-
-	currValue := strings.TrimSpace(stdout)
-	if currValue == value {
-		return nil
-	}
-
-	_, _, err = NewCommand(DefaultContext, "config", "--global").AddDynamicArguments(key, value).RunStdString(nil)
-	if err != nil {
-		return fmt.Errorf("failed to set git global config %s, err: %w", key, err)
-	}
-
-	return nil
-}
-
-func configSetNonExist(key, value string) error {
-	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
-	if err == nil {
-		// already exist
-		return nil
-	}
-	if IsErrorExitCode(err, 1) {
-		// not exist, set new config
-		_, _, err = NewCommand(DefaultContext, "config", "--global").AddDynamicArguments(key, value).RunStdString(nil)
-		if err != nil {
-			return fmt.Errorf("failed to set git global config %s, err: %w", key, err)
-		}
-		return nil
-	}
-
-	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
-}
-
-func configAddNonExist(key, value string) error {
-	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key, regexp.QuoteMeta(value)).RunStdString(nil)
-	if err == nil {
-		// already exist
-		return nil
-	}
-	if IsErrorExitCode(err, 1) {
-		// not exist, add new config
-		_, _, err = NewCommand(DefaultContext, "config", "--global", "--add").AddDynamicArguments(key, value).RunStdString(nil)
-		if err != nil {
-			return fmt.Errorf("failed to add git global config %s, err: %w", key, err)
-		}
-		return nil
-	}
-	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
-}
-
-func configUnsetAll(key, value string) error {
-	_, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
-	if err == nil {
-		// exist, need to remove
-		_, _, err = NewCommand(DefaultContext, "config", "--global", "--unset-all").AddDynamicArguments(key, regexp.QuoteMeta(value)).RunStdString(nil)
-		if err != nil {
-			return fmt.Errorf("failed to unset git global config %s, err: %w", key, err)
-		}
-		return nil
-	}
-	if IsErrorExitCode(err, 1) {
-		// not exist
-		return nil
-	}
-	return fmt.Errorf("failed to get git config %s, err: %w", key, err)
-}
-
-// Fsck verifies the connectivity and validity of the objects in the database
-func Fsck(ctx context.Context, repoPath string, timeout time.Duration, args TrustedCmdArgs) error {
-	return NewCommand(ctx, "fsck").AddArguments(args...).Run(&RunOpts{Timeout: timeout, Dir: repoPath})
-}

modules/git/git_test.go

@@ -7,7 +7,6 @@ import (
 	"context"
 	"fmt"
 	"os"
-	"strings"
 	"testing"
 
 	"code.gitea.io/gitea/modules/setting"
@@ -43,58 +42,6 @@ func TestMain(m *testing.M) {
 	}
 }
 
-func gitConfigContains(sub string) bool {
-	if b, err := os.ReadFile(HomeDir() + "/.gitconfig"); err == nil {
-		return strings.Contains(string(b), sub)
-	}
-	return false
-}
-
-func TestGitConfig(t *testing.T) {
-	assert.False(t, gitConfigContains("key-a"))
-
-	assert.NoError(t, configSetNonExist("test.key-a", "val-a"))
-	assert.True(t, gitConfigContains("key-a = val-a"))
-
-	assert.NoError(t, configSetNonExist("test.key-a", "val-a-changed"))
-	assert.False(t, gitConfigContains("key-a = val-a-changed"))
-
-	assert.NoError(t, configSet("test.key-a", "val-a-changed"))
-	assert.True(t, gitConfigContains("key-a = val-a-changed"))
-
-	assert.NoError(t, configAddNonExist("test.key-b", "val-b"))
-	assert.True(t, gitConfigContains("key-b = val-b"))
-
-	assert.NoError(t, configAddNonExist("test.key-b", "val-2b"))
-	assert.True(t, gitConfigContains("key-b = val-b"))
-	assert.True(t, gitConfigContains("key-b = val-2b"))
-
-	assert.NoError(t, configUnsetAll("test.key-b", "val-b"))
-	assert.False(t, gitConfigContains("key-b = val-b"))
-	assert.True(t, gitConfigContains("key-b = val-2b"))
-
-	assert.NoError(t, configUnsetAll("test.key-b", "val-2b"))
-	assert.False(t, gitConfigContains("key-b = val-2b"))
-
-	assert.NoError(t, configSet("test.key-x", "*"))
-	assert.True(t, gitConfigContains("key-x = *"))
-	assert.NoError(t, configSetNonExist("test.key-x", "*"))
-	assert.NoError(t, configUnsetAll("test.key-x", "*"))
-	assert.False(t, gitConfigContains("key-x = *"))
-}
-
-func TestSyncConfig(t *testing.T) {
-	oldGitConfig := setting.GitConfig
-	defer func() {
-		setting.GitConfig = oldGitConfig
-	}()
-
-	setting.GitConfig.Options["sync-test.cfg-key-a"] = "CfgValA"
-	assert.NoError(t, syncGitConfig())
-	assert.True(t, gitConfigContains("[sync-test]"))
-	assert.True(t, gitConfigContains("cfg-key-a = CfgValA"))
-}
-
 func TestParseGitVersion(t *testing.T) {
 	v, err := parseGitVersionLine("git version 2.29.3")
 	assert.NoError(t, err)

modules/git/repo_base_gogit.go

@@ -28,7 +28,7 @@ const isGogit = true
 type Repository struct {
 	Path string
 
-	tagCache *ObjectCache
+	tagCache *ObjectCache[*Tag]
 
 	gogitRepo    *gogit.Repository
 	gogitStorage *filesystem.Storage
@@ -79,7 +79,7 @@ func OpenRepository(ctx context.Context, repoPath string) (*Repository, error) {
 		Path:         repoPath,
 		gogitRepo:    gogitRepo,
 		gogitStorage: storage,
-		tagCache:     newObjectCache(),
+		tagCache:     newObjectCache[*Tag](),
 		Ctx:          ctx,
 		objectFormat: ParseGogitHash(plumbing.ZeroHash).Type(),
 	}, nil

modules/git/repo_base_nogogit.go

@@ -21,7 +21,7 @@ const isGogit = false
 type Repository struct {
 	Path string
 
-	tagCache *ObjectCache
+	tagCache *ObjectCache[*Tag]
 
 	gpgSettings *GPGSettings
 
@@ -53,7 +53,7 @@ func OpenRepository(ctx context.Context, repoPath string) (*Repository, error) {
 
 	return &Repository{
 		Path:     repoPath,
-		tagCache: newObjectCache(),
+		tagCache: newObjectCache[*Tag](),
 		Ctx:      ctx,
 	}, nil
 }

modules/git/repo_tag_gogit.go

@@ -72,7 +72,7 @@ func (repo *Repository) getTag(tagID ObjectID, name string) (*Tag, error) {
 	t, ok := repo.tagCache.Get(tagID.String())
 	if ok {
 		log.Debug("Hit cache: %s", tagID)
-		tagClone := *t.(*Tag)
+		tagClone := *t
 		tagClone.Name = name // This is necessary because lightweight tags may have same id
 		return &tagClone, nil
 	}

modules/git/repo_tag_nogogit.go

@@ -51,7 +51,7 @@ func (repo *Repository) getTag(tagID ObjectID, name string) (*Tag, error) {
 	t, ok := repo.tagCache.Get(tagID.String())
 	if ok {
 		log.Debug("Hit cache: %s", tagID)
-		tagClone := *t.(*Tag)
+		tagClone := *t
 		tagClone.Name = name // This is necessary because lightweight tags may have same id
 		return &tagClone, nil
 	}

modules/git/utils.go

@@ -15,27 +15,25 @@ import (
 )
 
 // ObjectCache provides thread-safe cache operations.
-type ObjectCache struct {
+type ObjectCache[T any] struct {
 	lock  sync.RWMutex
-	cache map[string]any
+	cache map[string]T
 }
 
-func newObjectCache() *ObjectCache {
-	return &ObjectCache{
-		cache: make(map[string]any, 10),
-	}
+func newObjectCache[T any]() *ObjectCache[T] {
+	return &ObjectCache[T]{cache: make(map[string]T, 10)}
 }
 
-// Set add obj to cache
-func (oc *ObjectCache) Set(id string, obj any) {
+// Set adds obj to cache
+func (oc *ObjectCache[T]) Set(id string, obj T) {
 	oc.lock.Lock()
 	defer oc.lock.Unlock()
 
 	oc.cache[id] = obj
 }
 
-// Get get cached obj by id
-func (oc *ObjectCache) Get(id string) (any, bool) {
+// Get gets cached obj by id
+func (oc *ObjectCache[T]) Get(id string) (T, bool) {
 	oc.lock.RLock()
 	defer oc.lock.RUnlock()