Add git.DIFF_RENAME_SIMILARITY_THRESHOLD option (#36164)

Make the threshold value passed to `git diff --find-renames` configurable

Author: silverwind

custom/conf/app.example.ini

@@ -733,6 +733,9 @@ LEVEL = Info
 ;DISABLE_CORE_PROTECT_NTFS=false
 ;; Disable the usage of using partial clones for git.
 ;DISABLE_PARTIAL_CLONE = false
+;; Set the similarity threshold passed to git commands via `--find-renames=<threshold>`.
+;; Default is 50%, the same as git. Must be a integer percentage between 0% and 100%.
+;DIFF_RENAME_SIMILARITY_THRESHOLD = 50%
 
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;; Git Operation timeout in seconds

modules/git/diff.go

@@ -16,6 +16,7 @@ import (
 
 	"code.gitea.io/gitea/modules/git/gitcmd"
 	"code.gitea.io/gitea/modules/log"
+	"code.gitea.io/gitea/modules/setting"
 )
 
 // RawDiffType type of a raw diff.
@@ -47,15 +48,19 @@ func GetRepoRawDiffForFile(repo *Repository, startCommit, endCommit string, diff
 	switch diffType {
 	case RawDiffNormal:
 		if len(startCommit) != 0 {
-			cmd.AddArguments("diff", "-M").AddDynamicArguments(startCommit, endCommit).AddDashesAndList(files...)
+			cmd.AddArguments("diff").
+				AddOptionFormat("--find-renames=%s", setting.Git.DiffRenameSimilarityThreshold).
+				AddDynamicArguments(startCommit, endCommit).AddDashesAndList(files...)
 		} else if commit.ParentCount() == 0 {
 			cmd.AddArguments("show").AddDynamicArguments(endCommit).AddDashesAndList(files...)
 		} else {
 			c, err := commit.Parent(0)
 			if err != nil {
 				return err
 			}
-			cmd.AddArguments("diff", "-M").AddDynamicArguments(c.ID.String(), endCommit).AddDashesAndList(files...)
+			cmd.AddArguments("diff").
+				AddOptionFormat("--find-renames=%s", setting.Git.DiffRenameSimilarityThreshold).
+				AddDynamicArguments(c.ID.String(), endCommit).AddDashesAndList(files...)
 		}
 	case RawDiffPatch:
 		if len(startCommit) != 0 {

modules/setting/git.go

@@ -5,6 +5,7 @@ package setting
 
 import (
 	"path/filepath"
+	"regexp"
 	"strings"
 	"time"
 
@@ -17,20 +18,21 @@ var Git = struct {
 	HomePath             string
 	DisableDiffHighlight bool
 
-	MaxGitDiffLines           int
-	MaxGitDiffLineCharacters  int
-	MaxGitDiffFiles           int
-	CommitsRangeSize          int // CommitsRangeSize the default commits range size
-	BranchesRangeSize         int // BranchesRangeSize the default branches range size
-	VerbosePush               bool
-	VerbosePushDelay          time.Duration
-	GCArgs                    []string `ini:"GC_ARGS" delim:" "`
-	EnableAutoGitWireProtocol bool
-	PullRequestPushMessage    bool
-	LargeObjectThreshold      int64
-	DisableCoreProtectNTFS    bool
-	DisablePartialClone       bool
-	Timeout                   struct {
+	MaxGitDiffLines               int
+	MaxGitDiffLineCharacters      int
+	MaxGitDiffFiles               int
+	CommitsRangeSize              int // CommitsRangeSize the default commits range size
+	BranchesRangeSize             int // BranchesRangeSize the default branches range size
+	VerbosePush                   bool
+	VerbosePushDelay              time.Duration
+	GCArgs                        []string `ini:"GC_ARGS" delim:" "`
+	EnableAutoGitWireProtocol     bool
+	PullRequestPushMessage        bool
+	LargeObjectThreshold          int64
+	DisableCoreProtectNTFS        bool
+	DisablePartialClone           bool
+	DiffRenameSimilarityThreshold string
+	Timeout                       struct {
 		Default int
 		Migrate int
 		Mirror  int
@@ -39,19 +41,20 @@ var Git = struct {
 		GC      int `ini:"GC"`
 	} `ini:"git.timeout"`
 }{
-	DisableDiffHighlight:      false,
-	MaxGitDiffLines:           1000,
-	MaxGitDiffLineCharacters:  5000,
-	MaxGitDiffFiles:           100,
-	CommitsRangeSize:          50,
-	BranchesRangeSize:         20,
-	VerbosePush:               true,
-	VerbosePushDelay:          5 * time.Second,
-	GCArgs:                    []string{},
-	EnableAutoGitWireProtocol: true,
-	PullRequestPushMessage:    true,
-	LargeObjectThreshold:      1024 * 1024,
-	DisablePartialClone:       false,
+	DisableDiffHighlight:          false,
+	MaxGitDiffLines:               1000,
+	MaxGitDiffLineCharacters:      5000,
+	MaxGitDiffFiles:               100,
+	CommitsRangeSize:              50,
+	BranchesRangeSize:             20,
+	VerbosePush:                   true,
+	VerbosePushDelay:              5 * time.Second,
+	GCArgs:                        []string{},
+	EnableAutoGitWireProtocol:     true,
+	PullRequestPushMessage:        true,
+	LargeObjectThreshold:          1024 * 1024,
+	DisablePartialClone:           false,
+	DiffRenameSimilarityThreshold: "50%",
 	Timeout: struct {
 		Default int
 		Migrate int
@@ -117,4 +120,9 @@ func loadGitFrom(rootCfg ConfigProvider) {
 	} else {
 		Git.HomePath = filepath.Clean(Git.HomePath)
 	}
+
+	// validate for a integer percentage between 0% and 100%
+	if !regexp.MustCompile(`^([0-9]|[1-9][0-9]|100)%$`).MatchString(Git.DiffRenameSimilarityThreshold) {
+		log.Fatal("Invalid git.DIFF_RENAME_SIMILARITY_THRESHOLD: %s", Git.DiffRenameSimilarityThreshold)
+	}
 }

services/gitdiff/git_diff_tree.go

@@ -15,6 +15,7 @@ import (
 	"code.gitea.io/gitea/modules/git"
 	"code.gitea.io/gitea/modules/git/gitcmd"
 	"code.gitea.io/gitea/modules/log"
+	"code.gitea.io/gitea/modules/setting"
 )
 
 type DiffTree struct {
@@ -56,7 +57,9 @@ func runGitDiffTree(ctx context.Context, gitRepo *git.Repository, useMergeBase b
 		return nil, err
 	}
 
-	cmd := gitcmd.NewCommand("diff-tree", "--raw", "-r", "--find-renames", "--root")
+	cmd := gitcmd.NewCommand("diff-tree", "--raw", "-r", "--root").
+		AddOptionFormat("--find-renames=%s", setting.Git.DiffRenameSimilarityThreshold)
+
 	if useMergeBase {
 		cmd.AddArguments("--merge-base")
 	}

services/gitdiff/gitdiff.go

@@ -1225,8 +1225,9 @@ func getDiffBasic(ctx context.Context, gitRepo *git.Repository, opts *DiffOption
 	}
 
 	cmdDiff := gitcmd.NewCommand().
-		AddArguments("diff", "--src-prefix=\\a/", "--dst-prefix=\\b/", "-M").
-		AddArguments(opts.WhitespaceBehavior...)
+		AddArguments("diff", "--src-prefix=\\a/", "--dst-prefix=\\b/").
+		AddArguments(opts.WhitespaceBehavior...).
+		AddOptionFormat("--find-renames=%s", setting.Git.DiffRenameSimilarityThreshold)
 
 	// In git 2.31, git diff learned --skip-to which we can use to shortcut skip to file
 	// so if we are using at least this version of git we don't have to tell ParsePatch to do