Use merge tree to improve conflict checking performance when possible

modules/git/commit_test.go

@@ -4,7 +4,6 @@
 package git
 
 import (
-	"os"
 	"path/filepath"
 	"strings"
 	"testing"
@@ -339,18 +338,3 @@ func TestGetCommitFileStatusMerges(t *testing.T) {
 	assert.Equal(t, expected.Removed, commitFileStatus.Removed)
 	assert.Equal(t, expected.Modified, commitFileStatus.Modified)
 }
-
-func Test_GetCommitBranchStart(t *testing.T) {
-	bareRepo1Path := filepath.Join(testReposDir, "repo1_bare")
-	repo, err := OpenRepository(t.Context(), bareRepo1Path)
-	assert.NoError(t, err)
-	defer repo.Close()
-	commit, err := repo.GetBranchCommit("branch1")
-	assert.NoError(t, err)
-	assert.Equal(t, "2839944139e0de9737a044f78b0e4b40d989a9e3", commit.ID.String())
-
-	startCommitID, err := repo.GetCommitBranchStart(os.Environ(), "branch1", commit.ID.String())
-	assert.NoError(t, err)
-	assert.NotEmpty(t, startCommitID)
-	assert.Equal(t, "95bb4d39648ee7e325106df01a621c530863a653", startCommitID)
-}

modules/git/diff.go

@@ -7,6 +7,7 @@ import (
 	"bufio"
 	"bytes"
 	"context"
+	"errors"
 	"fmt"
 	"io"
 	"os"
@@ -289,20 +290,18 @@ func CutDiffAroundLine(originalDiff io.Reader, line int64, old bool, numbersOfLi
 }
 
 // GetAffectedFiles returns the affected files between two commits
-func GetAffectedFiles(repo *Repository, branchName, oldCommitID, newCommitID string, env []string) ([]string, error) {
-	if oldCommitID == emptySha1ObjectID.String() || oldCommitID == emptySha256ObjectID.String() {
-		startCommitID, err := repo.GetCommitBranchStart(env, branchName, newCommitID)
-		if err != nil {
-			return nil, err
-		}
-		if startCommitID == "" {
-			return nil, fmt.Errorf("cannot find the start commit of %s", newCommitID)
-		}
-		oldCommitID = startCommitID
+func GetAffectedFiles(ctx context.Context, repoPath, oldCommitID, newCommitID string, env []string) ([]string, error) {
+	if oldCommitID == emptySha1ObjectID.String() {
+		oldCommitID = emptySha1ObjectID.Type().EmptyTree().String()
+	} else if oldCommitID == emptySha256ObjectID.String() {
+		oldCommitID = emptySha256ObjectID.Type().EmptyTree().String()
+	} else if oldCommitID == "" {
+		return nil, errors.New("oldCommitID is empty")
 	}
+
 	stdoutReader, stdoutWriter, err := os.Pipe()
 	if err != nil {
-		log.Error("Unable to create os.Pipe for %s", repo.Path)
+		log.Error("Unable to create os.Pipe for %s", repoPath)
 		return nil, err
 	}
 	defer func() {
@@ -314,9 +313,9 @@ func GetAffectedFiles(repo *Repository, branchName, oldCommitID, newCommitID str
 
 	// Run `git diff --name-only` to get the names of the changed files
 	err = gitcmd.NewCommand("diff", "--name-only").AddDynamicArguments(oldCommitID, newCommitID).
-		Run(repo.Ctx, &gitcmd.RunOpts{
+		Run(ctx, &gitcmd.RunOpts{
 			Env:    env,
-			Dir:    repo.Path,
+			Dir:    repoPath,
 			Stdout: stdoutWriter,
 			PipelineFunc: func(ctx context.Context, cancel context.CancelFunc) error {
 				// Close the writer end of the pipe to begin processing
@@ -338,7 +337,7 @@ func GetAffectedFiles(repo *Repository, branchName, oldCommitID, newCommitID str
 			},
 		})
 	if err != nil {
-		log.Error("Unable to get affected files for commits from %s to %s in %s: %v", oldCommitID, newCommitID, repo.Path, err)
+		log.Error("Unable to get affected files for commits from %s to %s in %s: %v", oldCommitID, newCommitID, repoPath, err)
 	}
 
 	return affectedFiles, err

modules/git/git.go

@@ -31,6 +31,7 @@ type Features struct {
 	SupportHashSha256      bool           // >= 2.42, SHA-256 repositories no longer an ‘experimental curiosity’
 	SupportedObjectFormats []ObjectFormat // sha1, sha256
 	SupportCheckAttrOnBare bool           // >= 2.40
+	SupportGitMergeTree    bool           // >= 2.38
 }
 
 var defaultFeatures *Features
@@ -75,6 +76,7 @@ func loadGitVersionFeatures() (*Features, error) {
 		features.SupportedObjectFormats = append(features.SupportedObjectFormats, Sha256ObjectFormat)
 	}
 	features.SupportCheckAttrOnBare = features.CheckVersionAtLeast("2.40")
+	features.SupportGitMergeTree = features.CheckVersionAtLeast("2.38")
 	return features, nil
 }
 

modules/git/repo_commit.go

@@ -534,35 +534,3 @@ func (repo *Repository) AddLastCommitCache(cacheKey, fullName, sha string) error
 	}
 	return nil
 }
-
-// GetCommitBranchStart returns the commit where the branch diverged
-func (repo *Repository) GetCommitBranchStart(env []string, branch, endCommitID string) (string, error) {
-	cmd := gitcmd.NewCommand("log", prettyLogFormat)
-	cmd.AddDynamicArguments(endCommitID)
-
-	stdout, _, runErr := cmd.RunStdBytes(repo.Ctx, &gitcmd.RunOpts{
-		Dir: repo.Path,
-		Env: env,
-	})
-	if runErr != nil {
-		return "", runErr
-	}
-
-	parts := bytes.SplitSeq(bytes.TrimSpace(stdout), []byte{'\n'})
-
-	// check the commits one by one until we find a commit contained by another branch
-	// and we think this commit is the divergence point
-	for commitID := range parts {
-		branches, err := repo.getBranches(env, string(commitID), 2)
-		if err != nil {
-			return "", err
-		}
-		for _, b := range branches {
-			if b != branch {
-				return string(commitID), nil
-			}
-		}
-	}
-
-	return "", nil
-}

modules/gitrepo/fetch.go

@@ -0,0 +1,18 @@
+// Copyright 2025 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package gitrepo
+
+import (
+	"context"
+
+	"code.gitea.io/gitea/modules/git/gitcmd"
+)
+
+func FetchRemoteCommit(ctx context.Context, repo, remoteRepo Repository, commitID string) error {
+	_, _, err := gitcmd.NewCommand("fetch", "--no-tags").
+		AddDynamicArguments(repoPath(remoteRepo)).
+		AddDynamicArguments(commitID).
+		RunStdString(ctx, &gitcmd.RunOpts{Dir: repoPath(repo)})
+	return err
+}

modules/gitrepo/merge.go

@@ -0,0 +1,61 @@
+// Copyright 2025 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package gitrepo
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"strings"
+
+	"code.gitea.io/gitea/modules/git"
+	"code.gitea.io/gitea/modules/git/gitcmd"
+	"code.gitea.io/gitea/modules/log"
+)
+
+func MergeBase(ctx context.Context, repo Repository, commit1, commit2 string) (string, error) {
+	mergeBase, _, err := gitcmd.NewCommand("merge-base", "--").
+		AddDynamicArguments(commit1, commit2).
+		RunStdString(ctx, &gitcmd.RunOpts{Dir: repoPath(repo)})
+	if err != nil {
+		return "", fmt.Errorf("get merge-base of %s and %s failed: %w", commit1, commit2, err)
+	}
+	return strings.TrimSpace(mergeBase), nil
+}
+
+func MergeTree(ctx context.Context, repo Repository, base, ours, theirs string) (string, bool, []string, error) {
+	cmd := gitcmd.NewCommand("merge-tree", "--write-tree", "-z", "--name-only", "--no-messages")
+	// https://git-scm.com/docs/git-merge-tree/2.40.0#_mistakes_to_avoid
+	if git.DefaultFeatures().CheckVersionAtLeast("2.40") && !git.DefaultFeatures().CheckVersionAtLeast("2.41") {
+		cmd.AddOptionFormat("--merge-base=%s", base)
+	}
+
+	stdout := &bytes.Buffer{}
+	gitErr := cmd.AddDynamicArguments(ours, theirs).Run(ctx, &gitcmd.RunOpts{
+		Dir:    repoPath(repo),
+		Stdout: stdout,
+	})
+	if gitErr != nil && !gitcmd.IsErrorExitCode(gitErr, 1) {
+		log.Error("run merge-tree failed: %v", gitErr)
+		return "", false, nil, fmt.Errorf("run merge-tree failed: %w", gitErr)
+	}
+
+	// There are two situations that we consider for the output:
+	// 1. Clean merge and the output is <OID of toplevel tree>NUL
+	// 2. Merge conflict and the output is <OID of toplevel tree>NUL<Conflicted file info>NUL
+	treeOID, conflictedFileInfo, _ := strings.Cut(stdout.String(), "\x00")
+	if len(conflictedFileInfo) == 0 {
+		return treeOID, gitcmd.IsErrorExitCode(gitErr, 1), nil, nil
+	}
+
+	// Remove last NULL-byte from conflicted file info, then split with NULL byte as separator.
+	return treeOID, true, strings.Split(conflictedFileInfo[:len(conflictedFileInfo)-1], "\x00"), nil
+}
+
+func DiffTree(ctx context.Context, repo Repository, treeHash, mergeBase string) error {
+	return gitcmd.NewCommand("diff-tree", "--quiet").AddDynamicArguments(treeHash, mergeBase).
+		Run(ctx, &gitcmd.RunOpts{
+			Dir: repoPath(repo),
+		})
+}

routers/private/hook_pre_receive.go

@@ -237,7 +237,7 @@ func preReceiveBranch(ctx *preReceiveContext, oldCommitID, newCommitID string, r
 
 	globs := protectBranch.GetProtectedFilePatterns()
 	if len(globs) > 0 {
-		_, err := pull_service.CheckFileProtection(gitRepo, branchName, oldCommitID, newCommitID, globs, 1, ctx.env)
+		_, err := pull_service.CheckFileProtection(ctx, repo.RepoPath(), oldCommitID, newCommitID, globs, 1, ctx.env)
 		if err != nil {
 			if !pull_service.IsErrFilePathProtected(err) {
 				log.Error("Unable to check file protection for commits from %s to %s in %-v: %v", oldCommitID, newCommitID, repo, err)
@@ -295,7 +295,7 @@ func preReceiveBranch(ctx *preReceiveContext, oldCommitID, newCommitID string, r
 			// Allow commits that only touch unprotected files
 			globs := protectBranch.GetUnprotectedFilePatterns()
 			if len(globs) > 0 {
-				unprotectedFilesOnly, err := pull_service.CheckUnprotectedFiles(gitRepo, branchName, oldCommitID, newCommitID, globs, ctx.env)
+				unprotectedFilesOnly, err := pull_service.CheckUnprotectedFiles(ctx, repo, oldCommitID, newCommitID, globs, ctx.env)
 				if err != nil {
 					log.Error("Unable to check file protection for commits from %s to %s in %-v: %v", oldCommitID, newCommitID, repo, err)
 					ctx.JSON(http.StatusInternalServerError, private.Response{

services/pull/check.go

@@ -437,7 +437,7 @@ func checkPullRequestMergeable(id int64) {
 		return
 	}
 
-	if err := testPullRequestBranchMergeable(pr); err != nil {
+	if err := checkPullRequestMergeableAndUpdateStatus(ctx, pr); err != nil {
 		log.Error("testPullRequestTmpRepoBranchMergeable[%-v]: %v", pr, err)
 		pr.Status = issues_model.PullRequestStatusError
 		if err := pr.UpdateCols(ctx, "status"); err != nil {

services/pull/conflicts.go

@@ -0,0 +1,152 @@
+// Copyright 2025 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package pull
+
+import (
+	"context"
+	"errors"
+	"fmt"
+
+	issues_model "code.gitea.io/gitea/models/issues"
+	"code.gitea.io/gitea/modules/git"
+	"code.gitea.io/gitea/modules/git/gitcmd"
+	"code.gitea.io/gitea/modules/gitrepo"
+	"code.gitea.io/gitea/modules/log"
+)
+
+// checkPullRequestMergeableAndUpdateStatus checks whether a pull request is mergeable and updates its status accordingly.
+// It uses 'git merge-tree' if supported by the Git version, otherwise it falls back to using a temporary repository.
+// This function updates the pr.Status, pr.MergeBase and pr.ConflictedFiles fields as necessary.
+// The pull request parameter may not be created yet in the database, so do not assume it has an ID.
+func checkPullRequestMergeableAndUpdateStatus(ctx context.Context, pr *issues_model.PullRequest) error {
+	if git.DefaultFeatures().SupportGitMergeTree {
+		return checkPullRequestMergeableAndUpdateStatusMergeTree(ctx, pr)
+	}
+
+	return checkPullRequestMergeableAndUpdateStatusTmpRepo(ctx, pr)
+}
+
+// checkConflictsMergeTree uses git merge-tree to check for conflicts and if none are found checks if the patch is empty
+// return true if there is conflicts otherwise return false
+// pr.Status and pr.ConflictedFiles will be updated as necessary
+func checkConflictsMergeTree(ctx context.Context, pr *issues_model.PullRequest, baseCommitID string) (bool, error) {
+	treeHash, conflict, conflictFiles, err := gitrepo.MergeTree(ctx, pr.BaseRepo, pr.MergeBase, baseCommitID, pr.HeadCommitID)
+	if err != nil {
+		return false, fmt.Errorf("MergeTree: %w", err)
+	}
+	if conflict {
+		pr.Status = issues_model.PullRequestStatusConflict
+		pr.ConflictedFiles = conflictFiles
+
+		log.Trace("Found %d files conflicted: %v", len(pr.ConflictedFiles), pr.ConflictedFiles)
+		return true, nil
+	}
+
+	// No conflicts were detected, now check if the pull request actually
+	// contains anything useful via a diff. git-diff-tree(1) with --quiet
+	// will return exit code 0 if there's no diff and exit code 1 if there's
+	// a diff.
+	isEmpty := true
+	if err = gitrepo.DiffTree(ctx, pr.BaseRepo, treeHash, pr.MergeBase); err != nil {
+		if !gitcmd.IsErrorExitCode(err, 1) {
+			return false, fmt.Errorf("DiffTree: %w", err)
+		}
+		isEmpty = false
+	}
+
+	if isEmpty {
+		log.Debug("PullRequest[%d]: Patch is empty - ignoring", pr.ID)
+		pr.Status = issues_model.PullRequestStatusEmpty
+	}
+	return false, nil
+}
+
+func checkPullRequestMergeableAndUpdateStatusMergeTree(ctx context.Context, pr *issues_model.PullRequest) error {
+	// 1. Get head commit
+	if err := pr.LoadHeadRepo(ctx); err != nil {
+		return err
+	}
+	headGitRepo, err := gitrepo.OpenRepository(ctx, pr.HeadRepo)
+	if err != nil {
+		return fmt.Errorf("OpenRepository: %w", err)
+	}
+	defer headGitRepo.Close()
+
+	// 2. Get base commit id
+	var baseGitRepo *git.Repository
+	if pr.IsSameRepo() {
+		baseGitRepo = headGitRepo
+	} else {
+		baseGitRepo, err = gitrepo.OpenRepository(ctx, pr.BaseRepo)
+		if err != nil {
+			return fmt.Errorf("OpenRepository: %w", err)
+		}
+		defer baseGitRepo.Close()
+	}
+
+	// 3. Get head commit id
+	if pr.Flow == issues_model.PullRequestFlowGithub {
+		pr.HeadCommitID, err = headGitRepo.GetRefCommitID(git.BranchPrefix + pr.HeadBranch)
+		if err != nil {
+			return fmt.Errorf("GetBranchCommitID: can't find commit ID for head: %w", err)
+		}
+	} else {
+		if pr.ID > 0 {
+			pr.HeadCommitID, err = baseGitRepo.GetRefCommitID(pr.GetGitHeadRefName())
+			if err != nil {
+				return fmt.Errorf("GetRefCommitID: can't find commit ID for head: %w", err)
+			}
+		} else if pr.HeadCommitID == "" { // for new pull request with agit, the head commit id must be provided
+			return errors.New("head commit ID is empty for pull request Agit flow")
+		}
+	}
+
+	// 4. fetch head commit id into the current repository
+	// it will be checked in 2 weeks by default from git if the pull request created failure.
+	if !pr.IsSameRepo() {
+		if err := gitrepo.FetchRemoteCommit(ctx, pr.BaseRepo, pr.HeadRepo, pr.HeadCommitID); err != nil {
+			return fmt.Errorf("FetchRemoteCommit: %w", err)
+		}
+	}
+
+	// 5. update merge base
+	baseCommitID, err := baseGitRepo.GetRefCommitID(git.BranchPrefix + pr.BaseBranch)
+	if err != nil {
+		return fmt.Errorf("GetBranchCommitID: can't find commit ID for base: %w", err)
+	}
+
+	pr.MergeBase, err = gitrepo.MergeBase(ctx, pr.BaseRepo, baseCommitID, pr.HeadCommitID)
+	if err != nil {
+		log.Error("GetMergeBase: %v and can't find commit ID for base: %v", err, baseCommitID)
+		pr.Status = issues_model.PullRequestStatusEmpty // if there is no merge base, then it's empty but we still need to allow the pull request created
+		return nil
+	}
+
+	// 6. if base == head, then it's an ancestor
+	if pr.HeadCommitID == pr.MergeBase {
+		pr.Status = issues_model.PullRequestStatusAncestor
+		return nil
+	}
+
+	// 7. Check for conflicts
+	conflicted, err := checkConflictsMergeTree(ctx, pr, baseCommitID)
+	if err != nil {
+		log.Error("checkConflictsMergeTree: %v", err)
+		pr.Status = issues_model.PullRequestStatusEmpty // if there is no merge base, then it's empty but we still need to allow the pull request created
+	}
+	if conflicted || pr.Status == issues_model.PullRequestStatusEmpty {
+		return nil
+	}
+
+	// 7. Check for protected files changes
+	if err = checkPullFilesProtection(ctx, pr, pr.BaseRepo.RepoPath()); err != nil {
+		return fmt.Errorf("pr.CheckPullFilesProtection(): %v", err)
+	}
+	if len(pr.ChangedProtectedFiles) > 0 {
+		log.Trace("Found %d protected files changed", len(pr.ChangedProtectedFiles))
+	}
+
+	pr.Status = issues_model.PullRequestStatusMergeable
+	return nil
+}