From 74bb292fe3f4c02fc1dc5f32622c74d820cadd78 Mon Sep 17 00:00:00 2001
From: zeripath <art27@cantab.net>
Date: Mon, 11 Nov 2019 11:46:28 +0000
Subject: [PATCH] Migrate temp_repo.go to use git.NewCommand  (#8918)

This PR migrates temp_repo.go to use git.NewCommand instead creating processes by itself - this fixes the problem underlying PR #8905.

There are other places that run git outside of the controlled locale defined in #8548 but temp_repo.go is the only cause of failure of local testing in cases where English is not the default - implying that error messages from those other commands are not interpreted.

Replaces #8905
---
 modules/git/command.go         |  11 ++
 modules/repofiles/temp_repo.go | 241 +++++++++++----------------------
 2 files changed, 91 insertions(+), 161 deletions(-)

diff --git a/modules/git/command.go b/modules/git/command.go
index 2b5288aea..7772abd2d 100644
--- a/modules/git/command.go
+++ b/modules/git/command.go
@@ -67,6 +67,13 @@ func (c *Command) RunInDirTimeoutEnvPipeline(env []string, timeout time.Duration
 // RunInDirTimeoutEnvFullPipeline executes the command in given directory with given timeout,
 // it pipes stdout and stderr to given io.Writer and passes in an io.Reader as stdin.
 func (c *Command) RunInDirTimeoutEnvFullPipeline(env []string, timeout time.Duration, dir string, stdout, stderr io.Writer, stdin io.Reader) error {
+	return c.RunInDirTimeoutEnvFullPipelineFunc(env, timeout, dir, stdout, stderr, stdin, nil)
+}
+
+// RunInDirTimeoutEnvFullPipelineFunc executes the command in given directory with given timeout,
+// it pipes stdout and stderr to given io.Writer and passes in an io.Reader as stdin. Between cmd.Start and cmd.Wait the passed in function is run.
+func (c *Command) RunInDirTimeoutEnvFullPipelineFunc(env []string, timeout time.Duration, dir string, stdout, stderr io.Writer, stdin io.Reader, fn func(context.Context, context.CancelFunc)) error {
+
 	if timeout == -1 {
 		timeout = DefaultCommandExecutionTimeout
 	}
@@ -98,6 +105,10 @@ func (c *Command) RunInDirTimeoutEnvFullPipeline(env []string, timeout time.Dura
 	pid := process.GetManager().Add(fmt.Sprintf("%s %s %s [repo_path: %s]", GitExecutable, c.name, strings.Join(c.args, " "), dir), cmd)
 	defer process.GetManager().Remove(pid)
 
+	if fn != nil {
+		fn(ctx, cancel)
+	}
+
 	if err := cmd.Wait(); err != nil {
 		return err
 	}
diff --git a/modules/repofiles/temp_repo.go b/modules/repofiles/temp_repo.go
index b07d2a897..abc224c2c 100644
--- a/modules/repofiles/temp_repo.go
+++ b/modules/repofiles/temp_repo.go
@@ -10,7 +10,6 @@ import (
 	"fmt"
 	"io"
 	"os"
-	"os/exec"
 	"regexp"
 	"strings"
 	"time"
@@ -18,7 +17,6 @@ import (
 	"code.gitea.io/gitea/models"
 	"code.gitea.io/gitea/modules/git"
 	"code.gitea.io/gitea/modules/log"
-	"code.gitea.io/gitea/modules/process"
 	"code.gitea.io/gitea/modules/setting"
 	"code.gitea.io/gitea/services/gitdiff"
 
@@ -51,9 +49,8 @@ func (t *TemporaryUploadRepository) Close() {
 
 // Clone the base repository to our path and set branch as the HEAD
 func (t *TemporaryUploadRepository) Clone(branch string) error {
-	if _, stderr, err := process.GetManager().ExecTimeout(5*time.Minute,
-		fmt.Sprintf("Clone (git clone -s --bare): %s", t.basePath),
-		git.GitExecutable, "clone", "-s", "--bare", "-b", branch, t.repo.RepoPath(), t.basePath); err != nil {
+	if _, err := git.NewCommand("clone", "-s", "--bare", "-b", branch, t.repo.RepoPath(), t.basePath).Run(); err != nil {
+		stderr := err.Error()
 		if matched, _ := regexp.MatchString(".*Remote branch .* not found in upstream origin.*", stderr); matched {
 			return git.ErrBranchNotExist{
 				Name: branch,
@@ -79,11 +76,8 @@ func (t *TemporaryUploadRepository) Clone(branch string) error {
 
 // SetDefaultIndex sets the git index to our HEAD
 func (t *TemporaryUploadRepository) SetDefaultIndex() error {
-	if _, stderr, err := process.GetManager().ExecDir(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("SetDefaultIndex (git read-tree HEAD): %s", t.basePath),
-		git.GitExecutable, "read-tree", "HEAD"); err != nil {
-		return fmt.Errorf("SetDefaultIndex: %v %s", err, stderr)
+	if _, err := git.NewCommand("read-tree", "HEAD").RunInDir(t.basePath); err != nil {
+		return fmt.Errorf("SetDefaultIndex: %v", err)
 	}
 	return nil
 }
@@ -93,10 +87,6 @@ func (t *TemporaryUploadRepository) LsFiles(filenames ...string) ([]string, erro
 	stdOut := new(bytes.Buffer)
 	stdErr := new(bytes.Buffer)
 
-	timeout := 5 * time.Minute
-	ctx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
-
 	cmdArgs := []string{"ls-files", "-z", "--"}
 	for _, arg := range filenames {
 		if arg != "" {
@@ -104,22 +94,9 @@ func (t *TemporaryUploadRepository) LsFiles(filenames ...string) ([]string, erro
 		}
 	}
 
-	cmd := exec.CommandContext(ctx, git.GitExecutable, cmdArgs...)
-	desc := fmt.Sprintf("lsFiles: (git ls-files) %v", cmdArgs)
-	cmd.Dir = t.basePath
-	cmd.Stdout = stdOut
-	cmd.Stderr = stdErr
-
-	if err := cmd.Start(); err != nil {
-		return nil, fmt.Errorf("exec(%s) failed: %v(%v)", desc, err, ctx.Err())
-	}
-
-	pid := process.GetManager().Add(desc, cmd)
-	err := cmd.Wait()
-	process.GetManager().Remove(pid)
-
-	if err != nil {
-		err = fmt.Errorf("exec(%d:%s) failed: %v(%v) stdout: %v stderr: %v", pid, desc, err, ctx.Err(), stdOut, stdErr)
+	if err := git.NewCommand(cmdArgs...).RunInDirPipeline(t.basePath, stdOut, stdErr); err != nil {
+		log.Error("Unable to run git ls-files for temporary repo: %s (%s) Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), t.basePath, err, stdOut.String(), stdErr.String())
+		err = fmt.Errorf("Unable to run git ls-files for temporary repo of: %s Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), err, stdOut.String(), stdErr.String())
 		return nil, err
 	}
 
@@ -128,7 +105,7 @@ func (t *TemporaryUploadRepository) LsFiles(filenames ...string) ([]string, erro
 		filelist = append(filelist, string(line))
 	}
 
-	return filelist, err
+	return filelist, nil
 }
 
 // RemoveFilesFromIndex removes the given files from the index
@@ -144,90 +121,50 @@ func (t *TemporaryUploadRepository) RemoveFilesFromIndex(filenames ...string) er
 		}
 	}
 
-	timeout := 5 * time.Minute
-	ctx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
-
-	cmdArgs := []string{"update-index", "--remove", "-z", "--index-info"}
-	cmd := exec.CommandContext(ctx, git.GitExecutable, cmdArgs...)
-	desc := fmt.Sprintf("removeFilesFromIndex: (git update-index) %v", filenames)
-	cmd.Dir = t.basePath
-	cmd.Stdout = stdOut
-	cmd.Stderr = stdErr
-	cmd.Stdin = bytes.NewReader(stdIn.Bytes())
-
-	if err := cmd.Start(); err != nil {
-		return fmt.Errorf("exec(%s) failed: %v(%v)", desc, err, ctx.Err())
+	if err := git.NewCommand("update-index", "--remove", "-z", "--index-info").RunInDirFullPipeline(t.basePath, stdOut, stdErr, stdIn); err != nil {
+		log.Error("Unable to update-index for temporary repo: %s (%s) Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), t.basePath, err, stdOut.String(), stdErr.String())
+		return fmt.Errorf("Unable to update-index for temporary repo: %s Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), err, stdOut.String(), stdErr.String())
 	}
-
-	pid := process.GetManager().Add(desc, cmd)
-	err := cmd.Wait()
-	process.GetManager().Remove(pid)
-
-	if err != nil {
-		err = fmt.Errorf("exec(%d:%s) failed: %v(%v) stdout: %v stderr: %v", pid, desc, err, ctx.Err(), stdOut, stdErr)
-	}
-
-	return err
+	return nil
 }
 
 // HashObject writes the provided content to the object db and returns its hash
 func (t *TemporaryUploadRepository) HashObject(content io.Reader) (string, error) {
-	timeout := 5 * time.Minute
-	ctx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
+	stdOut := new(bytes.Buffer)
+	stdErr := new(bytes.Buffer)
 
-	hashCmd := exec.CommandContext(ctx, git.GitExecutable, "hash-object", "-w", "--stdin")
-	hashCmd.Dir = t.basePath
-	hashCmd.Stdin = content
-	stdOutBuffer := new(bytes.Buffer)
-	stdErrBuffer := new(bytes.Buffer)
-	hashCmd.Stdout = stdOutBuffer
-	hashCmd.Stderr = stdErrBuffer
-	desc := fmt.Sprintf("hashObject: (git hash-object)")
-	if err := hashCmd.Start(); err != nil {
-		return "", fmt.Errorf("git hash-object: %s", err)
+	if err := git.NewCommand("hash-object", "-w", "--stdin").RunInDirFullPipeline(t.basePath, stdOut, stdErr, content); err != nil {
+		log.Error("Unable to hash-object to temporary repo: %s (%s) Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), t.basePath, err, stdOut.String(), stdErr.String())
+		return "", fmt.Errorf("Unable to hash-object to temporary repo: %s Error: %v\nstdout: %s\nstderr: %s", t.repo.FullName(), err, stdOut.String(), stdErr.String())
 	}
 
-	pid := process.GetManager().Add(desc, hashCmd)
-	err := hashCmd.Wait()
-	process.GetManager().Remove(pid)
-
-	if err != nil {
-		err = fmt.Errorf("exec(%d:%s) failed: %v(%v) stdout: %v stderr: %v", pid, desc, err, ctx.Err(), stdOutBuffer, stdErrBuffer)
-		return "", err
-	}
-
-	return strings.TrimSpace(stdOutBuffer.String()), nil
+	return strings.TrimSpace(stdOut.String()), nil
 }
 
 // AddObjectToIndex adds the provided object hash to the index with the provided mode and path
 func (t *TemporaryUploadRepository) AddObjectToIndex(mode, objectHash, objectPath string) error {
-	if _, stderr, err := process.GetManager().ExecDir(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("addObjectToIndex (git update-index): %s", t.basePath),
-		git.GitExecutable, "update-index", "--add", "--replace", "--cacheinfo", mode, objectHash, objectPath); err != nil {
+	if _, err := git.NewCommand("update-index", "--add", "--replace", "--cacheinfo", mode, objectHash, objectPath).RunInDir(t.basePath); err != nil {
+		stderr := err.Error()
 		if matched, _ := regexp.MatchString(".*Invalid path '.*", stderr); matched {
 			return models.ErrFilePathInvalid{
 				Message: objectPath,
 				Path:    objectPath,
 			}
 		}
-		return fmt.Errorf("git update-index: %s", stderr)
+		log.Error("Unable to add object to index: %s %s %s in temporary repo %s(%s) Error: %v", mode, objectHash, objectPath, t.repo.FullName(), t.basePath, err)
+		return fmt.Errorf("Unable to add object to index at %s in temporary repo %s Error: %v", objectPath, t.repo.FullName(), err)
 	}
 	return nil
 }
 
 // WriteTree writes the current index as a tree to the object db and returns its hash
 func (t *TemporaryUploadRepository) WriteTree() (string, error) {
-	treeHash, stderr, err := process.GetManager().ExecDir(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("WriteTree (git write-tree): %s", t.basePath),
-		git.GitExecutable, "write-tree")
+	stdout, err := git.NewCommand("write-tree").RunInDir(t.basePath)
 	if err != nil {
-		return "", fmt.Errorf("git write-tree: %s", stderr)
+		log.Error("Unable to write tree in temporary repo: %s(%s): Error: %v", t.repo.FullName(), t.basePath, err)
+		return "", fmt.Errorf("Unable to write-tree in temporary repo for: %s Error: %v", t.repo.FullName(), err)
 	}
-	return strings.TrimSpace(treeHash), nil
+	return strings.TrimSpace(stdout), nil
 }
 
 // GetLastCommit gets the last commit ID SHA of the repo
@@ -240,14 +177,12 @@ func (t *TemporaryUploadRepository) GetLastCommitByRef(ref string) (string, erro
 	if ref == "" {
 		ref = "HEAD"
 	}
-	treeHash, stderr, err := process.GetManager().ExecDir(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("GetLastCommit (git rev-parse %s): %s", ref, t.basePath),
-		git.GitExecutable, "rev-parse", ref)
+	stdout, err := git.NewCommand("rev-parse", ref).RunInDir(t.basePath)
 	if err != nil {
-		return "", fmt.Errorf("git rev-parse %s: %s", ref, stderr)
+		log.Error("Unable to get last ref for %s in temporary repo: %s(%s): Error: %v", ref, t.repo.FullName(), t.basePath, err)
+		return "", fmt.Errorf("Unable to rev-parse %s in temporary repo for: %s Error: %v", ref, t.repo.FullName(), err)
 	}
-	return strings.TrimSpace(treeHash), nil
+	return strings.TrimSpace(stdout), nil
 }
 
 // CommitTree creates a commit from a given tree for the user with provided message
@@ -287,16 +222,15 @@ func (t *TemporaryUploadRepository) CommitTree(author, committer *models.User, t
 		}
 	}
 
-	commitHash, stderr, err := process.GetManager().ExecDirEnvStdIn(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("commitTree (git commit-tree): %s", t.basePath),
-		env,
-		messageBytes,
-		git.GitExecutable, args...)
-	if err != nil {
-		return "", fmt.Errorf("git commit-tree: %s", stderr)
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+	if err := git.NewCommand(args...).RunInDirTimeoutEnvFullPipeline(env, -1, t.basePath, stdout, stderr, messageBytes); err != nil {
+		log.Error("Unable to commit-tree in temporary repo: %s (%s) Error: %v\nStdout: %s\nStderr: %s",
+			t.repo.FullName(), t.basePath, err, stdout, stderr)
+		return "", fmt.Errorf("Unable to commit-tree in temporary repo: %s Error: %v\nStdout: %s\nStderr: %s",
+			t.repo.FullName(), err, stdout, stderr)
 	}
-	return strings.TrimSpace(commitHash), nil
+	return strings.TrimSpace(stdout.String()), nil
 }
 
 // Push the provided commitHash to the repository branch by the provided user
@@ -304,47 +238,48 @@ func (t *TemporaryUploadRepository) Push(doer *models.User, commitHash string, b
 	// Because calls hooks we need to pass in the environment
 	env := models.PushingEnvironment(doer, t.repo)
 
-	if _, stderr, err := process.GetManager().ExecDirEnv(5*time.Minute,
-		t.basePath,
-		fmt.Sprintf("actuallyPush (git push): %s", t.basePath),
-		env,
-		git.GitExecutable, "push", t.repo.RepoPath(), strings.TrimSpace(commitHash)+":refs/heads/"+strings.TrimSpace(branch)); err != nil {
-		return fmt.Errorf("git push: %s", stderr)
+	if _, err := git.NewCommand("push", t.repo.RepoPath(), strings.TrimSpace(commitHash)+":refs/heads/"+strings.TrimSpace(branch)).RunInDirWithEnv(t.basePath, env); err != nil {
+		log.Error("Unable to push back to repo from temporary repo: %s (%s) Error: %v",
+			t.repo.FullName(), t.basePath, err)
+		return fmt.Errorf("Unable to push back to repo from temporary repo: %s (%s) Error: %v",
+			t.repo.FullName(), t.basePath, err)
 	}
 	return nil
 }
 
 // DiffIndex returns a Diff of the current index to the head
-func (t *TemporaryUploadRepository) DiffIndex() (diff *gitdiff.Diff, err error) {
-	timeout := 5 * time.Minute
-	ctx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
-
-	stdErr := new(bytes.Buffer)
-
-	cmd := exec.CommandContext(ctx, git.GitExecutable, "diff-index", "--cached", "-p", "HEAD")
-	cmd.Dir = t.basePath
-	cmd.Stderr = stdErr
-
-	stdout, err := cmd.StdoutPipe()
+func (t *TemporaryUploadRepository) DiffIndex() (*gitdiff.Diff, error) {
+	stdoutReader, stdoutWriter, err := os.Pipe()
 	if err != nil {
-		return nil, fmt.Errorf("StdoutPipe: %v stderr %s", err, stdErr.String())
+		log.Error("Unable to open stdout pipe: %v", err)
+		return nil, fmt.Errorf("Unable to open stdout pipe: %v", err)
 	}
+	defer func() {
+		_ = stdoutReader.Close()
+		_ = stdoutWriter.Close()
+	}()
+	stderr := new(bytes.Buffer)
+	var diff *gitdiff.Diff
+	var finalErr error
 
-	if err = cmd.Start(); err != nil {
-		return nil, fmt.Errorf("Start: %v stderr %s", err, stdErr.String())
-	}
-
-	pid := process.GetManager().Add(fmt.Sprintf("diffIndex [repo_path: %s]", t.repo.RepoPath()), cmd)
-	defer process.GetManager().Remove(pid)
-
-	diff, err = gitdiff.ParsePatch(setting.Git.MaxGitDiffLines, setting.Git.MaxGitDiffLineCharacters, setting.Git.MaxGitDiffFiles, stdout)
-	if err != nil {
-		return nil, fmt.Errorf("ParsePatch: %v", err)
-	}
-
-	if err = cmd.Wait(); err != nil {
-		return nil, fmt.Errorf("Wait: %v", err)
+	if err := git.NewCommand("diff-index", "--cached", "-p", "HEAD").
+		RunInDirTimeoutEnvFullPipelineFunc(nil, 30*time.Second, t.basePath, stdoutWriter, stderr, nil, func(ctx context.Context, cancel context.CancelFunc) {
+			_ = stdoutWriter.Close()
+			diff, finalErr = gitdiff.ParsePatch(setting.Git.MaxGitDiffLines, setting.Git.MaxGitDiffLineCharacters, setting.Git.MaxGitDiffFiles, stdoutReader)
+			if finalErr != nil {
+				log.Error("ParsePatch: %v", finalErr)
+				cancel()
+			}
+			_ = stdoutReader.Close()
+		}); err != nil {
+		if finalErr != nil {
+			log.Error("Unable to ParsePatch in temporary repo %s (%s). Error: %v", t.repo.FullName(), t.basePath, finalErr)
+			return nil, finalErr
+		}
+		log.Error("Unable to run diff-index pipeline in temporary repo %s (%s). Error: %v\nStderr: %s",
+			t.repo.FullName(), t.basePath, err, stderr)
+		return nil, fmt.Errorf("Unable to run diff-index pipeline in temporary repo %s. Error: %v\nStderr: %s",
+			t.repo.FullName(), err, stderr)
 	}
 
 	return diff, nil
@@ -358,12 +293,8 @@ func (t *TemporaryUploadRepository) CheckAttribute(attribute string, args ...str
 		return nil, err
 	}
 
-	stdOut := new(bytes.Buffer)
-	stdErr := new(bytes.Buffer)
-
-	timeout := 5 * time.Minute
-	ctx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
 
 	cmdArgs := []string{"check-attr", "-z", attribute}
 
@@ -379,26 +310,14 @@ func (t *TemporaryUploadRepository) CheckAttribute(attribute string, args ...str
 		}
 	}
 
-	cmd := exec.CommandContext(ctx, git.GitExecutable, cmdArgs...)
-	desc := fmt.Sprintf("checkAttr: (git check-attr) %s %v", attribute, cmdArgs)
-	cmd.Dir = t.basePath
-	cmd.Stdout = stdOut
-	cmd.Stderr = stdErr
-
-	if err := cmd.Start(); err != nil {
-		return nil, fmt.Errorf("exec(%s) failed: %v(%v)", desc, err, ctx.Err())
+	if err := git.NewCommand(cmdArgs...).RunInDirPipeline(t.basePath, stdout, stderr); err != nil {
+		log.Error("Unable to check-attr in temporary repo: %s (%s) Error: %v\nStdout: %s\nStderr: %s",
+			t.repo.FullName(), t.basePath, err, stdout, stderr)
+		return nil, fmt.Errorf("Unable to check-attr in temporary repo: %s Error: %v\nStdout: %s\nStderr: %s",
+			t.repo.FullName(), err, stdout, stderr)
 	}
 
-	pid := process.GetManager().Add(desc, cmd)
-	err = cmd.Wait()
-	process.GetManager().Remove(pid)
-
-	if err != nil {
-		err = fmt.Errorf("exec(%d:%s) failed: %v(%v) stdout: %v stderr: %v", pid, desc, err, ctx.Err(), stdOut, stdErr)
-		return nil, err
-	}
-
-	fields := bytes.Split(stdOut.Bytes(), []byte{'\000'})
+	fields := bytes.Split(stdout.Bytes(), []byte{'\000'})
 
 	if len(fields)%3 != 1 {
 		return nil, fmt.Errorf("Wrong number of fields in return from check-attr")