Merge pull request #5738 from bk2204/log-scanner-stderr

Avoid deadlocking on log scanning with lots of output on stderr

git/git.go

@@ -212,6 +212,10 @@ func gitNoLFSBuffered(args ...string) (*subprocess.BufferedCmd, error) {
 	return subprocess.BufferedExec("git", gitConfigNoLFS(args...)...)
 }
 
+func gitNoLFSBufferedStdout(args ...string) (*subprocess.BufferedCmd, error) {
+	return subprocess.StdoutBufferedExec("git", gitConfigNoLFS(args...)...)
+}
+
 // Invoke Git with enabled LFS filters
 func git(args ...string) (*subprocess.Cmd, error) {
 	return subprocess.ExecCommand("git", args...)
@@ -225,6 +229,10 @@ func gitBuffered(args ...string) (*subprocess.BufferedCmd, error) {
 	return subprocess.BufferedExec("git", args...)
 }
 
+func gitBufferedStdout(args ...string) (*subprocess.BufferedCmd, error) {
+	return subprocess.StdoutBufferedExec("git", args...)
+}
+
 func CatFile() (*subprocess.BufferedCmd, error) {
 	return gitNoLFSBuffered("cat-file", "--batch-check")
 }
@@ -247,7 +255,7 @@ func DiffIndex(ref string, cached bool, refresh bool, workingDir string) (*bufio
 		args = append([]string{"-C", workingDir}, args...)
 	}
 
-	cmd, err := gitBuffered(args...)
+	cmd, err := gitBufferedStdout(args...)
 	if err != nil {
 		return nil, err
 	}

lfs/gitscanner_log.go

@@ -137,6 +137,14 @@ func scanStashed(cb GitScannerFoundPointer) error {
 func parseScannerLogOutput(cb GitScannerFoundPointer, direction LogDiffDirection, cmd *subprocess.BufferedCmd, filter *filepathfilter.Filter) {
 	ch := make(chan gitscannerResult, chanBufSize)
 
+	cherr := make(chan []byte)
+
+	go func() {
+		stderr, _ := io.ReadAll(cmd.Stderr)
+		cherr <- stderr
+		close(cherr)
+	}()
+
 	go func() {
 		scanner := newLogScanner(direction, cmd.Stdout)
 		scanner.Filter = filter
@@ -149,7 +157,7 @@ func parseScannerLogOutput(cb GitScannerFoundPointer, direction LogDiffDirection
 			io.ReadAll(cmd.Stdout)
 			ch <- gitscannerResult{Err: errors.New(tr.Tr.Get("error while scanning `git log`: %v", err))}
 		}
-		stderr, _ := io.ReadAll(cmd.Stderr)
+		stderr := <-cherr
 		err := cmd.Wait()
 		if err != nil {
 			ch <- gitscannerResult{Err: errors.New(tr.Tr.Get("error in `git log`: %v %v", err, string(stderr)))}

subprocess/subprocess.go

@@ -50,6 +50,36 @@ func BufferedExec(name string, args ...string) (*BufferedCmd, error) {
 	}, nil
 }
 
+// StdoutBufferedExec starts up a command and creates a stdin pipe and a
+// buffered stdout pipe, with stderr directed to /dev/null, wrapped in a
+// BufferedCmd. The stdout buffer will be of stdoutBufSize bytes.
+func StdoutBufferedExec(name string, args ...string) (*BufferedCmd, error) {
+	cmd, err := ExecCommand(name, args...)
+	if err != nil {
+		return nil, err
+	}
+	stdout, err := cmd.StdoutPipe()
+	if err != nil {
+		return nil, err
+	}
+
+	stdin, err := cmd.StdinPipe()
+	if err != nil {
+		return nil, err
+	}
+
+	if err := cmd.Start(); err != nil {
+		return nil, err
+	}
+
+	return &BufferedCmd{
+		cmd,
+		stdin,
+		bufio.NewReaderSize(stdout, stdoutBufSize),
+		nil,
+	}, nil
+}
+
 // SimpleExec is a small wrapper around os/exec.Command.
 func SimpleExec(name string, args ...string) (string, error) {
 	cmd, err := ExecCommand(name, args...)