refactor: ParseCatFileTreeLine returns slices from ReadBytes, remove pre-provided buffers

Replace the three caller-provided buffer parameters (modeBuf, fnameBuf, shaBuf)
with internal allocation. Use rd.ReadBytes('\x00') to read mode+fname in one
call and slice the result directly. Allocate sha internally. Update both callers
(catBatchParseTreeEntries and lfs_nogogit.go) to drop the buffer setup.

Co-Authored-By: GitHub Copilot <copilot@github.com>

Agent-Logs-Url: https://github.com/go-gitea/gitea/sessions/43dab9a5-8c12-41c7-8da6-77c268a65ada

Co-authored-by: wxiaoguang <2114189+wxiaoguang@users.noreply.github.com>

modules/git/catfile_batch_reader.go

@@ -179,18 +179,14 @@ headerLoop:
 // Therefore we need some method to convert these binary hashes to hex hashes
 
 // ParseCatFileTreeLine reads an entry from a tree in a cat-file --batch stream
-// This carefully avoids allocations - except where fnameBuf is too small.
-// It is recommended therefore to pass in an fnameBuf large enough to avoid almost all allocations
 //
 // Each line is composed of:
 // <mode-in-ascii-dropping-initial-zeros> SP <fname> NUL <binary HASH>
 //
 // We don't attempt to convert the raw HASH to save a lot of time
-func ParseCatFileTreeLine(objectFormat ObjectFormat, rd BufferedReader, modeBuf, fnameBuf, shaBuf []byte) (mode, fname, sha []byte, n int, err error) {
-	var readBytes []byte
-
-	// Read the Mode & fname
-	readBytes, err = rd.ReadSlice('\x00')
+func ParseCatFileTreeLine(objectFormat ObjectFormat, rd BufferedReader) (mode, fname, sha []byte, n int, err error) {
+	// Read the mode and fname up to and including the NUL separator
+	readBytes, err := rd.ReadBytes('\x00')
 	if err != nil {
 		return mode, fname, sha, n, err
 	}
@@ -200,48 +196,23 @@ func ParseCatFileTreeLine(objectFormat ObjectFormat, rd BufferedReader, modeBuf,
 		return mode, fname, sha, n, &ErrNotExist{}
 	}
 
-	n += idx + 1
-	copy(modeBuf, readBytes[:idx])
-	if len(modeBuf) >= idx {
-		modeBuf = modeBuf[:idx]
-	} else {
-		modeBuf = append(modeBuf, readBytes[len(modeBuf):idx]...)
-	}
-	mode = modeBuf
+	mode = readBytes[:idx]
+	fname = readBytes[idx+1 : len(readBytes)-1] // trim the NUL terminator
+	n = len(readBytes)
 
-	readBytes = readBytes[idx+1:]
-
-	// Deal with the fname
-	copy(fnameBuf, readBytes)
-	if len(fnameBuf) > len(readBytes) {
-		fnameBuf = fnameBuf[:len(readBytes)]
-	} else {
-		fnameBuf = append(fnameBuf, readBytes[len(fnameBuf):]...)
-	}
-	for err == bufio.ErrBufferFull {
-		readBytes, err = rd.ReadSlice('\x00')
-		fnameBuf = append(fnameBuf, readBytes...)
-	}
-	n += len(fnameBuf)
-	if err != nil {
-		return mode, fname, sha, n, err
-	}
-	fnameBuf = fnameBuf[:len(fnameBuf)-1]
-	fname = fnameBuf
-
-	// Deal with the binary hash
-	idx = 0
+	// Read the binary hash
 	length := objectFormat.FullLength() / 2
+	sha = make([]byte, length)
+	idx = 0
 	for idx < length {
 		var read int
-		read, err = rd.Read(shaBuf[idx:length])
+		read, err = rd.Read(sha[idx:length])
 		n += read
 		if err != nil {
 			return mode, fname, sha, n, err
 		}
 		idx += read
 	}
-	sha = shaBuf
 	return mode, fname, sha, n, err
 }
 

modules/git/parse_treeentry.go

@@ -47,14 +47,11 @@ func parseTreeEntries(data []byte, ptree *Tree) ([]*TreeEntry, error) {
 }
 
 func catBatchParseTreeEntries(objectFormat ObjectFormat, ptree *Tree, rd BufferedReader, sz int64) ([]*TreeEntry, error) {
-	fnameBuf := make([]byte, 4096)
-	modeBuf := make([]byte, 40)
-	shaBuf := make([]byte, objectFormat.FullLength())
 	entries := make([]*TreeEntry, 0, 10)
 
 loop:
 	for sz > 0 {
-		mode, fname, sha, count, err := ParseCatFileTreeLine(objectFormat, rd, modeBuf, fnameBuf, shaBuf)
+		mode, fname, sha, count, err := ParseCatFileTreeLine(objectFormat, rd)
 		if err != nil {
 			if err == io.EOF {
 				break loop

modules/git/pipeline/lfs_nogogit.go

@@ -46,10 +46,6 @@ func findLFSFileFunc(repo *git.Repository, objectID git.ObjectID, revListReader
 	trees := []string{}
 	paths := []string{}
 
-	fnameBuf := make([]byte, 4096)
-	modeBuf := make([]byte, 40)
-	workingShaBuf := make([]byte, objectID.Type().FullLength()/2)
-
 	for scan.Scan() {
 		// Get the next commit ID
 		commitID := scan.Text()
@@ -93,7 +89,7 @@ func findLFSFileFunc(repo *git.Repository, objectID git.ObjectID, revListReader
 			case "tree":
 				var n int64
 				for n < info.Size {
-					mode, fname, binObjectID, count, err := git.ParseCatFileTreeLine(objectID.Type(), batchReader, modeBuf, fnameBuf, workingShaBuf)
+					mode, fname, binObjectID, count, err := git.ParseCatFileTreeLine(objectID.Type(), batchReader)
 					if err != nil {
 						return nil, err
 					}