fix(gitdiff): render exact-limit diffs and zero-limit comments (#38838)

This commit fixes the treatment of various edge cases related to diff
limits.

- Exact-limit diffs: a file containing exactly the configured maximum
was incorrectly treated as too large and hidden behind “Load diff.” It
now renders normally.

Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>

services/gitdiff/gitdiff.go

@@ -1010,15 +1010,17 @@ func newDiffSectionForDiffFile(curFile *DiffFile) *DiffSection {
 func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharacters int, input *bufio.Reader) (lineBytes []byte, isFragment bool, err error) {
 	sb := strings.Builder{}
 
-	var (
-		curSection        *DiffSection
-		curFileLinesCount int
-		curFileLFSPrefix  bool
-	)
+	var curSection *DiffSection
+	curFileLFSPrefix := false
 
 	lastLeftIdx := -1
 	leftLine, rightLine := 1, 1
 
+	curFileLinesCount := 0
+	curFileLineReachesLimit := func() bool {
+		return maxLines > -1 && curFileLinesCount >= maxLines
+	}
+
 	for {
 		for isFragment {
 			curFile.IsIncomplete = true
@@ -1045,7 +1047,7 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 
 		switch lineBytes[0] {
 		case '@':
-			if maxLines > -1 && curFileLinesCount >= maxLines {
+			if curFileLineReachesLimit() {
 				curFile.IsIncomplete = true
 				continue
 			}
@@ -1068,8 +1070,8 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 			lastLeftIdx = -1
 			curFile.Sections = append(curFile.Sections, curSection)
 
-			// FIXME: the "-1" can't be right, these "line idx" are all 1-based, maybe there are other bugs that covers this bug.
-			lineSectionInfo := newDiffLineSectionInfo(curFile, line, leftLine-1, rightLine-1)
+			// use "idx-1" as "last idx" (the last line before this hunk)
+			lineSectionInfo := newDiffLineSectionInfo(curFile, line, leftLine-1 /*lastLeftIdx*/, rightLine-1 /*lastRightIdx*/)
 			diffLine := &DiffLine{
 				Type:        DiffLineSection,
 				Content:     line,
@@ -1082,10 +1084,6 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 			rightLine = lineSectionInfo.RightIdx
 			continue
 		case '\\':
-			if maxLines > -1 && curFileLinesCount >= maxLines {
-				curFile.IsIncomplete = true
-				continue
-			}
 			// This is used only to indicate that the current file does not have a terminal newline
 			if !bytes.Equal(lineBytes, []byte("\\ No newline at end of file")) {
 				return nil, false, fmt.Errorf("unexpected line in hunk: %s", string(lineBytes))
@@ -1094,12 +1092,13 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 			// FIXME: we should be putting a marker at the end of the file if there is no terminal new line
 			continue
 		case '+':
-			curFileLinesCount++
 			curFile.Addition++
-			if maxLines > -1 && curFileLinesCount >= maxLines {
+			if curFileLineReachesLimit() {
 				curFile.IsIncomplete = true
 				continue
 			}
+			curFileLinesCount++
+
 			diffLine := &DiffLine{Type: DiffLineAdd, RightIdx: rightLine, Match: -1}
 			rightLine++
 			if curSection == nil {
@@ -1125,12 +1124,13 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 				}
 			}
 		case '-':
-			curFileLinesCount++
 			curFile.Deletion++
-			if maxLines > -1 && curFileLinesCount >= maxLines {
+			if curFileLineReachesLimit() {
 				curFile.IsIncomplete = true
 				continue
 			}
+			curFileLinesCount++
+
 			diffLine := &DiffLine{Type: DiffLineDel, LeftIdx: leftLine, Match: -1}
 			if leftLine > 0 {
 				leftLine++
@@ -1153,11 +1153,11 @@ func parseHunks(ctx context.Context, curFile *DiffFile, maxLines, maxLineCharact
 				}
 			}
 		case ' ':
-			curFileLinesCount++
-			if maxLines > -1 && curFileLinesCount >= maxLines {
+			if curFileLineReachesLimit() {
 				curFile.IsIncomplete = true
 				continue
 			}
+			curFileLinesCount++
 			diffLine := &DiffLine{Type: DiffLinePlain, LeftIdx: leftLine, RightIdx: rightLine}
 			leftLine++
 			rightLine++

services/gitdiff/gitdiff_test.go

@@ -19,6 +19,7 @@ import (
 	"gitea.dev/modules/json"
 	"gitea.dev/modules/setting"
 	"gitea.dev/modules/translation"
+	"gitea.dev/modules/util"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -548,6 +549,43 @@ index 0000000..6bb8f39
 	}
 }
 
+func TestParsePatchExactLineLimit(t *testing.T) {
+	for _, test := range []struct {
+		name, hunk   string
+		limit, lines int
+		incomplete   bool
+	}{
+		{name: "zero", limit: 0, hunk: "@@ -1,3 +1,3 @@\n one\n two\n three\n", incomplete: true},
+		{name: "one", limit: 1, lines: 1, hunk: "@@ -1,3 +1,3 @@\n one\n two\n three\n", incomplete: true},
+		{name: "N plus one", limit: 2, lines: 2, hunk: "@@ -1,3 +1,3 @@\n one\n two\n three\n", incomplete: true},
+		{name: "N", limit: 3, lines: 3, hunk: "@@ -1,3 +1,3 @@\n one\n two\n three\n"},
+		{name: "addition", limit: 1, lines: 1, hunk: "@@ -0,0 +1 @@\n+one\n"},
+		{name: "deletion", limit: 1, lines: 1, hunk: "@@ -1 +0,0 @@\n-one\n"},
+		{name: "marker has no cost", limit: 1, lines: 1, hunk: "@@ -1 +1 @@\n line\n\\ No newline at end of file\n"},
+		{name: "hunk at capacity", limit: 1, lines: 1, hunk: "@@ -1 +1 @@\n one\n@@ -3 +3 @@\n three\n", incomplete: true},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			patch := "diff --git a/file b/file\n--- a/file\n+++ b/file\n" + test.hunk
+			diff, err := ParsePatch(t.Context(), test.limit, 5000, 10, strings.NewReader(patch), "")
+			require.NoError(t, err)
+			require.Len(t, diff.Files, 1)
+			diffFile := diff.Files[0]
+			if test.limit == 0 {
+				require.Len(t, diffFile.Sections, 0)
+			} else {
+				require.Len(t, diffFile.Sections, 1)
+				diffSection := diffFile.Sections[0]
+				lineSecCount := 0
+				for _, line := range diffSection.Lines {
+					lineSecCount += util.Iif(line.Type == DiffLineSection, 1, 0)
+				}
+				assert.Equal(t, test.lines, len(diffSection.Lines)-lineSecCount) // actual diff lines
+				assert.Equal(t, test.incomplete, diffFile.IsIncomplete)
+			}
+		})
+	}
+}
+
 func setupDefaultDiff() *Diff {
 	return &Diff{
 		Files: []*DiffFile{