// ReadAllHeaders reads the file headers and extended headers (if any)
// from a file unified diff. It does not read hunks, and the returned
// FileDiff's Hunks field is nil. To read the hunks, call the
// (*FileDiffReader).HunksReader() method to get a HunksReader and
// read hunks from that.
func (r *FileDiffReader) ReadAllHeaders() (*FileDiff, error) {
var err error
fd := &FileDiff{}
fd.Extended, err = r.ReadExtendedHeaders()
if err != nil {
return nil, err
}
var origTime, newTime *time.Time
fd.OrigName, fd.NewName, origTime, newTime, err = r.ReadFileHeaders()
if err != nil {
return nil, err
}
if origTime != nil {
ts := pbtypes.NewTimestamp(*origTime)
fd.OrigTime = &ts
}
if newTime != nil {
ts := pbtypes.NewTimestamp(*newTime)
fd.NewTime = &ts
}
return fd, nil
}
func (r *Repository) makeCommit(rec *hg_revlog.Rec) (*vcs.Commit, error) {
fb := hg_revlog.NewFileBuilder()
ce, err := hg_changelog.BuildEntry(rec, fb)
if err != nil {
return nil, err
}
addr, err := mail.ParseAddress(ce.Committer)
if err != nil {
// This occurs when the commit author specifier is
// malformed. Fall back to just using the whole committer
// string as the name.
addr = &mail.Address{
Name: ce.Committer,
Address: "",
}
}
var parents []vcs.CommitID
if !rec.IsStartOfBranch() {
if p := rec.Parent(); p != nil {
parents = append(parents, vcs.CommitID(hex.EncodeToString(rec.Parent().Id())))
}
if rec.Parent2Present() {
parents = append(parents, vcs.CommitID(hex.EncodeToString(rec.Parent2().Id())))
}
}
return &vcs.Commit{
ID: vcs.CommitID(ce.Id),
Author: vcs.Signature{addr.Name, addr.Address, pbtypes.NewTimestamp(ce.Date)},
Message: ce.Comment,
Parents: parents,
}, nil
}
func TestServeRepoTreeEntry_File(t *testing.T) {
setupHandlerTest()
defer teardownHandlerTest()
commitID := vcs.CommitID(strings.Repeat("a", 40))
repoPath := "a.b/c"
rm := &mockFileSystem{
t: t,
at: commitID,
fs: mapFS(map[string]string{"myfile": "mydata"}),
}
sm := &mockServiceForExistingRepo{
t: t,
repoPath: repoPath,
repo: rm,
}
testHandler.Service = sm
resp, err := http.Get(server.URL + testHandler.router.URLToRepoTreeEntry(repoPath, commitID, "myfile").String())
if err != nil {
t.Fatal(err)
}
defer resp.Body.Close()
if got, want := resp.StatusCode, http.StatusOK; got != want {
t.Errorf("got status code %d, want %d", got, want)
}
if !sm.opened {
t.Errorf("!opened")
}
if !rm.called {
t.Errorf("!called")
}
var e *vcsclient.TreeEntry
if err := json.NewDecoder(resp.Body).Decode(&e); err != nil {
t.Fatal(err)
}
wantEntry := &vcsclient.TreeEntry{
Name: "myfile",
Type: vcsclient.FileEntry,
Size: 6,
ModTime: pbtypes.NewTimestamp(time.Time{}),
Contents: []byte("mydata"),
}
if !reflect.DeepEqual(e, wantEntry) {
t.Errorf("got tree entry %+v, want %+v", e, wantEntry)
}
// used canonical commit ID, so should be long-cached
if cc := resp.Header.Get("cache-control"); cc != longCacheControl {
t.Errorf("got cache-control %q, want %q", cc, longCacheControl)
}
}
func (r *Repository) makeCommit(c *git2go.Commit) *vcs.Commit {
var parents []vcs.CommitID
if pc := c.ParentCount(); pc > 0 {
parents = make([]vcs.CommitID, pc)
for i := 0; i < int(pc); i++ {
parents[i] = vcs.CommitID(c.ParentId(uint(i)).String())
}
}
au, cm := c.Author(), c.Committer()
return &vcs.Commit{
ID: vcs.CommitID(c.Id().String()),
Author: vcs.Signature{au.Name, au.Email, pbtypes.NewTimestamp(au.When)},
Committer: &vcs.Signature{cm.Name, cm.Email, pbtypes.NewTimestamp(cm.When)},
Message: strings.TrimSuffix(c.Message(), "\n"),
Parents: parents,
}
}
// Convert a git.Commit to a vcs.Commit
func (r *Repository) vcsCommit(commit *git.Commit) *vcs.Commit {
var committer *vcs.Signature
if commit.Committer != nil {
committer = &vcs.Signature{
Name: commit.Committer.Name,
Email: commit.Committer.Email,
Date: pbtypes.NewTimestamp(commit.Committer.When),
}
}
n := commit.ParentCount()
parentIds := commit.ParentIds()
parents := make([]vcs.CommitID, 0, len(parentIds))
for _, id := range parentIds {
parents = append(parents, vcs.CommitID(id.String()))
}
if n == 0 {
// Required to make reflect.DeepEqual tests pass. :/
parents = nil
}
var author vcs.Signature
if commit.Author != nil {
author.Name = commit.Author.Name
author.Email = commit.Author.Email
author.Date = pbtypes.NewTimestamp(commit.Author.When)
}
return &vcs.Commit{
ID: vcs.CommitID(commit.Id.String()),
Author: author,
Committer: committer,
Message: strings.TrimSuffix(commit.Message(), "\n"),
Parents: parents,
}
}
func newTreeEntry(fi os.FileInfo) *TreeEntry {
e := &TreeEntry{
Name: fi.Name(),
Size: fi.Size(),
ModTime: pbtypes.NewTimestamp(fi.ModTime()),
}
if fi.Mode().IsDir() {
e.Type = DirEntry
} else if fi.Mode().IsRegular() {
e.Type = FileEntry
} else if fi.Mode()&os.ModeSymlink != 0 {
e.Type = SymlinkEntry
}
return e
}
func (r *Repository) BlameFile(path string, opt *vcs.BlameOptions) ([]*vcs.Hunk, error) {
if opt == nil {
opt = &vcs.BlameOptions{}
}
// TODO(sqs): implement OldestCommit
cmd := exec.Command("python", "-", r.Dir, string(opt.NewestCommit), path)
cmd.Dir = r.Dir
cmd.Stdin = strings.NewReader(hgRepoAnnotatePy)
stdout, err := cmd.StdoutPipe()
if err != nil {
return nil, err
}
stderr, err := cmd.StderrPipe()
if err != nil {
return nil, err
}
in := bufio.NewReader(stdout)
if err := cmd.Start(); err != nil {
return nil, err
}
var data struct {
Commits map[string]struct {
Author struct{ Name, Email string }
AuthorDate time.Time
}
Hunks map[string][]struct {
CommitID string
StartLine, EndLine int
StartByte, EndByte int
}
}
jsonErr := json.NewDecoder(in).Decode(&data)
errOut, _ := ioutil.ReadAll(stderr)
if jsonErr != nil {
cmd.Wait()
return nil, fmt.Errorf("%s (stderr: %s)", jsonErr, errOut)
}
if err := cmd.Wait(); err != nil {
return nil, fmt.Errorf("%s (stderr: %s)", err, errOut)
}
hunks := make([]*vcs.Hunk, len(data.Hunks[path]))
for i, hunk := range data.Hunks[path] {
c := data.Commits[hunk.CommitID]
hunks[i] = &vcs.Hunk{
StartLine: hunk.StartLine,
EndLine: hunk.EndLine,
StartByte: hunk.StartByte,
EndByte: hunk.EndByte,
CommitID: vcs.CommitID(hunk.CommitID),
Author: vcs.Signature{
Name: c.Author.Name,
Email: c.Author.Email,
Date: pbtypes.NewTimestamp(c.AuthorDate.In(time.UTC)),
},
}
}
return hunks, nil
}
func (r *Repository) commitLog(opt vcs.CommitsOptions) ([]*vcs.Commit, uint, error) {
revSpec := string(opt.Head)
if opt.Skip != 0 {
revSpec += "~" + strconv.FormatUint(uint64(opt.N), 10)
}
args := []string{"log", `--template={node}\x00{author|person}\x00{author|email}\x00{date|rfc3339date}\x00{desc}\x00{p1node}\x00{p2node}\x00`}
if opt.N != 0 {
args = append(args, "--limit", strconv.FormatUint(uint64(opt.N), 10))
}
args = append(args, "--rev="+revSpec+":0")
cmd := exec.Command("hg", args...)
cmd.Dir = r.Dir
out, err := cmd.CombinedOutput()
if err != nil {
out = bytes.TrimSpace(out)
if isUnknownRevisionError(string(out), revSpec) {
return nil, 0, vcs.ErrCommitNotFound
}
return nil, 0, fmt.Errorf("exec `hg log` failed: %s. Output was:\n\n%s", err, out)
}
const partsPerCommit = 7 // number of \x00-separated fields per commit
allParts := bytes.Split(out, []byte{'\x00'})
numCommits := len(allParts) / partsPerCommit
commits := make([]*vcs.Commit, numCommits)
for i := 0; i < numCommits; i++ {
parts := allParts[partsPerCommit*i : partsPerCommit*(i+1)]
id := vcs.CommitID(parts[0])
authorTime, err := time.Parse(time.RFC3339, string(parts[3]))
if err != nil {
log.Println(err)
//return nil, 0, err
}
parents, err := r.getParents(id)
if err != nil {
return nil, 0, fmt.Errorf("r.GetParents failed: %s. Output was:\n\n%s", err, out)
}
commits[i] = &vcs.Commit{
ID: id,
Author: vcs.Signature{string(parts[1]), string(parts[2]), pbtypes.NewTimestamp(authorTime)},
Message: string(parts[4]),
Parents: parents,
}
}
// Count commits.
var total uint
if !opt.NoTotal {
cmd = exec.Command("hg", "id", "--num", "--rev="+revSpec)
cmd.Dir = r.Dir
out, err = cmd.CombinedOutput()
if err != nil {
return nil, 0, fmt.Errorf("exec `hg id --num` failed: %s. Output was:\n\n%s", err, out)
}
out = bytes.TrimSpace(out)
total, err = parseUint(string(out))
if err != nil {
return nil, 0, err
}
total++ // sequence number is 1 less than total number of commits
// Add back however many we skipped.
total += opt.Skip
}
return commits, total, nil
}
func (r *Repository) BlameFile(path string, opt *vcs.BlameOptions) ([]*vcs.Hunk, error) {
r.editLock.RLock()
defer r.editLock.RUnlock()
if opt == nil {
opt = &vcs.BlameOptions{}
}
if opt.OldestCommit != "" {
return nil, fmt.Errorf("OldestCommit not implemented")
}
if err := checkSpecArgSafety(string(opt.NewestCommit)); err != nil {
return nil, err
}
if err := checkSpecArgSafety(string(opt.OldestCommit)); err != nil {
return nil, err
}
args := []string{"blame", "-w", "--porcelain"}
if opt.StartLine != 0 || opt.EndLine != 0 {
args = append(args, fmt.Sprintf("-L%d,%d", opt.StartLine, opt.EndLine))
}
args = append(args, string(opt.NewestCommit), "--", filepath.ToSlash(path))
cmd := exec.Command("git", args...)
cmd.Dir = r.Dir
out, err := cmd.CombinedOutput()
if err != nil {
return nil, fmt.Errorf("exec `git blame` failed: %s. Output was:\n\n%s", err, out)
}
if len(out) < 1 {
// go 1.8.5 changed the behavior of `git blame` on empty files.
// previously, it returned a boundary commit. now, it returns nothing.
// TODO(sqs) TODO(beyang): make `git blame` return the boundary commit
// on an empty file somehow, or come up with some other workaround.
st, err := os.Stat(filepath.Join(r.Dir, path))
if err == nil && st.Size() == 0 {
return nil, nil
}
return nil, fmt.Errorf("Expected git output of length at least 1")
}
commits := make(map[string]vcs.Commit)
hunks := make([]*vcs.Hunk, 0)
remainingLines := strings.Split(string(out[:len(out)-1]), "\n")
byteOffset := 0
for len(remainingLines) > 0 {
// Consume hunk
hunkHeader := strings.Split(remainingLines[0], " ")
if len(hunkHeader) != 4 {
fmt.Printf("Remaining lines: %+v, %d, '%s'\n", remainingLines, len(remainingLines), remainingLines[0])
return nil, fmt.Errorf("Expected at least 4 parts to hunkHeader, but got: '%s'", hunkHeader)
}
commitID := hunkHeader[0]
lineNoCur, _ := strconv.Atoi(hunkHeader[2])
nLines, _ := strconv.Atoi(hunkHeader[3])
hunk := &vcs.Hunk{
CommitID: vcs.CommitID(commitID),
StartLine: int(lineNoCur),
EndLine: int(lineNoCur + nLines),
StartByte: byteOffset,
}
if _, in := commits[commitID]; in {
// Already seen commit
byteOffset += len(remainingLines[1])
remainingLines = remainingLines[2:]
} else {
// New commit
author := strings.Join(strings.Split(remainingLines[1], " ")[1:], " ")
email := strings.Join(strings.Split(remainingLines[2], " ")[1:], " ")
if len(email) >= 2 && email[0] == '<' && email[len(email)-1] == '>' {
email = email[1 : len(email)-1]
}
authorTime, err := strconv.ParseInt(strings.Join(strings.Split(remainingLines[3], " ")[1:], " "), 10, 64)
if err != nil {
return nil, fmt.Errorf("Failed to parse author-time %q", remainingLines[3])
}
summary := strings.Join(strings.Split(remainingLines[9], " ")[1:], " ")
commit := vcs.Commit{
ID: vcs.CommitID(commitID),
Message: summary,
Author: vcs.Signature{
Name: author,
Email: email,
Date: pbtypes.NewTimestamp(time.Unix(authorTime, 0).In(time.UTC)),
},
}
if len(remainingLines) >= 13 && strings.HasPrefix(remainingLines[10], "previous ") {
byteOffset += len(remainingLines[12])
remainingLines = remainingLines[13:]
} else if len(remainingLines) >= 13 && remainingLines[10] == "boundary" {
byteOffset += len(remainingLines[12])
remainingLines = remainingLines[13:]
} else if len(remainingLines) >= 12 {
byteOffset += len(remainingLines[11])
remainingLines = remainingLines[12:]
} else if len(remainingLines) == 11 {
// Empty file
remainingLines = remainingLines[11:]
} else {
return nil, fmt.Errorf("Unexpected number of remaining lines (%d):\n%s", len(remainingLines), " "+strings.Join(remainingLines, "\n "))
}
commits[commitID] = commit
}
if commit, present := commits[commitID]; present {
// Should always be present, but check just to avoid
// panicking in case of a (somewhat likely) bug in our
// git-blame parser above.
hunk.CommitID = commit.ID
hunk.Author = commit.Author
}
// Consume remaining lines in hunk
for i := 1; i < nLines; i++ {
byteOffset += len(remainingLines[1])
remainingLines = remainingLines[2:]
}
hunk.EndByte = byteOffset
hunks = append(hunks, hunk)
}
return hunks, nil
}
// commitLog returns a list of commits, and total number of commits
// starting from Head until Base or beginning of branch (unless NoTotal is true).
//
// The caller is responsible for doing checkSpecArgSafety on opt.Head and opt.Base.
func (r *Repository) commitLog(opt vcs.CommitsOptions) ([]*vcs.Commit, uint, error) {
args := []string{"log", `--format=format:%H%x00%aN%x00%aE%x00%at%x00%cN%x00%cE%x00%ct%x00%B%x00%P%x00`}
if opt.N != 0 {
args = append(args, "-n", strconv.FormatUint(uint64(opt.N), 10))
}
if opt.Skip != 0 {
args = append(args, "--skip="+strconv.FormatUint(uint64(opt.Skip), 10))
}
if opt.Path != "" {
args = append(args, "--follow")
}
// Range
rng := string(opt.Head)
if opt.Base != "" {
rng += "..." + string(opt.Base)
}
args = append(args, rng)
if opt.Path != "" {
args = append(args, "--", opt.Path)
}
cmd := exec.Command("git", args...)
cmd.Dir = r.Dir
out, err := cmd.CombinedOutput()
if err != nil {
out = bytes.TrimSpace(out)
if isBadObjectErr(string(out), string(opt.Head)) {
return nil, 0, vcs.ErrCommitNotFound
}
return nil, 0, fmt.Errorf("exec `git log` failed: %s. Output was:\n\n%s", err, out)
}
const partsPerCommit = 9 // number of \x00-separated fields per commit
allParts := bytes.Split(out, []byte{'\x00'})
numCommits := len(allParts) / partsPerCommit
commits := make([]*vcs.Commit, numCommits)
for i := 0; i < numCommits; i++ {
parts := allParts[partsPerCommit*i : partsPerCommit*(i+1)]
// log outputs are newline separated, so all but the 1st commit ID part
// has an erroneous leading newline.
parts[0] = bytes.TrimPrefix(parts[0], []byte{'\n'})
authorTime, err := strconv.ParseInt(string(parts[3]), 10, 64)
if err != nil {
return nil, 0, fmt.Errorf("parsing git commit author time: %s", err)
}
committerTime, err := strconv.ParseInt(string(parts[6]), 10, 64)
if err != nil {
return nil, 0, fmt.Errorf("parsing git commit committer time: %s", err)
}
var parents []vcs.CommitID
if parentPart := parts[8]; len(parentPart) > 0 {
parentIDs := bytes.Split(parentPart, []byte{' '})
parents = make([]vcs.CommitID, len(parentIDs))
for i, id := range parentIDs {
parents[i] = vcs.CommitID(id)
}
}
commits[i] = &vcs.Commit{
ID: vcs.CommitID(parts[0]),
Author: vcs.Signature{string(parts[1]), string(parts[2]), pbtypes.NewTimestamp(time.Unix(authorTime, 0))},
Committer: &vcs.Signature{string(parts[4]), string(parts[5]), pbtypes.NewTimestamp(time.Unix(committerTime, 0))},
Message: string(bytes.TrimSuffix(parts[7], []byte{'\n'})),
Parents: parents,
}
}
// Count commits.
var total uint
if !opt.NoTotal {
cmd = exec.Command("git", "rev-list", "--count", rng)
if opt.Path != "" {
// This doesn't include --follow flag because rev-list doesn't support it, so the number may be slightly off.
cmd.Args = append(cmd.Args, "--", opt.Path)
}
cmd.Dir = r.Dir
out, err = cmd.CombinedOutput()
if err != nil {
return nil, 0, fmt.Errorf("exec `git rev-list --count` failed: %s. Output was:\n\n%s", err, out)
}
out = bytes.TrimSpace(out)
total, err = parseUint(string(out))
if err != nil {
return nil, 0, err
}
}
return commits, total, nil
}
func TestServeRepoTreeEntry_Dir(t *testing.T) {
setupHandlerTest()
defer teardownHandlerTest()
repoPath := "a.b/c"
rm := &mockFileSystem{
t: t,
at: "abcd",
fs: mapFS(map[string]string{"myfile": "mydata", "mydir/f": ""}),
}
sm := &mockServiceForExistingRepo{
t: t,
repoPath: repoPath,
repo: rm,
}
testHandler.Service = sm
resp, err := http.Get(server.URL + testHandler.router.URLToRepoTreeEntry(repoPath, "abcd", ".").String())
if err != nil {
t.Fatal(err)
}
defer resp.Body.Close()
if got, want := resp.StatusCode, http.StatusOK; got != want {
t.Errorf("got status code %d, want %d", got, want)
}
if !sm.opened {
t.Errorf("!opened")
}
if !rm.called {
t.Errorf("!called")
}
var e *vcsclient.TreeEntry
if err := json.NewDecoder(resp.Body).Decode(&e); err != nil {
t.Fatal(err)
}
wantEntry := &vcsclient.TreeEntry{
Name: ".",
Type: vcsclient.DirEntry,
ModTime: pbtypes.NewTimestamp(time.Time{}),
Entries: []*vcsclient.TreeEntry{
{
Name: "myfile",
Type: vcsclient.FileEntry,
Size: 6,
ModTime: pbtypes.NewTimestamp(time.Time{}),
},
{
Name: "mydir",
Type: vcsclient.DirEntry,
ModTime: pbtypes.NewTimestamp(time.Time{}),
},
},
}
sort.Sort(vcsclient.TreeEntriesByTypeByName(e.Entries))
sort.Sort(vcsclient.TreeEntriesByTypeByName(wantEntry.Entries))
if !reflect.DeepEqual(e, wantEntry) {
t.Errorf("got tree entry %+v, want %+v", e, wantEntry)
}
// used short commit ID, so should not be long-cached
if cc := resp.Header.Get("cache-control"); cc != shortCacheControl {
t.Errorf("got cache-control %q, want %q", cc, shortCacheControl)
}
}
func (r *Repository) BlameFile(path string, opt *vcs.BlameOptions) ([]*vcs.Hunk, error) {
r.editLock.RLock()
defer r.editLock.RUnlock()
gopt := git2go.BlameOptions{}
if opt != nil {
var err error
if opt.NewestCommit != "" {
gopt.NewestCommit, err = git2go.NewOid(string(opt.NewestCommit))
if err != nil {
return nil, err
}
}
if opt.OldestCommit != "" {
gopt.OldestCommit, err = git2go.NewOid(string(opt.OldestCommit))
if err != nil {
return nil, err
}
}
gopt.MinLine = uint32(opt.StartLine)
gopt.MaxLine = uint32(opt.EndLine)
}
blame, err := r.u.BlameFile(path, &gopt)
if err != nil {
return nil, err
}
defer blame.Free()
// Read file contents so we can set hunk byte start and end.
fs, err := r.FileSystem(vcs.CommitID(gopt.NewestCommit.String()))
if err != nil {
return nil, err
}
b, err := fs.(*gitFSLibGit2).readFileBytes(path)
if err != nil {
return nil, err
}
lines := bytes.SplitAfter(b, []byte{'\n'})
byteOffset := 0
hunks := make([]*vcs.Hunk, blame.HunkCount())
for i := 0; i < len(hunks); i++ {
hunk, err := blame.HunkByIndex(i)
if err != nil {
return nil, err
}
hunkBytes := 0
for j := uint16(0); j < hunk.LinesInHunk; j++ {
hunkBytes += len(lines[j])
}
endByteOffset := byteOffset + hunkBytes
hunks[i] = &vcs.Hunk{
StartLine: int(hunk.FinalStartLineNumber),
EndLine: int(hunk.FinalStartLineNumber + hunk.LinesInHunk),
StartByte: byteOffset,
EndByte: endByteOffset,
CommitID: vcs.CommitID(hunk.FinalCommitId.String()),
Author: vcs.Signature{
Name: hunk.FinalSignature.Name,
Email: hunk.FinalSignature.Email,
Date: pbtypes.NewTimestamp(hunk.FinalSignature.When.In(time.UTC)),
},
}
byteOffset = endByteOffset
lines = lines[hunk.LinesInHunk:]
}
return hunks, nil
}