Improve sync performance for pull-mirrors (#19125)

This addresses https://github.com/go-gitea/gitea/issues/18352 It aims to improve performance (and resource use) of the `SyncReleasesWithTags` operation for pull-mirrors. For large repositories with many tags, `SyncReleasesWithTags` can be a costly operation (taking several minutes to complete). The reason is two-fold: 1. on sync, every upstream repo tag is compared (for changes) against existing local entries in the release table to ensure that they are up-to-date. 2. the procedure for getting _each tag_ involves a series of git operations ```bash git show-ref --tags -- v8.2.4477 git cat-file -t 29ab6ce9f36660cffaad3c8789e71162e5db5d2f git cat-file -p 29ab6ce9f36660cffaad3c8789e71162e5db5d2f git rev-list --count 29ab6ce9f36660cffaad3c8789e71162e5db5d2f ``` of which the `git rev-list --count` can be particularly heavy. This PR optimizes performance for pull-mirrors. We utilize the fact that a pull-mirror is always identical to its upstream and rebuild the entire release table on every sync and use a batch `git for-each-ref .. refs/tags` call to retrieve all tags in one go. For large mirror repos, with hundreds of annotated tags, this brings down the duration of the sync operation from several minutes to a few seconds. A few unscientific examples run on my local machine: - https://github.com/spring-projects/spring-boot (223 tags) - before: `0m28,673s` - after: `0m2,244s` - https://github.com/kubernetes/kubernetes (890 tags) - before: `8m00s` - after: `0m8,520s` - https://github.com/vim/vim (13954 tags) - before: `14m20,383s` - after: `0m35,467s` I added a `foreachref` package which contains a flexible way of specifying which reference fields are of interest (`git-for-each-ref(1)`) and to produce a parser for the expected output. These could be reused in other places where `for-each-ref` is used. I'll add unit tests for those if the overall PR looks promising.
2024-11-10 20:59:31 +01:00 · 2022-03-31 14:30:40 +02:00 · 2022-03-31 14:30:40 +02:00 · e28cc79c92
commit e28cc79c92
parent b877504b03
7 changed files with 836 additions and 21 deletions
--- a/modules/git/foreachref/format.go
+++ b/modules/git/foreachref/format.go
@ -0,0 +1,84 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package foreachref
+
+import (
+	"encoding/hex"
+	"fmt"
+	"io"
+	"strings"
+)
+
+var (
+	nullChar     = []byte("\x00")
+	dualNullChar = []byte("\x00\x00")
+)
+
+// Format supports specifying and parsing an output format for 'git
+// for-each-ref'. See See git-for-each-ref(1) for available fields.
+type Format struct {
+	// fieldNames hold %(fieldname)s to be passed to the '--format' flag of
+	// for-each-ref. See git-for-each-ref(1) for available fields.
+	fieldNames []string
+
+	// fieldDelim is the character sequence that is used to separate fields
+	// for each reference. fieldDelim and refDelim should be selected to not
+	// interfere with each other and to not be present in field values.
+	fieldDelim []byte
+	// fieldDelimStr is a string representation of fieldDelim. Used to save
+	// us from repetitive reallocation whenever we need the delimiter as a
+	// string.
+	fieldDelimStr string
+	// refDelim is the character sequence used to separate reference from
+	// each other in the output. fieldDelim and refDelim should be selected
+	// to not interfere with each other and to not be present in field
+	// values.
+	refDelim []byte
+}
+
+// NewFormat creates a forEachRefFormat using the specified fieldNames. See
+// git-for-each-ref(1) for available fields.
+func NewFormat(fieldNames ...string) Format {
+	return Format{
+		fieldNames:    fieldNames,
+		fieldDelim:    nullChar,
+		fieldDelimStr: string(nullChar),
+		refDelim:      dualNullChar,
+	}
+}
+
+// Flag returns a for-each-ref --format flag value that captures the fieldNames.
+func (f Format) Flag() string {
+	var formatFlag strings.Builder
+	for i, field := range f.fieldNames {
+		// field key and field value
+		formatFlag.WriteString(fmt.Sprintf("%s %%(%s)", field, field))
+
+		if i < len(f.fieldNames)-1 {
+			// note: escape delimiters to allow control characters as
+			// delimiters. For example, '%00' for null character or '%0a'
+			// for newline.
+			formatFlag.WriteString(f.hexEscaped(f.fieldDelim))
+		}
+	}
+	formatFlag.WriteString(f.hexEscaped(f.refDelim))
+	return formatFlag.String()
+}
+
+// Parser returns a Parser capable of parsing 'git for-each-ref' output produced
+// with this Format.
+func (f Format) Parser(r io.Reader) *Parser {
+	return NewParser(r, f)
+}
+
+// hexEscaped produces hex-escpaed characters from a string. For example, "\n\0"
+// would turn into "%0a%00".
+func (f Format) hexEscaped(delim []byte) string {
+	escaped := ""
+	for i := 0; i < len(delim); i++ {
+		escaped += "%" + hex.EncodeToString([]byte{delim[i]})
+	}
+	return escaped
+}
--- a/modules/git/foreachref/format_test.go
+++ b/modules/git/foreachref/format_test.go
@ -0,0 +1,67 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package foreachref_test
+
+import (
+	"testing"
+
+	"code.gitea.io/gitea/modules/git/foreachref"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestFormat_Flag(t *testing.T) {
+	tests := []struct {
+		name string
+
+		givenFormat foreachref.Format
+
+		wantFlag string
+	}{
+		{
+			name: "references are delimited by dual null chars",
+
+			// no reference fields requested
+			givenFormat: foreachref.NewFormat(),
+
+			// only a reference delimiter field in --format
+			wantFlag: "%00%00",
+		},
+
+		{
+			name: "a field is a space-separated key-value pair",
+
+			givenFormat: foreachref.NewFormat("refname:short"),
+
+			// only a reference delimiter field
+			wantFlag: "refname:short %(refname:short)%00%00",
+		},
+
+		{
+			name: "fields are separated by a null char field-delimiter",
+
+			givenFormat: foreachref.NewFormat("refname:short", "author"),
+
+			wantFlag: "refname:short %(refname:short)%00author %(author)%00%00",
+		},
+
+		{
+			name: "multiple fields",
+
+			givenFormat: foreachref.NewFormat("refname:short", "objecttype", "objectname"),
+
+			wantFlag: "refname:short %(refname:short)%00objecttype %(objecttype)%00objectname %(objectname)%00%00",
+		},
+	}
+
+	for _, test := range tests {
+		tc := test // don't close over loop variable
+		t.Run(tc.name, func(t *testing.T) {
+			gotFlag := tc.givenFormat.Flag()
+
+			require.Equal(t, tc.wantFlag, gotFlag, "unexpected for-each-ref --format string. wanted: '%s', got: '%s'", tc.wantFlag, gotFlag)
+		})
+	}
+}
--- a/modules/git/foreachref/parser.go
+++ b/modules/git/foreachref/parser.go
@ -0,0 +1,131 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package foreachref
+
+import (
+	"bufio"
+	"bytes"
+	"fmt"
+	"io"
+	"strings"
+)
+
+// Parser parses 'git for-each-ref' output according to a given output Format.
+type Parser struct {
+	//  tokenizes 'git for-each-ref' output into "reference paragraphs".
+	scanner *bufio.Scanner
+
+	// format represents the '--format' string that describes the expected
+	// 'git for-each-ref' output structure.
+	format Format
+
+	// err holds the last encountered error during parsing.
+	err error
+}
+
+// NewParser creates a 'git for-each-ref' output parser that will parse all
+// references in the provided Reader. The references in the output are assumed
+// to follow the specified Format.
+func NewParser(r io.Reader, format Format) *Parser {
+	scanner := bufio.NewScanner(r)
+
+	// in addition to the reference delimiter we specified in the --format,
+	// `git for-each-ref` will always add a newline after every reference.
+	refDelim := make([]byte, 0, len(format.refDelim)+1)
+	refDelim = append(refDelim, format.refDelim...)
+	refDelim = append(refDelim, '\n')
+
+	// Split input into delimiter-separated "reference blocks".
+	scanner.Split(
+		func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+			// Scan until delimiter, marking end of reference.
+			delimIdx := bytes.Index(data, refDelim)
+			if delimIdx >= 0 {
+				token := data[:delimIdx]
+				advance := delimIdx + len(refDelim)
+				return advance, token, nil
+			}
+			// If we're at EOF, we have a final, non-terminated reference. Return it.
+			if atEOF {
+				return len(data), data, nil
+			}
+			// Not yet a full field. Request more data.
+			return 0, nil, nil
+		})
+
+	return &Parser{
+		scanner: scanner,
+		format:  format,
+		err:     nil,
+	}
+}
+
+// Next returns the next reference as a collection of key-value pairs. nil
+// denotes EOF but is also returned on errors. The Err method should always be
+// consulted after Next returning nil.
+//
+// It could, for example return something like:
+//
+//  { "objecttype": "tag", "refname:short": "v1.16.4", "object": "f460b7543ed500e49c133c2cd85c8c55ee9dbe27" }
+//
+func (p *Parser) Next() map[string]string {
+	if !p.scanner.Scan() {
+		return nil
+	}
+	fields, err := p.parseRef(p.scanner.Text())
+	if err != nil {
+		p.err = err
+		return nil
+	}
+	return fields
+}
+
+// Err returns the latest encountered parsing error.
+func (p *Parser) Err() error {
+	return p.err
+}
+
+// parseRef parses out all key-value pairs from a single reference block, such as
+//
+//   "objecttype tag\0refname:short v1.16.4\0object f460b7543ed500e49c133c2cd85c8c55ee9dbe27"
+//
+func (p *Parser) parseRef(refBlock string) (map[string]string, error) {
+	if refBlock == "" {
+		// must be at EOF
+		return nil, nil
+	}
+
+	fieldValues := make(map[string]string)
+
+	fields := strings.Split(refBlock, p.format.fieldDelimStr)
+	if len(fields) != len(p.format.fieldNames) {
+		return nil, fmt.Errorf("unexpected number of reference fields: wanted %d, was %d",
+			len(fields), len(p.format.fieldNames))
+	}
+	for i, field := range fields {
+		field = strings.TrimSpace(field)
+
+		var fieldKey string
+		var fieldVal string
+		firstSpace := strings.Index(field, " ")
+		if firstSpace > 0 {
+			fieldKey = field[:firstSpace]
+			fieldVal = field[firstSpace+1:]
+		} else {
+			// could be the case if the requested field had no value
+			fieldKey = field
+		}
+
+		// enforce the format order of fields
+		if p.format.fieldNames[i] != fieldKey {
+			return nil, fmt.Errorf("unexpected field name at position %d: wanted: '%s', was: '%s'",
+				i, p.format.fieldNames[i], fieldKey)
+		}
+
+		fieldValues[fieldKey] = fieldVal
+	}
+
+	return fieldValues, nil
+}
--- a/modules/git/foreachref/parser_test.go
+++ b/modules/git/foreachref/parser_test.go
@ -0,0 +1,228 @@
+// Copyright 2022 The Gitea Authors. All rights reserved.
+// Use of this source code is governed by a MIT-style
+// license that can be found in the LICENSE file.
+
+package foreachref_test
+
+import (
+	"errors"
+	"fmt"
+	"io"
+	"strings"
+	"testing"
+
+	"code.gitea.io/gitea/modules/git/foreachref"
+	"code.gitea.io/gitea/modules/json"
+
+	"github.com/stretchr/testify/require"
+)
+
+type refSlice = []map[string]string
+
+func TestParser(t *testing.T) {
+	tests := []struct {
+		name string
+
+		givenFormat foreachref.Format
+		givenInput  io.Reader
+
+		wantRefs    refSlice
+		wantErr     bool
+		expectedErr error
+	}{
+		// this would, for example, be the result when running `git
+		// for-each-ref refs/tags` on a repo without tags.
+		{
+			name: "no references on empty input",
+
+			givenFormat: foreachref.NewFormat("refname:short"),
+			givenInput:  strings.NewReader(``),
+
+			wantRefs: []map[string]string{},
+		},
+
+		// note: `git for-each-ref` will add a newline between every
+		// reference (in addition to the ref-delimiter we've chosen)
+		{
+			name: "single field requested, single reference in output",
+
+			givenFormat: foreachref.NewFormat("refname:short"),
+			givenInput:  strings.NewReader("refname:short v0.0.1\x00\x00" + "\n"),
+
+			wantRefs: []map[string]string{
+				{"refname:short": "v0.0.1"},
+			},
+		},
+		{
+			name: "single field requested, multiple references in output",
+
+			givenFormat: foreachref.NewFormat("refname:short"),
+			givenInput: strings.NewReader(
+				"refname:short v0.0.1\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00\x00" + "\n"),
+
+			wantRefs: []map[string]string{
+				{"refname:short": "v0.0.1"},
+				{"refname:short": "v0.0.2"},
+				{"refname:short": "v0.0.3"},
+			},
+		},
+
+		{
+			name: "multiple fields requested for each reference",
+
+			givenFormat: foreachref.NewFormat("refname:short", "objecttype", "objectname"),
+			givenInput: strings.NewReader(
+
+				"refname:short v0.0.1\x00objecttype commit\x00objectname 7b2c5ac9fc04fc5efafb60700713d4fa609b777b\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00objecttype commit\x00objectname a1f051bc3eba734da4772d60e2d677f47cf93ef4\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00objecttype commit\x00objectname ef82de70bb3f60c65fb8eebacbb2d122ef517385\x00\x00" + "\n",
+			),
+
+			wantRefs: []map[string]string{
+				{
+					"refname:short": "v0.0.1",
+					"objecttype":    "commit",
+					"objectname":    "7b2c5ac9fc04fc5efafb60700713d4fa609b777b",
+				},
+				{
+					"refname:short": "v0.0.2",
+					"objecttype":    "commit",
+					"objectname":    "a1f051bc3eba734da4772d60e2d677f47cf93ef4",
+				},
+				{
+					"refname:short": "v0.0.3",
+					"objecttype":    "commit",
+					"objectname":    "ef82de70bb3f60c65fb8eebacbb2d122ef517385",
+				},
+			},
+		},
+
+		{
+			name: "must handle multi-line fields such as 'content'",
+
+			givenFormat: foreachref.NewFormat("refname:short", "contents", "author"),
+			givenInput: strings.NewReader(
+				"refname:short v0.0.1\x00contents Create new buffer if not present yet (#549)\n\nFixes a nil dereference when ProcessFoo is used\nwith multiple commands.\x00author Foo Bar <foo@bar.com> 1507832733 +0200\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00contents Update CI config (#651)\n\n\x00author John Doe <john.doe@foo.com> 1521643174 +0000\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00contents Fixed code sample for bash completion (#687)\n\n\x00author Foo Baz <foo@baz.com> 1524836750 +0200\x00\x00" + "\n",
+			),
+
+			wantRefs: []map[string]string{
+				{
+					"refname:short": "v0.0.1",
+					"contents":      "Create new buffer if not present yet (#549)\n\nFixes a nil dereference when ProcessFoo is used\nwith multiple commands.",
+					"author":        "Foo Bar <foo@bar.com> 1507832733 +0200",
+				},
+				{
+					"refname:short": "v0.0.2",
+					"contents":      "Update CI config (#651)",
+					"author":        "John Doe <john.doe@foo.com> 1521643174 +0000",
+				},
+				{
+					"refname:short": "v0.0.3",
+					"contents":      "Fixed code sample for bash completion (#687)",
+					"author":        "Foo Baz <foo@baz.com> 1524836750 +0200",
+				},
+			},
+		},
+
+		{
+			name: "must handle fields without values",
+
+			givenFormat: foreachref.NewFormat("refname:short", "object", "objecttype"),
+			givenInput: strings.NewReader(
+				"refname:short v0.0.1\x00object \x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00object \x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00object \x00objecttype commit\x00\x00" + "\n",
+			),
+
+			wantRefs: []map[string]string{
+				{
+					"refname:short": "v0.0.1",
+					"object":        "",
+					"objecttype":    "commit",
+				},
+				{
+					"refname:short": "v0.0.2",
+					"object":        "",
+					"objecttype":    "commit",
+				},
+				{
+					"refname:short": "v0.0.3",
+					"object":        "",
+					"objecttype":    "commit",
+				},
+			},
+		},
+
+		{
+			name: "must fail when the number of fields in the input doesn't match expected format",
+
+			givenFormat: foreachref.NewFormat("refname:short", "objecttype", "objectname"),
+			givenInput: strings.NewReader(
+				"refname:short v0.0.1\x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00objecttype commit\x00\x00" + "\n",
+			),
+
+			wantErr:     true,
+			expectedErr: errors.New("unexpected number of reference fields: wanted 2, was 3"),
+		},
+
+		{
+			name: "must fail input fields don't match expected format",
+
+			givenFormat: foreachref.NewFormat("refname:short", "objectname"),
+			givenInput: strings.NewReader(
+				"refname:short v0.0.1\x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.2\x00objecttype commit\x00\x00" + "\n" +
+					"refname:short v0.0.3\x00objecttype commit\x00\x00" + "\n",
+			),
+
+			wantErr:     true,
+			expectedErr: errors.New("unexpected field name at position 1: wanted: 'objectname', was: 'objecttype'"),
+		},
+	}
+
+	for _, test := range tests {
+		tc := test // don't close over loop variable
+		t.Run(tc.name, func(t *testing.T) {
+			parser := tc.givenFormat.Parser(tc.givenInput)
+
+			//
+			// parse references from input
+			//
+			gotRefs := make([]map[string]string, 0)
+			for {
+				ref := parser.Next()
+				if ref == nil {
+					break
+				}
+				gotRefs = append(gotRefs, ref)
+			}
+			err := parser.Err()
+
+			//
+			// verify expectations
+			//
+			if tc.wantErr {
+				require.Error(t, err)
+				require.EqualError(t, err, tc.expectedErr.Error())
+			} else {
+				require.NoError(t, err, "for-each-ref parser unexpectedly failed with: %v", err)
+				require.Equal(t, tc.wantRefs, gotRefs, "for-each-ref parser produced unexpected reference set. wanted: %v, got: %v", pretty(tc.wantRefs), pretty(gotRefs))
+			}
+		})
+	}
+}
+
+func pretty(v interface{}) string {
+	data, err := json.MarshalIndent(v, "", "  ")
+	if err != nil {
+		// shouldn't happen
+		panic(fmt.Sprintf("json-marshalling failed: %v", err))
+	}
+	return string(data)
+}
--- a/modules/git/repo_tag.go
+++ b/modules/git/repo_tag.go
@ -8,8 +8,10 @@ package git
 import (
 	"context"
 	"fmt"
+	"io"
 	"strings"

+	"code.gitea.io/gitea/modules/git/foreachref"
 	"code.gitea.io/gitea/modules/util"
 )

@ -111,37 +113,98 @@ func (repo *Repository) GetTagWithID(idStr, name string) (*Tag, error) {

 // GetTagInfos returns all tag infos of the repository.
 func (repo *Repository) GetTagInfos(page, pageSize int) ([]*Tag, int, error) {
-	// TODO this a slow implementation, makes one git command per tag
-	stdout, err := NewCommand(repo.Ctx, "tag").RunInDir(repo.Path)
-	if err != nil {
-		return nil, 0, err
-	}
+	forEachRefFmt := foreachref.NewFormat("objecttype", "refname:short", "object", "objectname", "creator", "contents", "contents:signature")

-	tagNames := strings.Split(strings.TrimRight(stdout, "\n"), "\n")
-	tagsTotal := len(tagNames)
+	stdoutReader, stdoutWriter := io.Pipe()
+	defer stdoutReader.Close()
+	defer stdoutWriter.Close()
+	stderr := strings.Builder{}
+	rc := &RunContext{Dir: repo.Path, Stdout: stdoutWriter, Stderr: &stderr, Timeout: -1}

-	if page != 0 {
-		tagNames = util.PaginateSlice(tagNames, page, pageSize).([]string)
-	}
-
-	tags := make([]*Tag, 0, len(tagNames))
-	for _, tagName := range tagNames {
-		tagName = strings.TrimSpace(tagName)
-		if len(tagName) == 0 {
-			continue
-		}
-
-		tag, err := repo.GetTag(tagName)
+	go func() {
+		err := NewCommand(repo.Ctx, "for-each-ref", "--format", forEachRefFmt.Flag(), "--sort", "-*creatordate", "refs/tags").RunWithContext(rc)
 		if err != nil {
-			return nil, tagsTotal, err
+			_ = stdoutWriter.CloseWithError(ConcatenateError(err, stderr.String()))
+		} else {
+			_ = stdoutWriter.Close()
+		}
+	}()
+
+	var tags []*Tag
+	parser := forEachRefFmt.Parser(stdoutReader)
+	for {
+		ref := parser.Next()
+		if ref == nil {
+			break
+		}
+
+		tag, err := parseTagRef(ref)
+		if err != nil {
+			return nil, 0, fmt.Errorf("GetTagInfos: parse tag: %w", err)
 		}
-		tag.Name = tagName
 		tags = append(tags, tag)
 	}
+	if err := parser.Err(); err != nil {
+		return nil, 0, fmt.Errorf("GetTagInfos: parse output: %w", err)
+	}
+
 	sortTagsByTime(tags)
+	tagsTotal := len(tags)
+	if page != 0 {
+		tags = util.PaginateSlice(tags, page, pageSize).([]*Tag)
+	}
+
 	return tags, tagsTotal, nil
 }

+// parseTagRef parses a tag from a 'git for-each-ref'-produced reference.
+func parseTagRef(ref map[string]string) (tag *Tag, err error) {
+	tag = &Tag{
+		Type: ref["objecttype"],
+		Name: ref["refname:short"],
+	}
+
+	tag.ID, err = NewIDFromString(ref["objectname"])
+	if err != nil {
+		return nil, fmt.Errorf("parse objectname '%s': %v", ref["objectname"], err)
+	}
+
+	if tag.Type == "commit" {
+		// lightweight tag
+		tag.Object = tag.ID
+	} else {
+		// annotated tag
+		tag.Object, err = NewIDFromString(ref["object"])
+		if err != nil {
+			return nil, fmt.Errorf("parse object '%s': %v", ref["object"], err)
+		}
+	}
+
+	tag.Tagger, err = newSignatureFromCommitline([]byte(ref["creator"]))
+	if err != nil {
+		return nil, fmt.Errorf("parse tagger: %w", err)
+	}
+
+	tag.Message = ref["contents"]
+	// strip PGP signature if present in contents field
+	pgpStart := strings.Index(tag.Message, beginpgp)
+	if pgpStart >= 0 {
+		tag.Message = tag.Message[0:pgpStart]
+	}
+
+	// annotated tag with GPG signature
+	if tag.Type == "tag" && ref["contents:signature"] != "" {
+		payload := fmt.Sprintf("object %s\ntype commit\ntag %s\ntagger %s\n\n%s\n",
+			tag.Object, tag.Name, ref["creator"], strings.TrimSpace(tag.Message))
+		tag.Signature = &CommitGPGSignature{
+			Signature: ref["contents:signature"],
+			Payload:   payload,
+		}
+	}
+
+	return tag, nil
+}
+
 // GetAnnotatedTag returns a Git tag by its SHA, must be an annotated tag
 func (repo *Repository) GetAnnotatedTag(sha string) (*Tag, error) {
 	id, err := NewIDFromString(sha)
--- a/modules/git/repo_tag_test.go
+++ b/modules/git/repo_tag_test.go
@ -11,6 +11,7 @@ import (
 	"code.gitea.io/gitea/modules/util"

 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )

 func TestRepository_GetTags(t *testing.T) {
@ -195,3 +196,184 @@ func TestRepository_GetAnnotatedTag(t *testing.T) {
 	assert.True(t, IsErrNotExist(err))
 	assert.Nil(t, tag4)
 }
+
+func TestRepository_parseTagRef(t *testing.T) {
+	tests := []struct {
+		name string
+
+		givenRef map[string]string
+
+		want        *Tag
+		wantErr     bool
+		expectedErr error
+	}{
+		{
+			name: "lightweight tag",
+
+			givenRef: map[string]string{
+				"objecttype":    "commit",
+				"refname:short": "v1.9.1",
+				// object will be empty for lightweight tags
+				"object":     "",
+				"objectname": "ab23e4b7f4cd0caafe0174c0e7ef6d651ba72889",
+				"creator":    "Foo Bar <foo@bar.com> 1565789218 +0300",
+				"contents": `Add changelog of v1.9.1 (#7859)
+
+* add changelog of v1.9.1
+* Update CHANGELOG.md
+`,
+				"contents:signature": "",
+			},
+
+			want: &Tag{
+				Name:      "v1.9.1",
+				ID:        MustIDFromString("ab23e4b7f4cd0caafe0174c0e7ef6d651ba72889"),
+				Object:    MustIDFromString("ab23e4b7f4cd0caafe0174c0e7ef6d651ba72889"),
+				Type:      "commit",
+				Tagger:    parseAuthorLine(t, "Foo Bar <foo@bar.com> 1565789218 +0300"),
+				Message:   "Add changelog of v1.9.1 (#7859)\n\n* add changelog of v1.9.1\n* Update CHANGELOG.md\n",
+				Signature: nil,
+			},
+		},
+
+		{
+			name: "annotated tag",
+
+			givenRef: map[string]string{
+				"objecttype":    "tag",
+				"refname:short": "v0.0.1",
+				// object will refer to commit hash for annotated tag
+				"object":     "3325fd8a973321fd59455492976c042dde3fd1ca",
+				"objectname": "8c68a1f06fc59c655b7e3905b159d761e91c53c9",
+				"creator":    "Foo Bar <foo@bar.com> 1565789218 +0300",
+				"contents": `Add changelog of v1.9.1 (#7859)
+
+* add changelog of v1.9.1
+* Update CHANGELOG.md
+`,
+				"contents:signature": "",
+			},
+
+			want: &Tag{
+				Name:      "v0.0.1",
+				ID:        MustIDFromString("8c68a1f06fc59c655b7e3905b159d761e91c53c9"),
+				Object:    MustIDFromString("3325fd8a973321fd59455492976c042dde3fd1ca"),
+				Type:      "tag",
+				Tagger:    parseAuthorLine(t, "Foo Bar <foo@bar.com> 1565789218 +0300"),
+				Message:   "Add changelog of v1.9.1 (#7859)\n\n* add changelog of v1.9.1\n* Update CHANGELOG.md\n",
+				Signature: nil,
+			},
+		},
+
+		{
+			name: "annotated tag with signature",
+
+			givenRef: map[string]string{
+				"objecttype":    "tag",
+				"refname:short": "v0.0.1",
+				"object":        "3325fd8a973321fd59455492976c042dde3fd1ca",
+				"objectname":    "8c68a1f06fc59c655b7e3905b159d761e91c53c9",
+				"creator":       "Foo Bar <foo@bar.com> 1565789218 +0300",
+				"contents": `Add changelog of v1.9.1 (#7859)
+
+* add changelog of v1.9.1
+* Update CHANGELOG.md
+-----BEGIN PGP SIGNATURE-----
+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+=2yGi
+-----END PGP SIGNATURE-----
+
+`,
+				"contents:signature": `-----BEGIN PGP SIGNATURE-----
+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+=2yGi
+-----END PGP SIGNATURE-----
+
+`,
+			},
+
+			want: &Tag{
+				Name:    "v0.0.1",
+				ID:      MustIDFromString("8c68a1f06fc59c655b7e3905b159d761e91c53c9"),
+				Object:  MustIDFromString("3325fd8a973321fd59455492976c042dde3fd1ca"),
+				Type:    "tag",
+				Tagger:  parseAuthorLine(t, "Foo Bar <foo@bar.com> 1565789218 +0300"),
+				Message: "Add changelog of v1.9.1 (#7859)\n\n* add changelog of v1.9.1\n* Update CHANGELOG.md",
+				Signature: &CommitGPGSignature{
+					Signature: `-----BEGIN PGP SIGNATURE-----
+
+aBCGzBAABCgAdFiEEyWRwv/q1Q6IjSv+D4IPOwzt33PoFAmI8jbIACgkQ4IPOwzt3
+3PoRuAv9FVSbPBXvzECubls9KQd7urwEvcfG20Uf79iBwifQJUv+egNQojrs6APT
+T4CdIXeGRpwJZaGTUX9RWnoDO1SLXAWnc82CypWraNwrHq8Go2YeoVu0Iy3vb0EU
+REdob/tXYZecMuP8AjhUR0XfdYaERYAvJ2dYsH/UkFrqDjM3V4kPXWG+R5DCaZiE
+slB5U01i4Dwb/zm/ckzhUGEcOgcnpOKX8SnY5kYRVDY47dl/yJZ1u2XWir3mu60G
+1geIitH7StBddHi/8rz+sJwTfcVaLjn2p59p/Dr9aGbk17GIaKq1j0pZA2lKT0Xt
+f9jDqU+9vCxnKgjSDhrwN69LF2jT47ZFjEMGV/wFPOa1EBxVWpgQ/CfEolBlbUqx
+yVpbxi/6AOK2lmG130e9jEZJcu+WeZUeq851WgKSEkf2d5f/JpwtSTEOlOedu6V6
+kl845zu5oE2nKM4zMQ7XrYQn538I31ps+VGQ0H8R07WrZP8WKUWugL2cU8KmXFwg
+qbHDASXl
+=2yGi
+-----END PGP SIGNATURE-----
+
+`,
+					Payload: `object 3325fd8a973321fd59455492976c042dde3fd1ca
+type commit
+tag v0.0.1
+tagger Foo Bar <foo@bar.com> 1565789218 +0300
+
+Add changelog of v1.9.1 (#7859)
+
+* add changelog of v1.9.1
+* Update CHANGELOG.md
+`,
+				},
+			},
+		},
+	}
+
+	for _, test := range tests {
+		tc := test // don't close over loop variable
+		t.Run(tc.name, func(t *testing.T) {
+			got, err := parseTagRef(tc.givenRef)
+
+			if tc.wantErr {
+				require.Error(t, err)
+				require.ErrorIs(t, err, tc.expectedErr)
+			} else {
+				require.NoError(t, err)
+				require.Equal(t, tc.want, got)
+			}
+		})
+	}
+}
+
+func parseAuthorLine(t *testing.T, committer string) *Signature {
+	t.Helper()
+
+	sig, err := newSignatureFromCommitline([]byte(committer))
+	if err != nil {
+		t.Fatalf("parse author line '%s': %v", committer, err)
+	}
+
+	return sig
+}
--- a/modules/repository/repo.go
+++ b/modules/repository/repo.go
@ -150,6 +150,9 @@ func MigrateRepositoryGitData(ctx context.Context, u *user_model.User,
 		}

 		if !opts.Releases {
+			// note: this will greatly improve release (tag) sync
+			// for pull-mirrors with many tags
+			repo.IsMirror = opts.Mirror
 			if err = SyncReleasesWithTags(repo, gitRepo); err != nil {
 				log.Error("Failed to synchronize tags to releases for repository: %v", err)
 			}
@ -254,6 +257,14 @@ func CleanUpMigrateInfo(ctx context.Context, repo *repo_model.Repository) (*repo

 // SyncReleasesWithTags synchronizes release table with repository tags
 func SyncReleasesWithTags(repo *repo_model.Repository, gitRepo *git.Repository) error {
+	log.Debug("SyncReleasesWithTags: in Repo[%d:%s/%s]", repo.ID, repo.OwnerName, repo.Name)
+
+	// optimized procedure for pull-mirrors which saves a lot of time (in
+	// particular for repos with many tags).
+	if repo.IsMirror {
+		return pullMirrorReleaseSync(repo, gitRepo)
+	}
+
 	existingRelTags := make(map[string]struct{})
 	opts := models.FindReleasesOptions{
 		IncludeDrafts: true,
@ -450,3 +461,52 @@ func StoreMissingLfsObjectsInRepository(ctx context.Context, repo *repo_model.Re

 	return nil
 }
+
+// pullMirrorReleaseSync is a pull-mirror specific tag<->release table
+// synchronization which overwrites all Releases from the repository tags. This
+// can be relied on since a pull-mirror is always identical to its
+// upstream. Hence, after each sync we want the pull-mirror release set to be
+// identical to the upstream tag set. This is much more efficient for
+// repositories like https://github.com/vim/vim (with over 13000 tags).
+func pullMirrorReleaseSync(repo *repo_model.Repository, gitRepo *git.Repository) error {
+	log.Trace("pullMirrorReleaseSync: rebuilding releases for pull-mirror Repo[%d:%s/%s]", repo.ID, repo.OwnerName, repo.Name)
+	tags, numTags, err := gitRepo.GetTagInfos(0, 0)
+	if err != nil {
+		return fmt.Errorf("unable to GetTagInfos in pull-mirror Repo[%d:%s/%s]: %w", repo.ID, repo.OwnerName, repo.Name, err)
+	}
+	err = db.WithTx(func(ctx context.Context) error {
+		//
+		// clear out existing releases
+		//
+		if _, err := db.DeleteByBean(ctx, &models.Release{RepoID: repo.ID}); err != nil {
+			return fmt.Errorf("unable to clear releases for pull-mirror Repo[%d:%s/%s]: %w", repo.ID, repo.OwnerName, repo.Name, err)
+		}
+		//
+		// make release set identical to upstream tags
+		//
+		for _, tag := range tags {
+			release := models.Release{
+				RepoID:       repo.ID,
+				TagName:      tag.Name,
+				LowerTagName: strings.ToLower(tag.Name),
+				Sha1:         tag.Object.String(),
+				// NOTE: ignored, since NumCommits are unused
+				// for pull-mirrors (only relevant when
+				// displaying releases, IsTag: false)
+				NumCommits:  -1,
+				CreatedUnix: timeutil.TimeStamp(tag.Tagger.When.Unix()),
+				IsTag:       true,
+			}
+			if err := db.Insert(ctx, release); err != nil {
+				return fmt.Errorf("unable insert tag %s for pull-mirror Repo[%d:%s/%s]: %w", tag.Name, repo.ID, repo.OwnerName, repo.Name, err)
+			}
+		}
+		return nil
+	})
+	if err != nil {
+		return fmt.Errorf("unable to rebuild release table for pull-mirror Repo[%d:%s/%s]: %w", repo.ID, repo.OwnerName, repo.Name, err)
+	}
+
+	log.Trace("pullMirrorReleaseSync: done rebuilding %d releases", numTags)
+	return nil
+}