repository.rb 30.9 KB
Newer Older
1 2
require 'securerandom'

3
class Repository
Lin Jen-Shin's avatar
Lin Jen-Shin committed
4 5 6
  REF_MERGE_REQUEST = 'merge-requests'.freeze
  REF_KEEP_AROUND = 'keep-around'.freeze
  REF_ENVIRONMENTS = 'environments'.freeze
7 8 9 10

  RESERVED_REFS_NAMES = %W[
    heads
    tags
11
    replace
12 13 14 15 16
    #{REF_ENVIRONMENTS}
    #{REF_KEEP_AROUND}
    #{REF_ENVIRONMENTS}
  ].freeze

17 18
  include Gitlab::ShellAdapter

19
  attr_accessor :full_path, :disk_path, :project, :is_wiki
20

21 22
  delegate :ref_name_for_sha, to: :raw_repository

23
  CreateTreeError = Class.new(StandardError)
24

25 26 27 28 29 30
  # Methods that cache data from the Git repository.
  #
  # Each entry in this Array should have a corresponding method with the exact
  # same name. The cache key used by those methods must also match method's
  # name.
  #
31 32 33
  # For example, for entry `:commit_count` there's a method called `commit_count` which
  # stores its data in the `commit_count` cache key.
  CACHED_METHODS = %i(size commit_count rendered_readme contribution_guide
34 35
                      changelog license_blob license_key gitignore koding_yml
                      gitlab_ci_yml branch_names tag_names branch_count
36 37
                      tag_count avatar exists? empty? root_ref has_visible_content?
                      issue_template_names merge_request_template_names).freeze
38 39 40

  # Methods that use cache_method but only memoize the value
  MEMOIZED_CACHED_METHODS = %i(license empty_repo?).freeze
41 42 43 44 45

  # Certain method caches should be refreshed when certain types of files are
  # changed. This Hash maps file types (as returned by Gitlab::FileDetector) to
  # the corresponding methods to call for refreshing caches.
  METHOD_CACHES_FOR_FILE_TYPES = {
46
    readme: :rendered_readme,
47
    changelog: :changelog,
48
    license: %i(license_blob license_key license),
49 50 51 52
    contributing: :contribution_guide,
    gitignore: :gitignore,
    koding: :koding_yml,
    gitlab_ci: :gitlab_ci_yml,
53 54 55
    avatar: :avatar,
    issue_template: :issue_template_names,
    merge_request_template: :merge_request_template_names
Douwe Maan's avatar
Douwe Maan committed
56
  }.freeze
57 58 59 60 61

  # Wraps around the given method and caches its output in Redis and an instance
  # variable.
  #
  # This only works for methods that do not take any arguments.
62
  def self.cache_method(name, fallback: nil, memoize_only: false)
63
    original = :"_uncached_#{name}"
64

65
    alias_method(original, name)
66

67
    define_method(name) do
68 69 70
      cache_method_output(name, fallback: fallback, memoize_only: memoize_only) do
        __send__(original) # rubocop:disable GitlabSecurity/PublicSend
      end
71
    end
72
  end
73

74
  def initialize(full_path, project, disk_path: nil, is_wiki: false)
75
    @full_path = full_path
76
    @disk_path = disk_path || full_path
77
    @project = project
78
    @commit_cache = {}
79
    @is_wiki = is_wiki
80
  end
81

82
  def ==(other)
83 84 85
    @disk_path == other.disk_path
  end

86
  def raw_repository
87
    return nil unless full_path
88

89
    @raw_repository ||= initialize_raw_repository
90 91
  end

92 93
  alias_method :raw, :raw_repository

94
  # Return absolute path to repository
95
  def path_to_repo
96
    @path_to_repo ||= File.expand_path(
97
      File.join(repository_storage_path, disk_path + '.git')
98
    )
99 100
  end

101 102 103 104
  def inspect
    "#<#{self.class.name}:#{@disk_path}>"
  end

105
  def commit(ref = 'HEAD')
106
    return nil unless exists?
107
    return ref if ref.is_a?(::Commit)
108

109 110
    find_commit(ref)
  end
111

112 113 114 115 116 117
  # Finding a commit by the passed SHA
  # Also takes care of caching, based on the SHA
  def commit_by(oid:)
    return @commit_cache[oid] if @commit_cache.key?(oid)

    @commit_cache[oid] = find_commit(oid)
118 119
  end

120
  def commits(ref, path: nil, limit: nil, offset: nil, skip_merges: false, after: nil, before: nil)
121
    options = {
122 123 124 125 126
      repo: raw_repository,
      ref: ref,
      path: path,
      limit: limit,
      offset: offset,
127 128
      after: after,
      before: before,
129
      follow: Array(path).length == 1,
130
      skip_merges: skip_merges
131 132 133
    }

    commits = Gitlab::Git::Commit.where(options)
134
    commits = Commit.decorate(commits, @project) if commits.present?
135 136

    CommitCollection.new(project, commits, ref)
137 138
  end

139 140
  def commits_between(from, to)
    commits = Gitlab::Git::Commit.between(raw_repository, from, to)
141
    commits = Commit.decorate(commits, @project) if commits.present?
142 143 144
    commits
  end

Jacob Vosmaer's avatar
Jacob Vosmaer committed
145
  # Gitaly migration: https://gitlab.com/gitlab-org/gitaly/issues/384
146
  def find_commits_by_message(query, ref = nil, path = nil, limit = 1000, offset = 0)
147 148 149 150
    unless exists? && has_visible_content? && query.present?
      return []
    end

151
    raw_repository.gitaly_migrate(:commits_by_message) do |is_enabled|
152 153 154 155 156 157 158 159
      commits =
        if is_enabled
          find_commits_by_message_by_gitaly(query, ref, path, limit, offset)
        else
          find_commits_by_message_by_shelling_out(query, ref, path, limit, offset)
        end

      CommitCollection.new(project, commits, ref)
160
    end
161 162
  end

163 164 165 166 167 168 169
  def find_branch(name, fresh_repo: true)
    # Since the Repository object may have in-memory index changes, invalidating the memoized Repository object may
    # cause unintended side effects. Because finding a branch is a read-only operation, we can safely instantiate
    # a new repo here to ensure a consistent state to avoid a libgit2 bug where concurrent access (e.g. via git gc)
    # may cause the branch to "disappear" erroneously or have the wrong SHA.
    #
    # See: https://github.com/libgit2/libgit2/issues/1534 and https://gitlab.com/gitlab-org/gitlab-ce/issues/15392
170
    raw_repo = fresh_repo ? initialize_raw_repository : raw_repository
171 172

    raw_repo.find_branch(name)
173 174 175
  end

  def find_tag(name)
176
    tags.find { |tag| tag.name == name }
177 178
  end

179
  def add_branch(user, branch_name, ref)
180
    branch = raw_repository.add_branch(branch_name, user: user, target: ref)
181

182
    after_create_branch
183 184 185 186

    branch
  rescue Gitlab::Git::Repository::InvalidRef
    false
187 188
  end

189
  def add_tag(user, tag_name, target, message = nil)
190
    raw_repository.add_tag(tag_name, user: user, target: target, message: message)
191 192
  rescue Gitlab::Git::Repository::InvalidRef
    false
193 194
  end

195
  def rm_branch(user, branch_name)
196
    before_remove_branch
197

198
    raw_repository.rm_branch(branch_name, user: user)
199

200
    after_remove_branch
201
    true
202 203
  end

Lin Jen-Shin's avatar
Lin Jen-Shin committed
204
  def rm_tag(user, tag_name)
205
    before_remove_tag
206

207
    raw_repository.rm_tag(tag_name, user: user)
Lin Jen-Shin's avatar
Lin Jen-Shin committed
208 209 210

    after_remove_tag
    true
211 212
  end

213 214 215 216
  def ref_names
    branch_names + tag_names
  end

217
  def branch_exists?(branch_name)
218 219
    return false unless raw_repository

220
    branch_names.include?(branch_name)
221 222
  end

223
  def ref_exists?(ref)
224 225
    !!raw_repository&.ref_exists?(ref)
  rescue ArgumentError
226
    false
227 228
  end

229 230 231 232
  # Makes sure a commit is kept around when Git garbage collection runs.
  # Git GC will delete commits from the repository that are no longer in any
  # branches or tags, but we want to keep some of these commits around, for
  # example if they have comments or CI builds.
233
  def keep_around(sha)
234
    return unless sha && commit_by(oid: sha)
235 236 237

    return if kept_around?(sha)

238 239
    # This will still fail if the file is corrupted (e.g. 0 bytes)
    begin
240
      write_ref(keep_around_ref_name(sha), sha)
241
    rescue Rugged::ReferenceError => ex
242
      Rails.logger.error "Unable to create #{REF_KEEP_AROUND} reference for repository #{path}: #{ex}"
243 244
    rescue Rugged::OSError => ex
      raise unless ex.message =~ /Failed to create locked file/ && ex.message =~ /File exists/
245

246
      Rails.logger.error "Unable to create #{REF_KEEP_AROUND} reference for repository #{path}: #{ex}"
247
    end
248 249 250
  end

  def kept_around?(sha)
251
    ref_exists?(keep_around_ref_name(sha))
252
  end
253

254 255 256 257
  def write_ref(ref_path, sha)
    rugged.references.create(ref_path, sha, force: true)
  end

258
  def diverging_commit_counts(branch)
259
    root_ref_hash = raw_repository.commit(root_ref).id
Jeff Stubler's avatar
Jeff Stubler committed
260
    cache.fetch(:"diverging_commit_counts_#{branch.name}") do
261 262
      # Rugged seems to throw a `ReferenceError` when given branch_names rather
      # than SHA-1 hashes
263 264
      number_commits_behind = raw_repository
        .count_commits_between(branch.dereferenced_target.sha, root_ref_hash)
265

266 267
      number_commits_ahead = raw_repository
        .count_commits_between(root_ref_hash, branch.dereferenced_target.sha)
268

269 270 271
      { behind: number_commits_behind, ahead: number_commits_ahead }
    end
  end
272

273 274 275
  def expire_tags_cache
    expire_method_caches(%i(tag_names tag_count))
    @tags = nil
276
  end
277

278
  def expire_branches_cache
279
    expire_method_caches(%i(branch_names branch_count has_visible_content?))
280
    @local_branches = nil
281
    @branch_exists_memo = nil
282 283
  end

284 285
  def expire_statistics_caches
    expire_method_caches(%i(size commit_count))
286 287
  end

288 289
  def expire_all_method_caches
    expire_method_caches(CACHED_METHODS)
290 291
  end

292 293 294 295 296 297 298 299 300
  # Expires the caches of a specific set of methods
  def expire_method_caches(methods)
    methods.each do |key|
      cache.expire(key)

      ivar = cache_instance_variable_name(key)

      remove_instance_variable(ivar) if instance_variable_defined?(ivar)
    end
301 302
  end

303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
  def expire_avatar_cache
    expire_method_caches(%i(avatar))
  end

  # Refreshes the method caches of this repository.
  #
  # types - An Array of file types (e.g. `:readme`) used to refresh extra
  #         caches.
  def refresh_method_caches(types)
    to_refresh = []

    types.each do |type|
      methods = METHOD_CACHES_FOR_FILE_TYPES[type.to_sym]

      to_refresh.concat(Array(methods)) if methods
318
    end
319

320
    expire_method_caches(to_refresh)
321

322
    to_refresh.each { |method| send(method) } # rubocop:disable GitlabSecurity/PublicSend
323
  end
324

325 326 327 328 329 330 331
  def expire_branch_cache(branch_name = nil)
    # When we push to the root branch we have to flush the cache for all other
    # branches as their statistics are based on the commits relative to the
    # root branch.
    if !branch_name || branch_name == root_ref
      branches.each do |branch|
        cache.expire(:"diverging_commit_counts_#{branch.name}")
332
        cache.expire(:"commit_count_#{branch.name}")
333 334 335 336 337
      end
    # In case a commit is pushed to a non-root branch we only have to flush the
    # cache for said branch.
    else
      cache.expire(:"diverging_commit_counts_#{branch_name}")
338
      cache.expire(:"commit_count_#{branch_name}")
339
    end
Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
340 341
  end

342
  def expire_root_ref_cache
343
    expire_method_caches(%i(root_ref))
344 345
  end

346 347
  # Expires the cache(s) used to determine if a repository is empty or not.
  def expire_emptiness_caches
348
    return unless empty?
349

350
    expire_method_caches(%i(empty? has_visible_content?))
351 352
  end

353 354 355 356
  def lookup_cache
    @lookup_cache ||= {}
  end

357
  def expire_exists_cache
358
    expire_method_caches(%i(exists?))
359 360
  end

361 362 363 364 365 366 367
  # expire cache that doesn't depend on repository data (when expiring)
  def expire_content_cache
    expire_tags_cache
    expire_branches_cache
    expire_root_ref_cache
    expire_emptiness_caches
    expire_exists_cache
368
    expire_statistics_caches
369 370 371 372 373
  end

  # Runs code after a repository has been created.
  def after_create
    expire_exists_cache
374 375
    expire_root_ref_cache
    expire_emptiness_caches
Yorick Peterse's avatar
Yorick Peterse committed
376 377

    repository_event(:create_repository)
378 379
  end

380 381
  # Runs code just before a repository is deleted.
  def before_delete
382
    expire_exists_cache
383 384
    expire_all_method_caches
    expire_branch_cache if exists?
385
    expire_content_cache
Yorick Peterse's avatar
Yorick Peterse committed
386 387

    repository_event(:remove_repository)
388 389 390 391 392 393 394
  end

  # Runs code just before the HEAD of a repository is changed.
  def before_change_head
    # Cached divergent commit counts are based on repository head
    expire_branch_cache
    expire_root_ref_cache
Yorick Peterse's avatar
Yorick Peterse committed
395 396

    repository_event(:change_default_branch)
397 398
  end

399 400
  # Runs code before pushing (= creating or removing) a tag.
  def before_push_tag
401 402
    expire_statistics_caches
    expire_emptiness_caches
403
    expire_tags_cache
Yorick Peterse's avatar
Yorick Peterse committed
404 405

    repository_event(:push_tag)
406 407 408 409 410
  end

  # Runs code before removing a tag.
  def before_remove_tag
    expire_tags_cache
411
    expire_statistics_caches
Yorick Peterse's avatar
Yorick Peterse committed
412 413

    repository_event(:remove_tag)
414 415
  end

Lin Jen-Shin's avatar
Lin Jen-Shin committed
416 417 418 419 420
  # Runs code after removing a tag.
  def after_remove_tag
    expire_tags_cache
  end

421 422 423
  # Runs code after the HEAD of a repository is changed.
  def after_change_head
    expire_method_caches(METHOD_CACHES_FOR_FILE_TYPES.keys)
424 425
  end

426 427
  # Runs code after a repository has been forked/imported.
  def after_import
428
    expire_content_cache
429 430 431
  end

  # Runs code after a new commit has been pushed.
432 433 434
  def after_push_commit(branch_name)
    expire_statistics_caches
    expire_branch_cache(branch_name)
Yorick Peterse's avatar
Yorick Peterse committed
435 436

    repository_event(:push_commit, branch: branch_name)
437 438 439 440
  end

  # Runs code after a new branch has been created.
  def after_create_branch
441
    expire_branches_cache
Yorick Peterse's avatar
Yorick Peterse committed
442 443

    repository_event(:push_branch)
444 445
  end

446 447 448
  # Runs code before removing an existing branch.
  def before_remove_branch
    expire_branches_cache
Yorick Peterse's avatar
Yorick Peterse committed
449 450

    repository_event(:remove_branch)
451 452
  end

453 454
  # Runs code after an existing branch has been removed.
  def after_remove_branch
455
    expire_branches_cache
456 457
  end

458
  def method_missing(m, *args, &block)
459 460
    if m == :lookup && !block_given?
      lookup_cache[m] ||= {}
461
      lookup_cache[m][args.join(":")] ||= raw_repository.__send__(m, *args, &block) # rubocop:disable GitlabSecurity/PublicSend
462
    else
463
      raw_repository.__send__(m, *args, &block) # rubocop:disable GitlabSecurity/PublicSend
464
    end
465 466
  end

467 468
  def respond_to_missing?(method, include_private = false)
    raw_repository.respond_to?(method, include_private) || super
469
  end
Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
470 471

  def blob_at(sha, path)
472
    Blob.decorate(raw_repository.blob_at(sha, path), project)
Douwe Maan's avatar
Douwe Maan committed
473 474
  rescue Gitlab::Git::Repository::NoRepository
    nil
Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
475
  end
476

477 478 479 480 481
  # items is an Array like: [[oid, path], [oid1, path1]]
  def blobs_at(items)
    raw_repository.batch_blobs(items).map { |blob| Blob.decorate(blob, project) }
  end

482 483 484 485 486 487 488
  def root_ref
    if raw_repository
      raw_repository.root_ref
    else
      # When the repo does not exist we raise this error so no data is cached.
      raise Rugged::ReferenceError
    end
489
  end
490
  cache_method :root_ref
491

492
  # Gitaly migration: https://gitlab.com/gitlab-org/gitaly/issues/314
493
  def exists?
494
    return false unless full_path
495

496
    raw_repository.exists?
497 498 499
  end
  cache_method :exists?

Douwe Maan's avatar
Douwe Maan committed
500
  delegate :empty?, to: :raw_repository
501 502 503 504 505 506 507 508 509 510 511 512 513
  cache_method :empty?

  # The size of this repository in megabytes.
  def size
    exists? ? raw_repository.size : 0.0
  end
  cache_method :size, fallback: 0.0

  def commit_count
    root_ref ? raw_repository.commit_count(root_ref) : 0
  end
  cache_method :commit_count, fallback: 0

514
  def commit_count_for_ref(ref)
515
    return 0 unless exists?
516

517 518 519 520
    begin
      cache.fetch(:"commit_count_#{ref}") { raw_repository.commit_count(ref) }
    rescue Rugged::ReferenceError
      0
521 522 523
    end
  end

524
  delegate :branch_names, to: :raw_repository
525 526
  cache_method :branch_names, fallback: []

Douwe Maan's avatar
Douwe Maan committed
527
  delegate :tag_names, to: :raw_repository
528 529
  cache_method :tag_names, fallback: []

530
  delegate :branch_count, :tag_count, :has_visible_content?, to: :raw_repository
531 532
  cache_method :branch_count, fallback: 0
  cache_method :tag_count, fallback: 0
533
  cache_method :has_visible_content?, fallback: false
534 535

  def avatar
536 537 538 539 540
    # n+1: https://gitlab.com/gitlab-org/gitlab-ce/issues/38327
    Gitlab::GitalyClient.allow_n_plus_1_calls do
      if tree = file_on_head(:avatar)
        tree.path
      end
541 542
    end
  end
543
  cache_method :avatar
544

545 546 547 548 549 550 551 552 553 554
  def issue_template_names
    Gitlab::Template::IssueTemplate.dropdown_names(project)
  end
  cache_method :issue_template_names, fallback: []

  def merge_request_template_names
    Gitlab::Template::MergeRequestTemplate.dropdown_names(project)
  end
  cache_method :merge_request_template_names, fallback: []

555
  def readme
556 557
    if readme = tree(:head)&.readme
      ReadmeBlob.new(readme, self)
558
    end
559 560
  end

561
  def rendered_readme
562
    MarkupHelper.markup_unsafe(readme.name, readme.data, project: project) if readme
563 564
  end
  cache_method :rendered_readme
565

566
  def contribution_guide
567
    file_on_head(:contributing)
568
  end
569
  cache_method :contribution_guide
570 571

  def changelog
572
    file_on_head(:changelog)
573
  end
574
  cache_method :changelog
575

576
  def license_blob
577
    file_on_head(:license)
578
  end
579
  cache_method :license_blob
580

581
  def license_key
582
    return unless exists?
583

584
    Licensee.license(path).try(:key)
585
  end
586
  cache_method :license_key
587

588 589
  def license
    return unless license_key
590

591
    Licensee::License.new(license_key)
592
  end
593
  cache_method :license, memoize_only: true
594 595

  def gitignore
596
    file_on_head(:gitignore)
597
  end
598
  cache_method :gitignore
599 600

  def koding_yml
601
    file_on_head(:koding)
602
  end
603
  cache_method :koding_yml
604

605
  def gitlab_ci_yml
606
    file_on_head(:gitlab_ci)
607
  end
608
  cache_method :gitlab_ci_yml
609

610
  def head_commit
611 612 613 614
    @head_commit ||= commit(self.root_ref)
  end

  def head_tree
615 616 617
    if head_commit
      @head_tree ||= Tree.new(self, head_commit.sha, nil)
    end
618 619
  end

620
  def tree(sha = :head, path = nil, recursive: false)
621
    if sha == :head
622 623
      return unless head_commit

624 625 626 627 628
      if path.nil?
        return head_tree
      else
        sha = head_commit.sha
      end
629 630
    end

631
    Tree.new(self, sha, path, recursive: recursive)
632
  end
633 634

  def blob_at_branch(branch_name, path)
635
    last_commit = commit(branch_name)
636

637 638 639 640 641
    if last_commit
      blob_at(last_commit.sha, path)
    else
      nil
    end
642
  end
643

644
  def last_commit_for_path(sha, path)
645 646 647 648 649 650 651
    raw_repository.gitaly_migrate(:last_commit_for_path) do |is_enabled|
      if is_enabled
        last_commit_for_path_by_gitaly(sha, path)
      else
        last_commit_for_path_by_rugged(sha, path)
      end
    end
652
  end
653

Hiroyuki Sato's avatar
Hiroyuki Sato committed
654 655
  def last_commit_id_for_path(sha, path)
    key = path.blank? ? "last_commit_id_for_path:#{sha}" : "last_commit_id_for_path:#{sha}:#{Digest::SHA1.hexdigest(path)}"
Hiroyuki Sato's avatar
Hiroyuki Sato committed
656

657
    cache.fetch(key) do
658 659 660 661 662 663 664
      raw_repository.gitaly_migrate(:last_commit_for_path) do |is_enabled|
        if is_enabled
          last_commit_for_path_by_gitaly(sha, path).id
        else
          last_commit_id_for_path_by_shelling_out(sha, path)
        end
      end
665 666 667
    end
  end

668
  def next_branch(name, opts = {})
669 670
    branch_ids = self.branch_names.map do |n|
      next 1 if n == name
671

672
      result = n.match(/\A#{name}-([0-9]+)\z/)
673 674 675
      result[1].to_i if result
    end.compact

676
    highest_branch_id = branch_ids.max || 0
677

678 679 680
    return name if opts[:mild] && 0 == highest_branch_id

    "#{name}-#{highest_branch_id + 1}"
681 682
  end

683
  def branches_sorted_by(value)
684
    raw_repository.local_branches(sort_by: value)
685
  end
686

687 688 689
  def tags_sorted_by(value)
    case value
    when 'name'
690
      VersionSorter.rsort(tags) { |tag| tag.name }
691 692 693 694 695 696 697 698 699
    when 'updated_desc'
      tags_sorted_by_committed_date.reverse
    when 'updated_asc'
      tags_sorted_by_committed_date
    else
      tags
    end
  end

700
  def contributors
701
    commits = self.commits(nil, limit: 2000, offset: 0, skip_merges: true)
702

Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
703
    commits.group_by(&:author_email).map do |email, commits|
704 705
      contributor = Gitlab::Contributor.new
      contributor.email = email
706

Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
707
      commits.each do |commit|
708
        if contributor.name.blank?
Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
709
          contributor.name = commit.author_name
710 711
        end

712
        contributor.commits += 1
713 714
      end

715 716
      contributor
    end
717
  end
Dmitriy Zaporozhets's avatar
Dmitriy Zaporozhets committed
718

719
  def refs_contains_sha(ref_type, sha)
720 721
    args = %W(#{ref_type} --contains #{sha})
    names = run_git(args).first
722 723 724 725 726 727 728 729 730 731 732 733 734

    if names.respond_to?(:split)
      names = names.split("\n").map(&:strip)

      names.each do |name|
        name.slice! '* '
      end

      names
    else
      []
    end
  end
735

736 737 738
  def branch_names_contains(sha)
    refs_contains_sha('branch', sha)
  end
739

740 741
  def tag_names_contains(sha)
    refs_contains_sha('tag', sha)
742
  end
743

744
  def local_branches
745
    @local_branches ||= raw_repository.local_branches
746 747
  end

748 749
  alias_method :branches, :local_branches

750 751 752 753
  def tags
    @tags ||= raw_repository.tags
  end

Douwe Maan's avatar
Douwe Maan committed
754 755 756
  def create_dir(user, path, **options)
    options[:user] = user
    options[:actions] = [{ action: :create_dir, file_path: path }]
757

Douwe Maan's avatar
Douwe Maan committed
758
    multi_action(**options)
Stan Hu's avatar
Stan Hu committed
759 760
  end

Douwe Maan's avatar
Douwe Maan committed
761 762 763
  def create_file(user, path, content, **options)
    options[:user] = user
    options[:actions] = [{ action: :create, file_path: path, content: content }]
764

Douwe Maan's avatar
Douwe Maan committed
765
    multi_action(**options)
Stan Hu's avatar
Stan Hu committed
766
  end
767

Douwe Maan's avatar
Douwe Maan committed
768 769 770
  def update_file(user, path, content, **options)
    previous_path = options.delete(:previous_path)
    action = previous_path && previous_path != path ? :move : :update
771

Douwe Maan's avatar
Douwe Maan committed
772 773
    options[:user] = user
    options[:actions] = [{ action: action, file_path: path, previous_path: previous_path, content: content }]
774

Douwe Maan's avatar
Douwe Maan committed
775
    multi_action(**options)
776 777
  end

Douwe Maan's avatar
Douwe Maan committed
778 779 780
  def delete_file(user, path, **options)
    options[:user] = user
    options[:actions] = [{ action: :delete, file_path: path }]
781

Douwe Maan's avatar
Douwe Maan committed
782
    multi_action(**options)
783 784
  end

785 786
  def with_cache_hooks
    result = yield
787

788
    return unless result
789

790 791
    after_create if result.repo_created?
    after_create_branch if result.branch_created?
792

793 794 795 796 797 798 799 800 801
    result.newrev
  end

  def with_branch(user, *args)
    with_cache_hooks do
      Gitlab::Git::OperationService.new(user, raw_repository).with_branch(*args) do |start_commit|
        yield start_commit
      end
    end
802 803
  end

804
  # rubocop:disable Metrics/ParameterLists
Lin Jen-Shin's avatar
Lin Jen-Shin committed
805
  def multi_action(
806
    user:, branch_name:, message:, actions:,
807
    author_email: nil, author_name: nil,
808
    start_branch_name: nil, start_project: project)
809

810 811
    with_branch(
      user,
812
      branch_name,
813
      start_branch_name: start_branch_name,
814
      start_repository: start_project.repository.raw_repository) do |start_commit|
815

816
      index = Gitlab::Git::Index.new(raw_repository)
817

818
      if start_commit
819
        index.read_tree(start_commit.rugged_commit.tree)
820 821 822
        parents = [start_commit.sha]
      else
        parents = []
Marc Siegfriedt's avatar
Marc Siegfriedt committed
823 824
      end

825
      actions.each do |options|
826
        index.public_send(options.delete(:action), options) # rubocop:disable GitlabSecurity/PublicSend
Marc Siegfriedt's avatar
Marc Siegfriedt committed
827 828 829
      end

      options = {
830
        tree: index.write_tree,
Marc Siegfriedt's avatar
Marc Siegfriedt committed
831 832 833 834 835
        message: message,
        parents: parents
      }
      options.merge!(get_committer_and_author(user, email: author_email, name: author_name))

836
      create_commit(options)
Marc Siegfriedt's avatar
Marc Siegfriedt committed
837 838
    end
  end
839
  # rubocop:enable Metrics/ParameterLists
Marc Siegfriedt's avatar
Marc Siegfriedt committed
840

841 842
  def get_committer_and_author(user, email: nil, name: nil)
    committer = user_to_committer(user)
Douwe Maan's avatar
Douwe Maan committed
843
    author = Gitlab::Git.committer_hash(email: email, name: name) || committer
844

845
    {
846 847
      author: author,
      committer: committer
848 849 850 851 852 853 854 855 856 857 858 859 860 861
    }
  end

  def can_be_merged?(source_sha, target_branch)
    our_commit = rugged.branches[target_branch].target
    their_commit = rugged.lookup(source_sha)

    if our_commit && their_commit
      !rugged.merge_commits(our_commit, their_commit).conflicts?
    else
      false
    end
  end

862 863 864 865 866 867
  def merge(user, source_sha, merge_request, message)
    with_cache_hooks do
      raw_repository.merge(user, source_sha, merge_request.target_branch, message) do |commit_id|
        merge_request.update(in_progress_merge_commit_sha: commit_id)
        nil # Return value does not matter.
      end
868
    end
869 870
  end

871
  def ff_merge(user, source, target_branch, merge_request: nil)
872 873
    their_commit_id = commit(source)&.id
    raise 'Invalid merge source' if their_commit_id.nil?
874

875
    merge_request&.update(in_progress_merge_commit_sha: their_commit_id)
876

877
    with_cache_hooks { raw.ff_merge(user, their_commit_id, target_branch) }
878 879
  end

880
  def revert(
881
    user, commit, branch_name, message,
882
    start_branch_name: nil, start_project: project)
883

884 885 886 887 888 889 890 891 892
    with_cache_hooks do
      raw_repository.revert(
        user: user,
        commit: commit.raw,
        branch_name: branch_name,
        message: message,
        start_branch_name: start_branch_name,
        start_repository: start_project.repository.raw_repository
      )
893
    end
894 895
  end

896
  def cherry_pick(
897
    user, commit, branch_name, message,
898
    start_branch_name: nil, start_project: project)
899

900 901 902 903 904 905 906 907 908
    with_cache_hooks do
      raw_repository.cherry_pick(
        user: user,
        commit: commit.raw,
        branch_name: branch_name,
        message: message,
        start_branch_name: start_branch_name,
        start_repository: start_project.repository.raw_repository
      )
909 910 911
    end
  end

912
  def merged_to_root_ref?(branch_or_name)
913 914 915
    branch = Gitlab::Git::Branch.find(self, branch_or_name)

    if branch
916 917
      @root_ref_sha ||= commit(root_ref).sha
      same_head = branch.target == @root_ref_sha
918
      merged = ancestor?(branch.target, @root_ref_sha)
919
      !same_head && merged
920 921 922 923 924
    else
      nil
    end
  end

925 926
  delegate :merged_branch_names, to: :raw_repository

927
  def merge_base(first_commit_id, second_commit_id)
928 929
    first_commit_id = commit(first_commit_id).try(:id) || first_commit_id
    second_commit_id = commit(second_commit_id).try(:id) || second_commit_id
930
    rugged.merge_base(first_commit_id, second_commit_id)
Douwe Maan's avatar
Douwe Maan committed
931 932
  rescue Rugged::ReferenceError
    nil
933 934
  end

935
  def ancestor?(ancestor_id, descendant_id)
936
    return false if ancestor_id.nil? || descendant_id.nil?
937

938 939
    Gitlab::GitalyClient.migrate(:is_ancestor) do |is_enabled|
      if is_enabled
940
        raw_repository.ancestor?(ancestor_id, descendant_id)
941
      else
942
        rugged_is_ancestor?(ancestor_id, descendant_id)
943 944
      end
    end
945 946
  end

947 948 949
  def empty_repo?
    !exists? || !has_visible_content?
  end
950
  cache_method :empty_repo?, memoize_only: true
951 952 953

  def search_files_by_content(query, ref)
    return [] if empty_repo? || query.blank?
Valery Sizov's avatar
Valery Sizov committed
954

955
    offset = 2
956 957 958
    args = %W(grep -i -I -n --before-context #{offset} --after-context #{offset} -E -e #{Regexp.escape(query)} #{ref || root_ref})

    run_git(args).first.scrub.split(/^--$/)
959 960
  end

961 962
  def search_files_by_name(query, ref)
    return [] if empty_repo? || query.blank?
963

964 965 966
    args = %W(ls-tree --full-tree -r #{ref || root_ref} --name-status | #{Regexp.escape(query)})

    run_git(args).first.lines.map(&:strip)
967 968
  end

Douwe Maan's avatar
Douwe Maan committed
969
  def fetch_as_mirror(url, forced: false, refmap: :all_refs, remote_name: nil)
970 971 972 973 974 975
    unless remote_name
      remote_name = "tmp-#{SecureRandom.hex}"
      tmp_remote_name = true
    end

    add_remote(remote_name, url)
Douwe Maan's avatar
Douwe Maan committed
976
    set_remote_as_mirror(remote_name, refmap: refmap)
977 978 979 980 981
    fetch_remote(remote_name, forced: forced)
  ensure
    remove_remote(remote_name) if tmp_remote_name
  end

982 983
  def fetch_remote(remote, forced: false, ssh_auth: nil, no_tags: false)
    gitlab_shell.fetch_remote(raw_repository, remote, ssh_auth: ssh_auth, forced: forced, no_tags: no_tags)
984 985
  end

986 987
  def fetch_source_branch!(source_repository, source_branch, local_ref)
    raw_repository.fetch_source_branch!(source_repository.raw_repository, source_branch, local_ref)
988
  end
989

990 991 992 993
  def remote_exists?(name)
    raw_repository.remote_exists?(name)
  end

994 995
  def compare_source_branch(target_branch_name, source_repository, source_branch_name, straight:)
    raw_repository.compare_source_branch(target_branch_name, source_repository.raw_repository, source_branch_name, straight: straight)
996
  end
997

998
  def create_ref(ref, ref_path)
999
    raw_repository.write_ref(ref_path, ref)
1000 1001
  end

1002 1003 1004 1005 1006
  def ls_files(ref)
    actual_ref = ref || root_ref
    raw_repository.ls_files(actual_ref)
  end

1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
  def copy_gitattributes(ref)
    actual_ref = ref || root_ref
    begin
      raw_repository.copy_gitattributes(actual_ref)
      true
    rescue Gitlab::Git::Repository::InvalidRef
      false
    end
  end

1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
  # Caches the supplied block both in a cache and in an instance variable.
  #
  # The cache key and instance variable are named the same way as the value of
  # the `key` argument.
  #
  # This method will return `nil` if the corresponding instance variable is also
  # set to `nil`. This ensures we don't keep yielding the block when it returns
  # `nil`.
  #
  # key - The name of the key to cache the data in.
  # fallback - A value to fall back to in the event of a Git error.
1028
  def cache_method_output(key, fallback: nil, memoize_only: false, &block)
1029
    ivar = cache_instance_variable_name(key)
1030

1031 1032 1033
    if instance_variable_defined?(ivar)
      instance_variable_get(ivar)
    else
1034 1035 1036 1037
      # If the repository doesn't exist and a fallback was specified we return
      # that value inmediately. This saves us Rugged/gRPC invocations.
      return fallback unless fallback.nil? || exists?

1038
      begin
1039 1040 1041 1042 1043 1044 1045
        value =
          if memoize_only
            yield
          else
            cache.fetch(key, &block)
          end
        instance_variable_set(ivar, value)
1046
      rescue Rugged::ReferenceError, Gitlab::Git::Repository::NoRepository
1047 1048 1049
        # Even if the above `#exists?` check passes these errors might still
        # occur (for example because of a non-existing HEAD). We want to
        # gracefully handle this and not cache anything
1050
        fallback
1051 1052 1053
      end
    end
  end
1054

1055 1056 1057
  def cache_instance_variable_name(key)
    :"@#{key.to_s.tr('?!', '')}"
  end
1058

1059 1060
  def file_on_head(type)
    if head = tree(:head)
1061 1062
      head.blobs.find do |blob|
        Gitlab::FileDetector.type_of(blob.path) == type
1063 1064 1065 1066
      end
    end
  end

Douwe Maan's avatar
Douwe Maan committed
1067 1068 1069 1070
  def route_map_for(sha)
    blob_data_at(sha, '.gitlab/route-map.yml')
  end

1071 1072
  def gitlab_ci_yml_for(sha, path = '.gitlab-ci.yml')
    blob_data_at(sha, path)
Douwe Maan's avatar
Douwe Maan committed
1073 1074
  end

1075 1076 1077 1078
  def fetch_ref(source_repository, source_ref:, target_ref:)
    raw_repository.fetch_ref(source_repository.raw_repository, source_ref: source_ref, target_ref: target_ref)
  end

1079 1080 1081 1082
  def repository_storage_path
    @project.repository_storage_path
  end

1083 1084
  private

1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
  # TODO Generice finder, later split this on finders by Ref or Oid
  # gitlab-org/gitlab-ce#39239
  def find_commit(oid_or_ref)
    commit = if oid_or_ref.is_a?(Gitlab::Git::Commit)
               oid_or_ref
             else
               Gitlab::Git::Commit.find(raw_repository, oid_or_ref)
             end

    ::Commit.new(commit, @project) if commit
  end

Douwe Maan's avatar
Douwe Maan committed
1097 1098
  def blob_data_at(sha, path)
    blob = blob_at(sha, path)
1099
    return unless blob
1100

1101
    blob.load_all_data!
1102
    blob.data
1103
  end
1104

1105
  def cache
1106 1107
    # TODO: should we use UUIDs here? We could move repositories without clearing this cache
    @cache ||= RepositoryCache.new(full_path, @project.id)
1108
  end
1109 1110

  def tags_sorted_by_committed_date
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
    tags.sort_by do |tag|
      # Annotated tags can point to any object (e.g. a blob), but generally
      # tags point to a commit. If we don't have a commit, then just default
      # to putting the tag at the end of the list.
      target = tag.dereferenced_target

      if target
        target.committed_date
      else
        Time.now
      end
    end
1123
  end
1124 1125

  def keep_around_ref_name(sha)
1126
    "refs/#{REF_KEEP_AROUND}/#{sha}"
1127
  end
Yorick Peterse's avatar
Yorick Peterse committed
1128 1129

  def repository_event(event, tags = {})
1130
    Gitlab::Metrics.add_event(event, { path: full_path }.merge(tags))
Yorick Peterse's avatar
Yorick Peterse committed
1131
  end
1132

1133 1134
  def last_commit_for_path_by_gitaly(sha, path)
    c = raw_repository.gitaly_commit_client.last_commit_for_path(sha, path)
1135
    commit_by(oid: c)
1136 1137 1138 1139
  end

  def last_commit_for_path_by_rugged(sha, path)
    sha = last_commit_id_for_path_by_shelling_out(sha, path)
1140
    commit_by(oid: sha)
1141 1142 1143
  end

  def last_commit_id_for_path_by_shelling_out(sha, path)
1144
    args = %W(rev-list --max-count=1 #{sha} -- #{path})
Andrew Newdigate's avatar
Andrew Newdigate committed
1145
    raw_repository.run_git_with_timeout(args, Gitlab::Git::Popen::FAST_GIT_PROCESS_TIMEOUT).first.strip
1146 1147
  end

1148
  def initialize_raw_repository
1149
    Gitlab::Git::Repository.new(project.repository_storage, disk_path + '.git', Gitlab::GlRepository.gl_repository(project, is_wiki))
1150
  end
1151

1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171
  def find_commits_by_message_by_shelling_out(query, ref, path, limit, offset)
    ref ||= root_ref

    args = %W(
      log #{ref} --pretty=%H --skip #{offset}
      --max-count #{limit} --grep=#{query} --regexp-ignore-case
    )
    args = args.concat(%W(-- #{path})) if path.present?

    git_log_results = run_git(args).first.lines

    git_log_results.map { |c| commit(c.chomp) }.compact
  end

  def find_commits_by_message_by_gitaly(query, ref, path, limit, offset)
    raw_repository
      .gitaly_commit_client
      .commits_by_message(query, revision: ref, path: path, limit: limit, offset: offset)
      .map { |c| commit(c) }
  end
1172
end