Files
gitea/services/repository/migrate.go
T
silverwindandbircni fc49368845 feat: Improve Gitea-compatible forge migrations, fix many migration bugs (#39424)
Supports migrating from Gitea-compatible forges like Forgejo and fixes
many migration bugs.

1. Pre-release and `+gitea-` suffixed server versions no longer drop
reviews, reactions and pagination.
2. Review states, dismissals, scoped and archived labels and merge bases
are migrated.
3. Disabled units, external trackers, unavailable release assets, and
GitLab confidential issues and internal notes are skipped.
4. Credentials are only sent to the source host, and `Retry-After` rate
limits are waited out.
5. GitHub migrations of only issues or pull requests, basic auth clones
and dismissed reviews work again.
6. The API returns 404 for releases of a disabled unit and shows
migrated authors as mannequins, replacing the never populated
`original_author` and `original_author_id` fields.

Fixes: https://github.com/go-gitea/gitea/issues/19533
Fixes: https://github.com/go-gitea/gitea/issues/35885
Fixes: https://github.com/go-gitea/gitea/issues/18275
Fixes: https://github.com/go-gitea/gitea/issues/35069
Fixes: https://github.com/go-gitea/gitea/issues/17763
Fixes: https://github.com/go-gitea/gitea/issues/39548
Co-authored-by: bircni <bircni@icloud.com>
2026-10-24 01:15:31 +00:00

289 lines
9.3 KiB
Go

// Copyright 2024 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package repository
import (
"context"
"errors"
"fmt"
"net/http"
"time"
"gitea.dev/models/db"
"gitea.dev/models/organization"
repo_model "gitea.dev/models/repo"
unit_model "gitea.dev/models/unit"
user_model "gitea.dev/models/user"
"gitea.dev/modules/git"
"gitea.dev/modules/git/gitcmd"
"gitea.dev/modules/lfs"
"gitea.dev/modules/log"
"gitea.dev/modules/migration"
repo_module "gitea.dev/modules/repository"
"gitea.dev/modules/setting"
"gitea.dev/modules/timeutil"
"gitea.dev/modules/util"
)
func cloneWiki(ctx context.Context, repo *repo_model.Repository, opts migration.MigrateOptions, migrateTimeout time.Duration) (string, error) {
wikiRemoteURL := repo_module.WikiRemoteURL(ctx, opts.CloneAddr)
if wikiRemoteURL == "" {
return "", nil
}
storageRepo := repo.WikiStorageRepo()
if err := git.DeleteRepository(ctx, storageRepo); err != nil {
return "", fmt.Errorf("failed to remove existing wiki for repo %q, err: %w", repo.FullName(), err)
}
cleanIncompleteWikiPath := func() {
if err := git.DeleteRepository(ctx, storageRepo); err != nil {
log.Error("Failed to remove incomplete wiki for repo %q, err: %v", repo.FullName(), err)
}
}
if err := git.CloneExternalRepo(ctx, wikiRemoteURL, storageRepo, git.CloneRepoOptions{
Mirror: true,
Quiet: true,
Timeout: migrateTimeout,
SkipTLSVerify: setting.Migrations.SkipTLSVerify,
}); err != nil {
log.Error("Clone wiki failed, err: %v", err)
cleanIncompleteWikiPath()
return "", err
}
if err := git.WriteCommitGraph(ctx, storageRepo); err != nil {
cleanIncompleteWikiPath()
return "", err
}
defaultBranch, err := git.GetDefaultBranch(ctx, storageRepo)
if err != nil {
cleanIncompleteWikiPath()
return "", fmt.Errorf("failed to get wiki repo default branch for %q, err: %w", repo.FullName(), err)
}
return defaultBranch, nil
}
// MigrateRepositoryGitData starts migrating git related data after created migrating repository
func MigrateRepositoryGitData(ctx context.Context, u *user_model.User,
repo *repo_model.Repository, opts migration.MigrateOptions,
httpTransport http.RoundTripper,
) (*repo_model.Repository, error) {
if u.IsOrganization() {
t, err := organization.OrgFromUser(u).GetOwnerTeam(ctx)
if err != nil {
return nil, err
}
repo.NumWatches = t.NumMembers
} else {
repo.NumWatches = 1
}
migrateTimeout := time.Duration(setting.Git.Timeout.Migrate) * time.Second
if err := git.DeleteRepository(ctx, repo); err != nil {
return repo, fmt.Errorf("failed to remove existing repo dir %q, err: %w", repo.FullName(), err)
}
if err := git.CloneExternalRepo(ctx, opts.CloneAddr, repo, git.CloneRepoOptions{
Mirror: true,
Quiet: true,
Timeout: migrateTimeout,
SkipTLSVerify: setting.Migrations.SkipTLSVerify,
}); err != nil {
if errors.Is(err, context.DeadlineExceeded) {
return repo, fmt.Errorf("clone timed out, consider increasing [git.timeout] MIGRATE in app.ini, underlying err: %w", err)
}
return repo, fmt.Errorf("clone error: %w", err)
}
if err := git.WriteCommitGraph(ctx, repo); err != nil {
return repo, err
}
if opts.Wiki {
defaultWikiBranch, err := cloneWiki(ctx, repo, opts, migrateTimeout)
if err != nil {
return repo, fmt.Errorf("clone wiki error: %w", err)
}
repo.DefaultWikiBranch = defaultWikiBranch
}
if repo.OwnerID == u.ID {
repo.Owner = u
}
if err := updateGitRepoAfterCreate(ctx, repo); err != nil {
return nil, fmt.Errorf("updateGitRepoAfterCreate: %w", err)
}
gitRepo, err := git.OpenRepository(ctx, repo)
if err != nil {
return repo, fmt.Errorf("OpenRepository: %w", err)
}
defer gitRepo.Close()
repo.IsEmpty, err = gitRepo.IsEmpty(ctx)
if err != nil {
return repo, fmt.Errorf("git.IsEmpty: %w", err)
}
if !repo.IsEmpty {
if len(repo.DefaultBranch) == 0 {
// Try to get HEAD branch and set it as default branch.
headBranchName, err := git.GetDefaultBranch(ctx, repo)
if err != nil {
return repo, fmt.Errorf("GetHEADBranch: %w", err)
}
if headBranchName != "" {
repo.DefaultBranch = headBranchName
}
}
if _, _, err := repo_module.SyncRepoBranchesWithRepo(ctx, repo, gitRepo, u.ID); err != nil {
return repo, fmt.Errorf("SyncRepoBranchesWithRepo: %w", err)
}
// if releases migration are not requested, we will sync all tags here
// otherwise, the releases sync will be done out of this function
if !opts.Releases {
repo.IsMirror = opts.Mirror
if _, err = repo_module.SyncReleasesWithTags(ctx, repo, gitRepo); err != nil {
log.Error("Failed to synchronize tags to releases for repository: %v", err)
}
}
if opts.LFS {
lfsClient, err := lfs.NewClientFromEndpoint(opts.CloneAddr, opts.LFSEndpoint, httpTransport)
if err != nil {
return repo, fmt.Errorf("NewClientFromEndpoint: %w", err)
}
if err = repo_module.StoreMissingLfsObjectsInRepository(ctx, repo, gitRepo, lfsClient); err != nil {
log.Error("Failed to store missing LFS objects for repository: %v", err)
return repo, fmt.Errorf("StoreMissingLfsObjectsInRepository: %w", err)
}
}
// Update repo license
if err := AddRepoToLicenseUpdaterQueue(&LicenseUpdaterOptions{RepoID: repo.ID}); err != nil {
log.Error("Failed to add repo to license updater queue: %v", err)
}
}
return db.WithTx2(ctx, func(ctx context.Context) (*repo_model.Repository, error) {
if opts.Mirror {
remoteAddress, err := util.SanitizeURL(opts.CloneAddr)
if err != nil {
return repo, err
}
mirrorModel := repo_model.Mirror{
RepoID: repo.ID,
Interval: setting.Mirror.DefaultInterval,
EnablePrune: true,
NextUpdateUnix: timeutil.TimeStampNow().AddDuration(setting.Mirror.DefaultInterval),
LFS: opts.LFS,
RemoteAddress: remoteAddress,
}
if opts.LFS {
mirrorModel.LFSEndpoint = opts.LFSEndpoint
}
if opts.MirrorInterval != "" {
parsedInterval, err := time.ParseDuration(opts.MirrorInterval)
if err != nil {
log.Error("Failed to set Interval: %v", err)
return repo, err
}
if parsedInterval == 0 {
mirrorModel.Interval = 0
mirrorModel.NextUpdateUnix = 0
} else if parsedInterval < setting.Mirror.MinInterval {
err := fmt.Errorf("interval %s is set below Minimum Interval of %s", parsedInterval, setting.Mirror.MinInterval)
log.Error("Interval: %s is too frequent", opts.MirrorInterval)
return repo, err
} else {
mirrorModel.Interval = parsedInterval
mirrorModel.NextUpdateUnix = timeutil.TimeStampNow().AddDuration(parsedInterval)
}
}
if err = repo_model.InsertMirror(ctx, &mirrorModel); err != nil {
return repo, fmt.Errorf("InsertOne: %w", err)
}
repo.IsMirror = true
if err = repo_model.UpdateRepositoryColsNoAutoTime(ctx, repo, "num_watches", "is_empty", "default_branch", "default_wiki_branch", "is_mirror"); err != nil {
return nil, err
}
if err = repo_module.UpdateRepoSize(ctx, repo); err != nil {
log.Error("Failed to update size for repository: %v", err)
}
// this is necessary for sync local tags from remote
configName := fmt.Sprintf("remote.%s.fetch", mirrorModel.GetRemoteName())
stdout, _, err := gitcmd.NewCommand("config").
AddOptionValues("--add", configName, `+refs/tags/*:refs/tags/*`).
WithRepo(repo).RunStdString(ctx)
if err != nil {
log.Error("MigrateRepositoryGitData(git config --add <remote> +refs/tags/*:refs/tags/*) in %v: Stdout: %s\nError: %v", repo, stdout, err)
return repo, fmt.Errorf("error in MigrateRepositoryGitData(git config --add <remote> +refs/tags/*:refs/tags/*): %w", err)
}
} else {
if err = repo_module.UpdateRepoSize(ctx, repo); err != nil {
log.Error("Failed to update size for repository: %v", err)
}
if repo, err = CleanUpMigrateInfo(ctx, repo); err != nil {
return nil, err
}
}
var enableRepoUnits []repo_model.RepoUnit
if opts.Releases && !unit_model.TypeReleases.UnitGlobalDisabled() {
enableRepoUnits = append(enableRepoUnits, repo_model.RepoUnit{RepoID: repo.ID, Type: unit_model.TypeReleases})
}
if opts.Wiki && !unit_model.TypeWiki.UnitGlobalDisabled() {
enableRepoUnits = append(enableRepoUnits, repo_model.RepoUnit{RepoID: repo.ID, Type: unit_model.TypeWiki})
}
if len(enableRepoUnits) > 0 {
err = UpdateRepositoryUnits(ctx, repo, enableRepoUnits, nil)
if err != nil {
return nil, err
}
}
return repo, nil
})
}
// CleanUpMigrateInfo finishes migrating repository and/or wiki with things that don't need to be done for mirrors.
func CleanUpMigrateInfo(ctx context.Context, repo *repo_model.Repository) (*repo_model.Repository, error) {
if err := git.CreateDelegateHooks(ctx, repo); err != nil {
return repo, fmt.Errorf("createDelegateHooks: %w", err)
}
hasWiki := HasWiki(ctx, repo)
if hasWiki {
if err := git.CreateDelegateHooks(ctx, repo.WikiStorageRepo()); err != nil {
return repo, fmt.Errorf("createDelegateHooks.(wiki): %w", err)
}
}
err := git.ManagedRemoteRemove(ctx, repo, "origin")
if err != nil && !git.IsRemoteNotExistError(err) {
return repo, fmt.Errorf("CleanUpMigrateInfo: %w", err)
}
if hasWiki {
err = git.ManagedRemoteRemove(ctx, repo.WikiStorageRepo(), "origin")
if err != nil && !git.IsRemoteNotExistError(err) {
return repo, fmt.Errorf("cleanUpMigrateGitConfig (wiki): %w", err)
}
}
return repo, UpdateRepository(ctx, repo, false)
}