Package com.google.gerrit.server.notedb

Class CommitRewriter

java.lang.Object

com.google.gerrit.server.notedb.CommitRewriter

public class CommitRewriter
extends Object

Rewrites ('backfills') commit history of change in NoteDb to not contain user data. Only fixes known cases, rewriting commits case by case.

The cases where we used to put user data in NoteDb can be found by https://gerrit-review.googlesource.com/q/hashtag:user-data-cleanup

As opposed to NoteDbRewriter implementations, which target a specific change and are used by REST endpoints, this rewriter is used as standalone tool, that bulk backfills changes by project.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	CommitRewriter.BackfillResult	Result of the backfill run for a project.
static class	CommitRewriter.CommitDiff	Diff result of a single commit rewrite
static class	CommitRewriter.RunOptions	Options to run backfillProject(com.google.gerrit.entities.Project.NameKey, org.eclipse.jgit.lib.Repository, com.google.gerrit.server.notedb.CommitRewriter.RunOptions).

Field Summary

Fields

Modifier and Type	Field	Description
static String	DEFAULT_ACCOUNT_REPLACEMENT

Method Summary

Modifier and Type	Method	Description
com.google.gerrit.server.notedb.CommitRewriter.ChangeFixProgress	backfillChange(com.google.gerrit.server.notedb.CommitRewriter.RefsUpdate refsUpdate, org.eclipse.jgit.lib.Ref ref, com.google.common.collect.ImmutableSet<AccountState> accountsInChange, CommitRewriter.RunOptions options)	Walks the ref history from oldest update to the most recent update, fixing the commits that contain user data case by case.
CommitRewriter.BackfillResult	backfillProject(Project.NameKey project, org.eclipse.jgit.lib.Repository repo, CommitRewriter.RunOptions options)	Rewrites commit history of RefNames.changeMetaRef(com.google.gerrit.entities.Change.Id)s in single repo.
static byte[]	cutTreeAndParents(byte[] b)	Cuts tree and parent lines from raw unparsed commit body, so they are not included in diff comparison.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

DEFAULT_ACCOUNT_REPLACEMENT

public static final String DEFAULT_ACCOUNT_REPLACEMENT

See Also:

Constant Field Values

Method Detail

backfillProject

public CommitRewriter.BackfillResult backfillProject(Project.NameKey project,
                                                     org.eclipse.jgit.lib.Repository repo,
                                                     CommitRewriter.RunOptions options)

Rewrites commit history of RefNames.changeMetaRef(com.google.gerrit.entities.Change.Id)s in single repo. Only rewrites branch if necessary, i.e. if there were any commits that contained user data.

See CommitRewriter.RunOptions for the execution and output options.

Parameters:

project - project to backfill

repo - repo to backfill

options - CommitRewriter.RunOptions to control how the run is executed.

Returns:

BackfillResult

backfillChange

public com.google.gerrit.server.notedb.CommitRewriter.ChangeFixProgress backfillChange(com.google.gerrit.server.notedb.CommitRewriter.RefsUpdate refsUpdate,
                                                                                       org.eclipse.jgit.lib.Ref ref,
                                                                                       com.google.common.collect.ImmutableSet<AccountState> accountsInChange,
                                                                                       CommitRewriter.RunOptions options)
                                                                                throws IOException,
                                                                                       org.eclipse.jgit.errors.ConfigInvalidException

Walks the ref history from oldest update to the most recent update, fixing the commits that contain user data case by case. Commit history is rewritten from the first commit, that needs to be updated, for all subsequent updates. The new ref tip is returned in CommitRewriter.ChangeFixProgress.newTipId.

Throws:

IOException

org.eclipse.jgit.errors.ConfigInvalidException

cutTreeAndParents

public static byte[] cutTreeAndParents(byte[] b)

Cuts tree and parent lines from raw unparsed commit body, so they are not included in diff comparison.

Parameters:

b - raw unparsed commit body, see RevCommit.getRawBuffer().

For parsing, see RawParseUtils.author(byte[], int), RawParseUtils.commitMessage(byte[], int), etc.

Returns:

raw unparsed commit body, without tree and parent lines.