git-annex/Annex/Branch.hs

1085 lines
39 KiB
Haskell
Raw Normal View History

2011-06-21 20:08:09 +00:00
{- management of the git-annex branch
-
- Copyright 2011-2023 Joey Hess <id@joeyh.name>
-
- Licensed under the GNU AGPL version 3 or higher.
-}
{-# LANGUAGE OverloadedStrings #-}
2011-10-04 04:40:47 +00:00
module Annex.Branch (
fullname,
name,
hasOrigin,
hasSibling,
siblingBranches,
create,
getBranch,
UpdateMade(..),
2011-06-21 20:08:09 +00:00
update,
forceUpdate,
updateTo,
get,
getHistorical,
getUnmergedRefs,
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
RegardingUUID(..),
change,
ChangeOrAppend(..),
changeOrAppend,
maybeChange,
commitMessage,
createMessage,
2011-06-22 21:47:06 +00:00
commit,
forceCommit,
files,
rememberTreeish,
performTransitions,
withIndex,
precache,
2021-04-21 18:19:58 +00:00
overBranchFileContents,
updatedFromTree,
2011-06-21 20:08:09 +00:00
) where
import qualified Data.ByteString as B
Fix encoding of data written to git-annex branch. Avoid truncating unicode characters to 8 bits. Allow any encoding to be used, as with filenames (but utf8 is the sane choice). Affects metadata and repository descriptions, and preferred content expressions. The question of what's the right encoding for the git-annex branch is a vexing one. utf-8 would be a nice choice, but this leaves the possibility of bad data getting into a git-annex branch somehow, and this resulting in git-annex crashing with encoding errors, which is a failure mode I want to avoid. (Also, preferred content expressions can refer to filenames, and filenames can have any encoding, so limiting to utf-8 would not be ideal.) The union merge code already took care to not assume any encoding for a file. Except it assumes that any \n is a literal newline, and not part of some encoding of a character that happens to contain a newline. (At least utf-8 avoids using newline for anything except liternal newlines.) Adapted the git-annex branch code to use this same approach. Note that there is a potential interop problem with Windows, since FileSystemEncoding doesn't work there, and instead things are always decoded as utf-8. If someone uses non-utf8 encoding for data on the git-annex branch, this can lead to an encoding error on windows. However, this commit doesn't actually make that any worse, because the union merge code would similarly fail with an encoding error on windows in that situation. This commit was sponsored by Kyle Meyer.
2014-05-27 18:16:33 +00:00
import qualified Data.ByteString.Lazy as L
import qualified Data.Set as S
import qualified Data.Map as M
import Data.Function
import Data.Char
import Data.ByteString.Builder
import Control.Concurrent (threadDelay)
import Control.Concurrent.MVar
import qualified System.FilePath.ByteString as P
import System.PosixCompat.Files (isRegularFile)
import Annex.Common
import Types.BranchState
2011-12-12 21:38:46 +00:00
import Annex.BranchState
2011-12-12 22:03:28 +00:00
import Annex.Journal
import Annex.GitOverlay
import Annex.Tmp
import qualified Git
2011-12-14 19:56:11 +00:00
import qualified Git.Command
2011-12-12 22:23:24 +00:00
import qualified Git.Ref
import qualified Git.RefLog
import qualified Git.Sha
import qualified Git.Branch
2011-12-13 01:24:55 +00:00
import qualified Git.UnionMerge
import qualified Git.UpdateIndex
import qualified Git.Tree
import qualified Git.LsTree
import Git.LsTree (lsTreeParams)
2016-03-14 20:29:19 +00:00
import qualified Git.HashObject
import Annex.HashObject
import Git.Types (Ref(..), fromRef, fromRef', RefDate, TreeItemType(..))
import Git.FilePath
2011-10-04 04:40:47 +00:00
import Annex.CatFile
2021-04-21 18:19:58 +00:00
import Git.CatFile (catObjectStreamLsTree)
import Annex.Perms
import Logs
import Logs.Transitions
import Logs.File
import Logs.Trust.Pure
import Logs.Remote.Pure
import Logs.Export.Pure
import Logs.Difference.Pure
import qualified Annex.Queue
import Types.Transitions
import Annex.Branch.Transitions
import qualified Annex
import Annex.Hook
import Utility.Directory.Stream
import Utility.Tmp
import qualified Utility.RawFilePath as R
{- Name of the branch that is used to store git-annex's information. -}
name :: Git.Ref
name = Git.Ref "git-annex"
{- Fully qualified name of the branch. -}
fullname :: Git.Ref
fullname = Git.Ref $ "refs/heads/" <> fromRef' name
{- Branch's name in origin. -}
originname :: Git.Ref
originname = Git.Ref $ "refs/remotes/origin/" <> fromRef' name
{- Does origin/git-annex exist? -}
hasOrigin :: Annex Bool
hasOrigin = inRepo $ Git.Ref.exists originname
{- Does the git-annex branch or a sibling foo/git-annex branch exist? -}
hasSibling :: Annex Bool
hasSibling = not . null <$> siblingBranches
{- List of git-annex (shas, branches), including the main one and any
- from remotes. Duplicates are filtered out. -}
siblingBranches :: Annex [(Git.Sha, Git.Branch)]
siblingBranches = inRepo $ Git.Ref.matchingUniq [name]
{- Creates the branch, if it does not already exist. -}
create :: Annex ()
2012-06-12 15:32:06 +00:00
create = void getBranch
{- Returns the sha of the branch, creating it first if necessary. -}
2012-01-10 19:36:54 +00:00
getBranch :: Annex Git.Ref
getBranch = maybe (hasOrigin >>= go >>= use) return =<< branchsha
2012-12-13 04:24:19 +00:00
where
go True = do
inRepo $ Git.Command.run
[ Param "branch"
, Param "--no-track"
, Param $ fromRef name
, Param $ fromRef originname
]
fromMaybe (giveup $ "failed to create " ++ fromRef name)
2012-12-13 04:24:19 +00:00
<$> branchsha
go False = withIndex' True $ do
-- Create the index file. This is not necessary,
-- except to avoid a bug in git 2.37 that causes
-- git write-tree to segfault when the index file does not
-- exist.
inRepo $ flip Git.UpdateIndex.streamUpdateIndex []
cmode <- annexCommitMode <$> Annex.getGitConfig
cmessage <- createMessage
inRepo $ Git.Branch.commitAlways cmode cmessage fullname []
2012-12-13 04:24:19 +00:00
use sha = do
setIndexSha sha
return sha
branchsha = inRepo $ Git.Ref.sha fullname
2012-09-15 19:40:13 +00:00
{- Ensures that the branch and index are up-to-date; should be
- called before data is read from it. Runs only once per git-annex run. -}
update :: Annex BranchState
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
update = runUpdateOnce $ updateTo =<< siblingBranches
{- Forces an update even if one has already been run. -}
forceUpdate :: Annex UpdateMade
forceUpdate = updateTo =<< siblingBranches
{- Merges the specified Refs into the index, if they have any changes not
- already in it. The Branch names are only used in the commit message;
- it's even possible that the provided Branches have not been updated to
- point to the Refs yet.
-
- The branch is fast-forwarded if possible, otherwise a merge commit is
- made.
-
- Before Refs are merged into the index, it's important to first stage the
merge: Use fast-forward merges when possible. Thanks Valentin Haenel for a test case showing how non-fast-forward merges could result in an ongoing pull/merge/push cycle. While the git-annex branch is fast-forwarded, git-annex's index file is still updated using the union merge strategy as before. There's no other way to update the index that would be any faster. It is possible that a union merge and a fast-forward result in different file contents: Files should have the same lines, but a union merge may change their order. If this happens, the next commit made to the git-annex branch will have some unnecessary changes to line orders, but the consistency of data should be preserved. Note that when the journal contains changes, a fast-forward is never attempted, which is fine, because committing those changes would be vanishingly unlikely to leave the git-annex branch at a commit that already exists in one of the remotes. The real difficulty is handling the case where multiple remotes have all changed. git-annex does find the best (ie, newest) one and fast forwards to it. If the remotes are diverged, no fast-forward is done at all. It would be possible to pick one, fast forward to it, and make a merge commit to the rest, I see no benefit to adding that complexity. Determining the best of N changed remotes requires N*2+1 calls to git-log, but these are fast git-log calls, and N is typically small. Also, typically some or all of the remote refs will be the same, and git-log is not called to compare those. In the real world I expect this will almost always add only 1 git-log call to the merge process. (Which already makes N anyway.)
2011-11-06 19:18:45 +00:00
- journal into the index. Otherwise, any changes in the journal would
- later get staged, and might overwrite changes made during the merge.
- This is only done if some of the Refs do need to be merged.
-
- Also handles performing any Transitions that have not yet been
- performed, in either the local branch, or the Refs.
-}
updateTo :: [(Git.Sha, Git.Branch)] -> Annex UpdateMade
updateTo pairs = ifM (annexMergeAnnexBranches <$> Annex.getGitConfig)
( updateTo' pairs
, return (UpdateMade False False)
)
updateTo' :: [(Git.Sha, Git.Branch)] -> Annex UpdateMade
updateTo' pairs = do
-- ensure branch exists, and get its current ref
branchref <- getBranch
ignoredrefs <- getIgnoredRefs
let unignoredrefs = excludeset ignoredrefs pairs
(tomerge, notnewer) <- if null unignoredrefs
then return ([], [])
else do
mergedrefs <- getMergedRefs
partitionM isnewer $
excludeset mergedrefs unignoredrefs
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
{- In a read-only repository, catching permission denied lets
- query operations still work, although they will need to do
- additional work since the refs are not merged. -}
catchPermissionDenied
(const (updatefailedperms tomerge))
(go branchref tomerge notnewer)
2012-12-13 04:24:19 +00:00
where
excludeset s = filter (\(r, _) -> S.notMember r s)
isnewer (r, _) = inRepo $ Git.Branch.changed fullname r
go branchref tomerge notnewer = do
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
dirty <- journalDirty gitAnnexJournalDir
journalcleaned <- if null tomerge
{- Even when no refs need to be merged, the index
- may still be updated if the branch has gotten ahead
- of the index, or just if the journal is dirty. -}
then ifM (needUpdateIndex branchref)
( lockJournal $ \jl -> do
forceUpdateIndex jl branchref
{- When there are journalled changes
- as well as the branch being updated,
- a commit needs to be done. -}
when dirty $
go' branchref dirty [] jl
return True
, if dirty
then ifM (annexAlwaysCommit <$> Annex.getGitConfig)
( lockJournal $ \jl -> do
go' branchref dirty [] jl
return True
, return False
)
else return True
)
else lockJournal $ \jl -> do
go' branchref dirty tomerge jl
return True
journalclean <- if journalcleaned
then not <$> privateUUIDsKnown
else pure False
addMergedRefs notnewer
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
return $ UpdateMade
{ refsWereMerged = not (null tomerge)
, journalClean = journalclean
}
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
go' branchref dirty tomerge jl = stagejournalwhen dirty jl $ do
let (refs, branches) = unzip tomerge
merge_desc <- if null tomerge
then commitMessage
else return $ "merging " ++
2012-12-13 04:24:19 +00:00
unwords (map Git.Ref.describe branches) ++
" into " ++ fromRef name
localtransitions <- getLocalTransitions
unless (null tomerge) $ do
showSideAction (UnquotedString merge_desc)
mapM_ checkBranchDifferences refs
mergeIndex jl refs
let commitrefs = nub $ fullname:refs
ifM (handleTransitions jl localtransitions commitrefs)
( runAnnexHook postUpdateAnnexHook
, do
ff <- if dirty
then return False
else inRepo $ Git.Branch.fastForward fullname refs
if ff
then updateIndex jl branchref
else commitIndex jl branchref merge_desc commitrefs
)
addMergedRefs tomerge
invalidateCache
stagejournalwhen dirty jl a
| dirty = stageJournal jl a
| otherwise = withIndex a
-- Preparing for read-only branch access with unmerged remote refs.
updatefailedperms tomerge = do
let refs = map fst tomerge
-- Gather any transitions that are new to either the
-- local branch or a remote ref, which will need to be
-- applied on the fly.
localts <- getLocalTransitions
remotets <- mapM getRefTransitions refs
ts <- if all (localts ==) remotets
then return []
else
let tcs = mapMaybe getTransitionCalculator $
knownTransitionList $
combineTransitions (localts:remotets)
in if null tcs
then return []
else do
config <- Annex.getGitConfig
trustmap <- calcTrustMap <$> getStaged trustLog
remoteconfigmap <- calcRemoteConfigMap <$> getStaged remoteLog
return $ map (\c -> c trustmap remoteconfigmap config) tcs
return $ UpdateFailedPermissions
{ refsUnmerged = refs
, newTransitions = ts
}
Fix a bug in the git-annex branch handling code that could cause info from a remote to not be merged and take effect immediately. This bug was turned up by the test suite, running fsck in direct mode. A repository was cloned, was put into direct mode, was fscked, and fsck incorrectly said that no copy existed of a file, that was actually present in origin. This turned out to occur because fsck first did a Annex.Branch.change, recording that it did not locally have the file. That was recorded in the journal. Since neither the git annex direct not the fsck had yet needed to read any info from the branch, but had only made changes to it, the origin/git-annex branch was not yet merged in. So the journal got a location log entry written to it, but this did not include the location log info for the origin. When fsck then did a Annex.Branch.get, it trusted the journal was cosnsitent, and returned it, again w/o merging from origin/git-annex. This latter behavior is the actual bug. Refer to commit e9bfa8eaed3ff59a4c0bc8d4d677bc493177807c for the thinking behind it being ok to make a change to a file on the branch, without first merging the branch. That thinking still stands. However, it means that files in the journal cannot be trusted to be consistent if the branch has not been merged. So, to fix, just enure the branch gets merged, even when reading from the journal. In tests, this does not seem to cause any extra merging. Except, of course, in the one case described above. But git annex add, etc, are able to make changes w/o first merging the branch.
2013-05-20 19:14:59 +00:00
{- Gets the content of a file, which may be in the journal, or in the index
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
- (and committed to the branch).
-
- Returns an empty string if the file doesn't exist yet.
-
- Updates the branch if necessary, to ensure the most up-to-date available
- content is returned.
-
- When permissions prevented updating the branch, reads the content from the
- journal, plus the branch, plus all unmerged refs. In this case, any
- transitions that have not been applied to all refs will be applied on
- the fly.
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
-}
get :: RawFilePath -> Annex L.ByteString
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
get file = do
st <- update
case getCache file st of
Just content -> return content
Nothing -> do
content <- if journalIgnorable st
then getRef fullname file
else if null (unmergedRefs st)
then getLocal file
else unmergedbranchfallback st
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
setCache file content
return content
where
unmergedbranchfallback st = do
l <- getLocal file
bs <- forM (unmergedRefs st) $ \ref -> getRef ref file
let content = l <> mconcat bs
return $ applytransitions (unhandledTransitions st) content
applytransitions [] content = content
applytransitions (changer:rest) content = case changer file content of
PreserveFile -> applytransitions rest content
ChangeFile builder -> do
let content' = toLazyByteString builder
if L.null content'
-- File is deleted, can't run any other
-- transitions on it.
then content'
else applytransitions rest content'
{- When the git-annex branch is unable to be updated due to permissions,
- and there are other git-annex branches that have not been merged into
- it, this gets the refs of those branches. -}
getUnmergedRefs :: Annex [Git.Ref]
getUnmergedRefs = unmergedRefs <$> update
{- Used to cache the value of a file, which has been read from the branch
- using some optimised method. The journal has to be checked, in case
- it has a newer version of the file that has not reached the branch yet.
-}
precache :: RawFilePath -> L.ByteString -> Annex ()
precache file branchcontent = do
st <- getState
content <- if journalIgnorable st
then pure branchcontent
else getJournalFileStale (GetPrivate True) file >>= return . \case
NoJournalledContent -> branchcontent
JournalledContent journalcontent -> journalcontent
PossiblyStaleJournalledContent journalcontent ->
branchcontent <> journalcontent
merge git-annex branch in memory in read-only repository Improved support for using git-annex in a read-only repository, git-annex branch information from remotes that cannot be merged into the git-annex branch will now not crash it, but will be merged in memory. To avoid this making git-annex behave one way in a read-only repository, and another way when it can write, it's important that Annex.Branch.get return the same thing (modulo log file compaction) in both cases. This manages that mostly. There are some exceptions: - When there is a transition in one of the remote git-annex branches that has not yet been applied to the local or other git-annex branches. Transitions are not handled. - `git-annex log` runs git log on the git-annex branch, and so it will not be able to show information coming from the other, not yet merged branches. - Annex.Branch.files only looks at files in the git-annex branch and not unmerged branches. This affects git-annex info output. - Annex.Branch.hs.overBranchFileContents ditto. Affects --all and also importfeed (but importfeed cannot work in a read-only repo anyway). - CmdLine.Seek.seekFilteredKeys when precaching location logs. Note use of Annex.Branch.fullname - Database.ContentIdentifier.needsUpdateFromLog and updateFromLog These warts make this not suitable to be merged yet. This readonly code path is more expensive, since it has to query several branches. The value does get cached, but still large queries will be slower in a read-only repository when there are unmerged git-annex branches. When annex.merge-annex-branches=false, updateTo skips doing anything, and so the read-only repository code does not get triggered. So a user who is bothered by the extra work can set that. Other writes to the repository can still result in permissions errors. This includes the initial creation of the git-annex branch, and of course any writes to the git-annex branch. Sponsored-by: Dartmouth College's Datalad project
2021-12-26 18:28:42 +00:00
setCache file content
{- Like get, but does not merge the branch, so the info returned may not
Fix a bug in the git-annex branch handling code that could cause info from a remote to not be merged and take effect immediately. This bug was turned up by the test suite, running fsck in direct mode. A repository was cloned, was put into direct mode, was fscked, and fsck incorrectly said that no copy existed of a file, that was actually present in origin. This turned out to occur because fsck first did a Annex.Branch.change, recording that it did not locally have the file. That was recorded in the journal. Since neither the git annex direct not the fsck had yet needed to read any info from the branch, but had only made changes to it, the origin/git-annex branch was not yet merged in. So the journal got a location log entry written to it, but this did not include the location log info for the origin. When fsck then did a Annex.Branch.get, it trusted the journal was cosnsitent, and returned it, again w/o merging from origin/git-annex. This latter behavior is the actual bug. Refer to commit e9bfa8eaed3ff59a4c0bc8d4d677bc493177807c for the thinking behind it being ok to make a change to a file on the branch, without first merging the branch. That thinking still stands. However, it means that files in the journal cannot be trusted to be consistent if the branch has not been merged. So, to fix, just enure the branch gets merged, even when reading from the journal. In tests, this does not seem to cause any extra merging. Except, of course, in the one case described above. But git annex add, etc, are able to make changes w/o first merging the branch.
2013-05-20 19:14:59 +00:00
- reflect changes in remotes.
- (Changing the value this returns, and then merging is always the
- same as using get, and then changing its value.) -}
getLocal :: RawFilePath -> Annex L.ByteString
getLocal = getLocal' (GetPrivate True)
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
getLocal' :: GetPrivate -> RawFilePath -> Annex L.ByteString
getLocal' getprivate file = do
fastDebug "Annex.Branch" ("read " ++ fromRawFilePath file)
go =<< getJournalFileStale getprivate file
2012-12-13 04:24:19 +00:00
where
go NoJournalledContent = getRef fullname file
go (JournalledContent journalcontent) = return journalcontent
go (PossiblyStaleJournalledContent journalcontent) = do
v <- getRef fullname file
return (v <> journalcontent)
{- Gets the content of a file as staged in the branch's index. -}
getStaged :: RawFilePath -> Annex L.ByteString
getStaged = getRef indexref
where
-- This makes git cat-file be run with ":file",
-- so it looks at the index.
indexref = Ref ""
getHistorical :: RefDate -> RawFilePath -> Annex L.ByteString
getHistorical date file =
-- This check avoids some ugly error messages when the reflog
-- is empty.
ifM (null <$> inRepo (Git.RefLog.get' [Param (fromRef fullname), Param "-n1"]))
( giveup ("No reflog for " ++ fromRef fullname)
, getRef (Git.Ref.dateRef fullname date) file
)
getRef :: Ref -> RawFilePath -> Annex L.ByteString
getRef ref file = withIndex $ catFile ref file
{- Applies a function to modify the content of a file.
-
- Note that this does not cause the branch to be merged, it only
2023-03-14 02:39:16 +00:00
- modifies the current content of the file on the branch.
-}
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
change :: Journalable content => RegardingUUID -> RawFilePath -> (L.ByteString -> content) -> Annex ()
change ru file f = lockJournal $ \jl -> f <$> getToChange ru file >>= set jl ru file
{- Applies a function which can modify the content of a file, or not. -}
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
maybeChange :: Journalable content => RegardingUUID -> RawFilePath -> (L.ByteString -> Maybe content) -> Annex ()
maybeChange ru file f = lockJournal $ \jl -> do
v <- getToChange ru file
case f v of
Just jv ->
let b = journalableByteString jv
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
in when (v /= b) $ set jl ru file b
_ -> noop
data ChangeOrAppend t = Change t | Append t
{- Applies a function that can either modify the content of the file,
- or append to the file. Appending can be more efficient when several
- lines are written to a file in succession.
-
- When annex.alwayscompact=false, the function is not passed the content
- of the journal file when the journal file already exists, and whatever
- value it provides is always appended to the journal file. That avoids
- reading the journal file, and so can be faster when many lines are being
- written to it. The information that is recorded will be effectively the
2023-03-14 02:39:16 +00:00
- same, only obsolete log lines will not get compacted.
-
- Currently, only appends when annex.alwayscompact=false. That is to
- avoid appending when an older version of git-annex is also in use in the
- same repository. An interrupted append could leave the journal file in a
- state that would confuse the older version. This is planned to be
- changed in a future repository version.
-}
changeOrAppend :: Journalable content => RegardingUUID -> RawFilePath -> (L.ByteString -> ChangeOrAppend content) -> Annex ()
changeOrAppend ru file f = lockJournal $ \jl ->
checkCanAppendJournalFile jl ru file >>= \case
Just appendable -> ifM (annexAlwaysCompact <$> Annex.getGitConfig)
( do
oldc <- getToChange ru file
case f oldc of
Change newc -> set jl ru file newc
Append toappend ->
set jl ru file $
oldc <> journalableByteString toappend
-- Use this instead in v11
-- or whatever.
-- append jl file appendable toappend
, case f mempty of
-- Append even though a change was
-- requested; since mempty was passed in,
-- the lines requested to change are
-- minimized.
Change newc -> append jl file appendable newc
Append toappend -> append jl file appendable toappend
)
Nothing -> do
oldc <- getToChange ru file
case f oldc of
Change newc -> set jl ru file newc
-- Journal file does not exist yet, so
-- cannot append and have to write it all.
Append toappend -> set jl ru file $
oldc <> journalableByteString toappend
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
{- Only get private information when the RegardingUUID is itself private. -}
getToChange :: RegardingUUID -> RawFilePath -> Annex L.ByteString
getToChange ru f = flip getLocal' f . GetPrivate =<< regardingPrivateUUID ru
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
{- Records new content of a file into the journal.
-
- This is not exported; all changes have to be made via change. This
- ensures that information that was written to the branch is not
- overwritten. Also, it avoids a get followed by a set without taking into
- account whether private information was gotten from the private
- git-annex index, and should not be written to the public git-annex
- branch.
-}
set :: Journalable content => JournalLocked -> RegardingUUID -> RawFilePath -> content -> Annex ()
set jl ru f c = do
journalChanged
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
setJournalFile jl ru f c
fastDebug "Annex.Branch" ("set " ++ fromRawFilePath f)
-- Could cache the new content, but it would involve
-- evaluating a Journalable Builder twice, which is not very
-- efficient. Instead, assume that it's not common to need to read
-- a log file immediately after writing it.
invalidateCache
{- Appends content to the journal file. -}
append :: Journalable content => JournalLocked -> RawFilePath -> AppendableJournalFile -> content -> Annex ()
append jl f appendable toappend = do
journalChanged
appendJournalFile jl appendable toappend
fastDebug "Annex.Branch" ("append " ++ fromRawFilePath f)
invalidateCache
{- Commit message used when making a commit of whatever data has changed
2023-03-14 02:39:16 +00:00
- to the git-annex branch. -}
commitMessage :: Annex String
commitMessage = fromMaybe "update" <$> getCommitMessage
{- Commit message used when creating the branch. -}
createMessage :: Annex String
createMessage = fromMaybe "branch created" <$> getCommitMessage
getCommitMessage :: Annex (Maybe String)
getCommitMessage = do
config <- Annex.getGitConfig
case annexCommitMessageCommand config of
Nothing -> return (annexCommitMessage config)
Just cmd -> catchDefaultIO (annexCommitMessage config) $
Just <$> liftIO (readProcess "sh" ["-c", cmd])
{- Stages the journal, and commits staged changes to the branch. -}
commit :: String -> Annex ()
2022-07-13 19:17:08 +00:00
commit = whenM (journalDirty gitAnnexJournalDir) . forceCommit
{- Commits the current index to the branch even without any journalled
- changes. -}
forceCommit :: String -> Annex ()
forceCommit message = lockJournal $ \jl ->
stageJournal jl $ do
ref <- getBranch
commitIndex jl ref message [fullname]
{- Commits the staged changes in the index to the branch.
-
- Ensures that the branch's index file is first updated to merge the state
- of the branch at branchref, before running the commit action. This
- is needed because the branch may have had changes pushed to it, that
- are not yet reflected in the index.
-
- The branchref value can have been obtained using getBranch at any
- previous point, though getting it a long time ago makes the race
- more likely to occur.
-
- Note that changes may be pushed to the branch at any point in time!
- So, there's a race. If the commit is made using the newly pushed tip of
- the branch as its parent, and that ref has not yet been merged into the
- index, then the result is that the commit will revert the pushed
- changes, since they have not been merged into the index. This race
- is detected and another commit made to fix it.
-
- (It's also possible for the branch to be overwritten,
- losing the commit made here. But that's ok; the data is still in the
- index and will get committed again later.)
-}
commitIndex :: JournalLocked -> Git.Ref -> String -> [Git.Ref] -> Annex ()
commitIndex jl branchref message parents = do
showStoringStateAction
commitIndex' jl branchref message message 0 parents
commitIndex' :: JournalLocked -> Git.Ref -> String -> String -> Integer -> [Git.Ref] -> Annex ()
commitIndex' jl branchref message basemessage retrynum parents = do
updateIndex jl branchref
cmode <- annexCommitMode <$> Annex.getGitConfig
committedref <- inRepo $ Git.Branch.commitAlways cmode message fullname parents
setIndexSha committedref
parentrefs <- commitparents <$> catObject committedref
when (racedetected branchref parentrefs) $
fixrace committedref parentrefs
2012-12-13 04:24:19 +00:00
where
-- look for "parent ref" lines and return the refs
commitparents = map (Git.Ref . snd) . filter isparent .
map (toassoc . L.toStrict) . L.split newline
newline = fromIntegral (ord '\n')
toassoc = separate' (== (fromIntegral (ord ' ')))
2012-12-13 04:24:19 +00:00
isparent (k,_) = k == "parent"
2012-12-13 04:24:19 +00:00
{- The race can be detected by checking the commit's
- parent, which will be the newly pushed branch,
- instead of the expected ref that the index was updated to. -}
racedetected expectedref parentrefs
| expectedref `elem` parentrefs = False -- good parent
| otherwise = True -- race!
2012-12-13 04:24:19 +00:00
{- To recover from the race, union merge the lost refs
- into the index. -}
2012-12-13 04:24:19 +00:00
fixrace committedref lostrefs = do
showSideAction "recovering from race"
let retrynum' = retrynum+1
-- small sleep to let any activity that caused
-- the race settle down
liftIO $ threadDelay (100000 + fromInteger retrynum')
mergeIndex jl lostrefs
let racemessage = basemessage ++ " (recovery from race #" ++ show retrynum' ++ "; expected commit parent " ++ show branchref ++ " but found " ++ show lostrefs ++ " )"
commitIndex' jl committedref racemessage basemessage retrynum' [committedref]
{- Lists all files on the branch. including ones in the journal
- that have not been committed yet.
-
- There may be duplicates in the list, when the journal has files that
- have not been written to the branch yet.
-
- In a read-only repository that has other git-annex branches that have
- not been merged in, returns Nothing, because it's not possible to
- efficiently handle that.
-}
files :: Annex (Maybe ([RawFilePath], IO Bool))
2012-09-15 19:40:13 +00:00
files = do
st <- update
if not (null (unmergedRefs st))
then return Nothing
else do
(bfs, cleanup) <- branchFiles
jfs <- journalledFiles
pjfs <- journalledFilesPrivate
-- ++ forces the content of the first list to be
-- buffered in memory, so use journalledFiles,
-- which should be much smaller most of the time.
-- branchFiles will stream as the list is consumed.
let l = jfs ++ pjfs ++ bfs
return (Just (l, cleanup))
{- Lists all files currently in the journal, but not files in the private
- journal. -}
journalledFiles :: Annex [RawFilePath]
journalledFiles = getJournalledFilesStale gitAnnexJournalDir
journalledFilesPrivate :: Annex [RawFilePath]
journalledFilesPrivate = ifM privateUUIDsKnown
( getJournalledFilesStale gitAnnexPrivateJournalDir
, return []
)
{- Files in the branch, not including any from journalled changes,
- and without updating the branch. -}
branchFiles :: Annex ([RawFilePath], IO Bool)
branchFiles = withIndex $ inRepo branchFiles'
branchFiles' :: Git.Repo -> IO ([RawFilePath], IO Bool)
branchFiles' = Git.Command.pipeNullSplit' $
lsTreeParams Git.LsTree.LsTreeRecursive (Git.LsTree.LsTreeLong False)
fullname
[Param "--name-only"]
{- Populates the branch's index file with the current branch contents.
-
- This is only done when the index doesn't yet exist, and the index
2023-03-14 02:39:16 +00:00
- is used to build up changes to be committed to the branch, and merge
- in changes from other branches.
-}
genIndex :: Git.Repo -> IO ()
genIndex g = Git.UpdateIndex.streamUpdateIndex g
[Git.UpdateIndex.lsTree fullname g]
{- Merges the specified refs into the index.
- Any changes staged in the index will be preserved. -}
mergeIndex :: JournalLocked -> [Git.Ref] -> Annex ()
mergeIndex jl branches = do
prepareModifyIndex jl
withHashObjectHandle $ \hashhandle ->
withCatFileHandle $ \ch ->
inRepo $ \g -> Git.UnionMerge.mergeIndex hashhandle ch g branches
{- Removes any stale git lock file, to avoid git falling over when
- updating the index.
-
- Since all modifications of the index are performed inside this module,
- and only when the journal is locked, the fact that the journal has to be
- locked when this is called ensures that no other process is currently
- modifying the index. So any index.lock file must be stale, caused
- by git running when the system crashed, or the repository's disk was
- removed, etc.
-}
prepareModifyIndex :: JournalLocked -> Annex ()
prepareModifyIndex _jl = do
index <- fromRepo gitAnnexIndex
void $ liftIO $ tryIO $ R.removeLink (index <> ".lock")
{- Runs an action using the branch's index file. -}
withIndex :: Annex a -> Annex a
withIndex = withIndex' False
withIndex' :: Bool -> Annex a -> Annex a
withIndex' bootstrapping a = withIndexFile AnnexIndexFile $ \f -> do
checkIndexOnce $ unlessM (liftIO $ doesFileExist f) $ do
unless bootstrapping create
createAnnexDirectory $ toRawFilePath $ takeDirectory f
unless bootstrapping $ inRepo genIndex
a
{- Updates the branch's index to reflect the current contents of the branch.
- Any changes staged in the index will be preserved.
-
- Compares the ref stored in the lock file with the current
- ref of the branch to see if an update is needed.
-}
updateIndex :: JournalLocked -> Git.Ref -> Annex ()
updateIndex jl branchref = whenM (needUpdateIndex branchref) $
forceUpdateIndex jl branchref
forceUpdateIndex :: JournalLocked -> Git.Ref -> Annex ()
forceUpdateIndex jl branchref = do
withIndex $ mergeIndex jl [fullname]
setIndexSha branchref
{- Checks if the index needs to be updated. -}
needUpdateIndex :: Git.Ref -> Annex Bool
needUpdateIndex branchref = do
f <- fromRawFilePath <$> fromRepo gitAnnexIndexStatus
committedref <- Git.Ref . firstLine' <$>
liftIO (catchDefaultIO mempty $ B.readFile f)
return (committedref /= branchref)
{- Record that the branch's index has been updated to correspond to a
- given ref of the branch. -}
setIndexSha :: Git.Ref -> Annex ()
setIndexSha ref = do
f <- fromRepo gitAnnexIndexStatus
writeLogFile f $ fromRef ref ++ "\n"
runAnnexHook postUpdateAnnexHook
{- Stages the journal into the index, and runs an action that
- commits the index to the branch. Note that the action is run
- inside withIndex so will automatically use the branch's index.
-
- Before staging, this removes any existing git index file lock.
- This is safe to do because stageJournal is the only thing that
- modifies this index file, and only one can run at a time, because
- the journal is locked. So any existing git index file lock must be
- stale, and the journal must contain any data that was in the process
- of being written to the index file when it crashed.
-}
stageJournal :: JournalLocked -> Annex () -> Annex ()
stageJournal jl commitindex = withIndex $ withOtherTmp $ \tmpdir -> do
prepareModifyIndex jl
g <- gitRepo
st <- getState
let dir = gitAnnexJournalDir st g
(jlogf, jlogh) <- openjlog (fromRawFilePath tmpdir)
withHashObjectHandle $ \h ->
withJournalHandle gitAnnexJournalDir $ \jh ->
Git.UpdateIndex.streamUpdateIndex g
[genstream dir h jh jlogh]
commitindex
liftIO $ cleanup (fromRawFilePath dir) jlogh jlogf
2012-12-13 04:24:19 +00:00
where
genstream dir h jh jlogh streamer = readDirectory jh >>= \case
Nothing -> return ()
Just file -> do
let path = dir P.</> toRawFilePath file
unless (dirCruft file) $ whenM (isfile path) $ do
sha <- Git.HashObject.hashFile h path
hPutStrLn jlogh file
streamer $ Git.UpdateIndex.updateIndexLine
sha TreeFile (asTopFilePath $ fileJournal $ toRawFilePath file)
genstream dir h jh jlogh streamer
isfile file = isRegularFile <$> R.getFileStatus file
-- Clean up the staged files, as listed in the temp log file.
-- The temp file is used to avoid needing to buffer all the
-- filenames in memory.
cleanup dir jlogh jlogf = do
hFlush jlogh
hSeek jlogh AbsoluteSeek 0
stagedfs <- lines <$> hGetContents jlogh
mapM_ (removeFile . (dir </>)) stagedfs
hClose jlogh
removeWhenExistsWith (R.removeLink) (toRawFilePath jlogf)
openjlog tmpdir = liftIO $ openTmpFileIn tmpdir "jlog"
getLocalTransitions :: Annex Transitions
getLocalTransitions =
2021-12-28 16:15:51 +00:00
parseTransitionsStrictly "local"
<$> getLocal transitionsLog
2021-12-28 16:15:51 +00:00
{- This is run after the refs have been merged into the index,
- but before the result is committed to the branch.
- (Which is why it's passed the contents of the local branches's
- transition log before that merge took place.)
-
- When the refs contain transitions that have not yet been done locally,
- the transitions are performed on the index, and a new branch
- is created from the result.
-
- When there are transitions recorded locally that have not been done
- to the remote refs, the transitions are performed in the index,
- and committed to the existing branch. In this case, the untransitioned
- remote refs cannot be merged into the branch (since transitions
- throw away history), so they are added to the list of refs to ignore,
- to avoid re-merging content from them again.
-}
handleTransitions :: JournalLocked -> Transitions -> [Git.Ref] -> Annex Bool
handleTransitions jl localts refs = do
remotets <- mapM getRefTransitions refs
if all (localts ==) remotets
then return False
else do
let m = M.fromList (zip refs remotets)
let allts = combineTransitions (localts:remotets)
let (transitionedrefs, untransitionedrefs) =
partition (\r -> M.lookup r m == Just allts) refs
performTransitionsLocked jl allts (localts /= allts) transitionedrefs
ignoreRefs untransitionedrefs
return True
{- Performs the specified transitions on the contents of the index file,
- commits it to the branch, or creates a new branch.
-}
performTransitions :: Transitions -> Bool -> [Ref] -> Annex ()
performTransitions ts neednewlocalbranch transitionedrefs = lockJournal $ \jl ->
performTransitionsLocked jl ts neednewlocalbranch transitionedrefs
performTransitionsLocked :: JournalLocked -> Transitions -> Bool -> [Ref] -> Annex ()
performTransitionsLocked jl ts neednewlocalbranch transitionedrefs = do
-- For simplicity & speed, we're going to use the Annex.Queue to
-- update the git-annex branch, while it usually holds changes
-- for the head branch. Flush any such changes.
Annex.Queue.flush
-- Stop any running git cat-files, to ensure that the
-- getStaged calls below use the current index, and not some older
-- one.
catFileStop
withIndex $ do
prepareModifyIndex jl
run $ mapMaybe getTransitionCalculator tlist
Annex.Queue.flush
if neednewlocalbranch
then do
cmode <- annexCommitMode <$> Annex.getGitConfig
-- Creating a new empty branch must happen
-- atomically, so if this is interrupted,
-- it will not leave the new branch created
-- but without exports grafted in.
c <- inRepo $ Git.Branch.commitShaAlways
cmode message transitionedrefs
void $ regraftexports c
else do
ref <- getBranch
ref' <- regraftexports ref
commitIndex jl ref' message
(nub $ fullname:transitionedrefs)
where
message
| neednewlocalbranch && null transitionedrefs = "new branch for transition " ++ tdesc
| otherwise = "continuing transition " ++ tdesc
tdesc = show $ map describeTransition tlist
tlist = knownTransitionList ts
{- The changes to make to the branch are calculated and applied to
- the branch directly, rather than going through the journal,
- which would be innefficient. (And the journal is not designed
- to hold changes to every file in the branch at once.)
-
- When a file in the branch is changed by transition code,
- its new content is remembered and fed into the code for subsequent
- transitions.
-}
run [] = noop
run changers = do
config <- Annex.getGitConfig
trustmap <- calcTrustMap <$> getStaged trustLog
remoteconfigmap <- calcRemoteConfigMap <$> getStaged remoteLog
-- partially apply, improves performance
let changers' = map (\c -> c trustmap remoteconfigmap config) changers
(fs, cleanup) <- branchFiles
forM_ fs $ \f -> do
content <- getStaged f
apply changers' f content
liftIO $ void cleanup
apply [] _ _ = return ()
apply (changer:rest) file content = case changer file content of
PreserveFile -> apply rest file content
ChangeFile builder -> do
let content' = toLazyByteString builder
if L.null content'
then do
Annex.Queue.addUpdateIndex
2023-02-27 19:02:53 +00:00
=<< inRepo (Git.UpdateIndex.unstageFile file)
-- File is deleted; can't run any other
-- transitions on it.
return ()
else do
sha <- hashBlob content'
Annex.Queue.addUpdateIndex $ Git.UpdateIndex.pureStreamer $
Git.UpdateIndex.updateIndexLine sha TreeFile (asTopFilePath file)
apply rest file content'
-- Trees mentioned in export.log were grafted into the old
-- git-annex branch to make sure they remain available.
-- Re-graft the trees.
regraftexports parent = do
l <- exportedTreeishes . M.elems . parseExportLogMap
<$> getStaged exportLog
c <- regraft l parent
inRepo $ Git.Branch.update' fullname c
setIndexSha c
return c
where
regraft [] c = pure c
regraft (et:ets) c =
-- Verify that the tree object exists.
catObjectDetails et >>= \case
Just _ ->
prepRememberTreeish et graftpoint c
>>= regraft ets
Nothing -> regraft ets c
graftpoint = asTopFilePath exportTreeGraftPoint
checkBranchDifferences :: Git.Ref -> Annex ()
checkBranchDifferences ref = do
theirdiffs <- allDifferences . parseDifferencesLog
<$> catFile ref differenceLog
mydiffs <- annexDifferences <$> Annex.getGitConfig
when (theirdiffs /= mydiffs) $
2017-02-11 09:38:49 +00:00
giveup "Remote repository is tuned in incompatible way; cannot be merged with local repository."
ignoreRefs :: [Git.Sha] -> Annex ()
ignoreRefs rs = do
old <- getIgnoredRefs
let s = S.unions [old, S.fromList rs]
f <- fromRepo gitAnnexIgnoredRefs
writeLogFile f $
unlines $ map fromRef $ S.elems s
getIgnoredRefs :: Annex (S.Set Git.Sha)
getIgnoredRefs =
Windows: Fix CRLF handling in some log files In particular, the mergedrefs file was written with CR added to each line, but read without CRLF handling. This resulted in each update of the file adding CR to each line in it, growing the number of lines, while also preventing the optimisation from working, so it remerged unncessarily. writeFile and readFile do NewlineMode translation on Windows. But the ByteString conversion prevented that from happening any longer. I've audited for other cases of this, and found three more (.git/annex/index.lck, .git/annex/ignoredrefs, and .git/annex/import/). All of those also only prevent optimisations from working. Some other files are currently both read and written with ByteString, but old git-annex may have written them with NewlineMode translation. Other files are at risk for breakage later if the reader gets converted to ByteString. This is a minimal fix, but should be enough, as long as I remember to use fileLines when splitting a ByteString into lines. This leaves files written using ByteString without CR added, but that's ok because old git-annex has no difficulty reading such files. When the mergedrefs file has gotten lines that end with "\r\r\r\n", this will eventually clean it up. Each update will remove a single trailing CR. Note that S8.lines is still used in eg Command.Unused, where it is parsing git show-ref, and similar in Git/*. git commands don't include CR in their output so that's ok. Sponsored-by: Joshua Antonishen on Patreon
2023-10-30 18:23:23 +00:00
S.fromList . mapMaybe Git.Sha.extractSha . fileLines' <$> content
where
content = do
f <- fromRawFilePath <$> fromRepo gitAnnexIgnoredRefs
liftIO $ catchDefaultIO mempty $ B.readFile f
addMergedRefs :: [(Git.Sha, Git.Branch)] -> Annex ()
addMergedRefs [] = return ()
addMergedRefs new = do
old <- getMergedRefs'
-- Keep only the newest sha for each branch.
let l = nubBy ((==) `on` snd) (new ++ old)
f <- fromRepo gitAnnexMergedRefs
writeLogFile f $
unlines $ map (\(s, b) -> fromRef s ++ '\t' : fromRef b) l
getMergedRefs :: Annex (S.Set Git.Sha)
getMergedRefs = S.fromList . map fst <$> getMergedRefs'
getMergedRefs' :: Annex [(Git.Sha, Git.Branch)]
getMergedRefs' = do
f <- fromRawFilePath <$> fromRepo gitAnnexMergedRefs
s <- liftIO $ catchDefaultIO mempty $ B.readFile f
Windows: Fix CRLF handling in some log files In particular, the mergedrefs file was written with CR added to each line, but read without CRLF handling. This resulted in each update of the file adding CR to each line in it, growing the number of lines, while also preventing the optimisation from working, so it remerged unncessarily. writeFile and readFile do NewlineMode translation on Windows. But the ByteString conversion prevented that from happening any longer. I've audited for other cases of this, and found three more (.git/annex/index.lck, .git/annex/ignoredrefs, and .git/annex/import/). All of those also only prevent optimisations from working. Some other files are currently both read and written with ByteString, but old git-annex may have written them with NewlineMode translation. Other files are at risk for breakage later if the reader gets converted to ByteString. This is a minimal fix, but should be enough, as long as I remember to use fileLines when splitting a ByteString into lines. This leaves files written using ByteString without CR added, but that's ok because old git-annex has no difficulty reading such files. When the mergedrefs file has gotten lines that end with "\r\r\r\n", this will eventually clean it up. Each update will remove a single trailing CR. Note that S8.lines is still used in eg Command.Unused, where it is parsing git show-ref, and similar in Git/*. git commands don't include CR in their output so that's ok. Sponsored-by: Joshua Antonishen on Patreon
2023-10-30 18:23:23 +00:00
return $ map parse $ fileLines' s
where
parse l =
let (s, b) = separate' (== (fromIntegral (ord '\t'))) l
in (Ref s, Ref b)
{- Grafts a treeish into the branch at the specified location,
- and then removes it. This ensures that the treeish won't get garbage
- collected, and will always be available as long as the git-annex branch
- is available.
-
- Returns the sha of the git commit made to the git-annex branch.
-}
rememberTreeish :: Git.Ref -> TopFilePath -> Annex Git.Sha
rememberTreeish treeish graftpoint = lockJournal $ \jl -> do
branchref <- getBranch
updateIndex jl branchref
c <- prepRememberTreeish treeish graftpoint branchref
inRepo $ Git.Branch.update' fullname c
-- The tree in c is the same as the tree in branchref,
-- and the index was updated to that above, so it's safe to
-- say that the index contains c.
setIndexSha c
return c
{- Create a series of commits that graft a tree onto the parent commit,
- and then remove it. -}
prepRememberTreeish :: Git.Ref -> TopFilePath -> Git.Ref -> Annex Git.Sha
prepRememberTreeish treeish graftpoint parent = do
origtree <- fromMaybe (giveup "unable to determine git-annex branch tree") <$>
inRepo (Git.Ref.tree parent)
addedt <- inRepo $ Git.Tree.graftTree treeish graftpoint origtree
cmode <- annexCommitMode <$> Annex.getGitConfig
c <- inRepo $ Git.Branch.commitTree cmode
["graft"] [parent] addedt
inRepo $ Git.Branch.commitTree cmode
["graft cleanup"] [c] origtree
2021-04-21 18:19:58 +00:00
{- Runs an action on the content of selected files from the branch.
- This is much faster than reading the content of each file in turn,
- because it lets git cat-file stream content without blocking.
2021-04-21 18:19:58 +00:00
-
- The action is passed a callback that it can repeatedly call to read
- the next file and its contents. When there are no more files, the
- callback will return Nothing.
-
- In some cases the callback may return the same file more than once,
- with different content. This happens rarely, only when the journal
- contains additional information, and the last version of the
- file it returns is the most current one.
-
- In a read-only repository that has other git-annex branches that have
- not been merged in, returns Nothing, because it's not possible to
- efficiently handle that.
2021-04-21 18:19:58 +00:00
-}
overBranchFileContents
:: (RawFilePath -> Maybe v)
-> (Annex (Maybe (v, RawFilePath, Maybe L.ByteString)) -> Annex a)
-> Annex (Maybe a)
2021-04-21 18:19:58 +00:00
overBranchFileContents select go = do
st <- update
if not (null (unmergedRefs st))
then return Nothing
else Just <$> overBranchFileContents' select go st
overBranchFileContents'
:: (RawFilePath -> Maybe v)
-> (Annex (Maybe (v, RawFilePath, Maybe L.ByteString)) -> Annex a)
-> BranchState
-> Annex a
overBranchFileContents' select go st = do
2021-04-21 18:19:58 +00:00
g <- Annex.gitRepo
(l, cleanup) <- inRepo $ Git.LsTree.lsTree
Git.LsTree.LsTreeRecursive
(Git.LsTree.LsTreeLong False)
fullname
let select' f = fmap (\v -> (v, f)) (select f)
buf <- liftIO newEmptyMVar
let go' reader = go $ liftIO reader >>= \case
Just ((v, f), content) -> do
content' <- checkjournal f content
return (Just (v, f, content'))
Nothing
| journalIgnorable st -> return Nothing
-- The journal did not get committed to the
-- branch, and may contain files that
-- are not present in the branch, which
-- need to be provided to the action still.
-- This can cause the action to be run a
-- second time with a file it already ran on.
| otherwise -> liftIO (tryTakeMVar buf) >>= \case
Nothing -> do
jfs <- journalledFiles
pjfs <- journalledFilesPrivate
drain buf jfs pjfs
Just (jfs, pjfs) -> drain buf jfs pjfs
2021-04-21 18:19:58 +00:00
catObjectStreamLsTree l (select' . getTopFilePath . Git.LsTree.file) g go'
`finally` liftIO (void cleanup)
where
-- Check the journal, in case it did not get committed to the branch
checkjournal f branchcontent
| journalIgnorable st = return branchcontent
| otherwise = getJournalFileStale (GetPrivate True) f >>= return . \case
NoJournalledContent -> branchcontent
JournalledContent journalledcontent ->
Just journalledcontent
PossiblyStaleJournalledContent journalledcontent ->
Just (fromMaybe mempty branchcontent <> journalledcontent)
drain buf fs pfs = case getnext fs pfs of
Just (v, f, fs', pfs') -> do
liftIO $ putMVar buf (fs', pfs')
content <- getJournalFileStale (GetPrivate True) f >>= \case
NoJournalledContent -> return Nothing
JournalledContent journalledcontent ->
return (Just journalledcontent)
PossiblyStaleJournalledContent journalledcontent -> do
-- This is expensive, but happens
-- only when there is a private
-- journal file.
content <- getRef fullname f
return (Just (content <> journalledcontent))
return (Just (v, f, content))
Nothing -> do
liftIO $ putMVar buf ([], [])
return Nothing
getnext [] [] = Nothing
getnext (f:fs) pfs = case select f of
Nothing -> getnext fs pfs
Just v -> Just (v, f, fs, pfs)
getnext [] (pf:pfs) = case select pf of
Nothing -> getnext [] pfs
Just v -> Just (v, pf, [], pfs)
{- Check if the git-annex branch has been updated from the oldtree.
- If so, returns the tuple of the old and new trees. -}
updatedFromTree :: Git.Sha -> Annex (Maybe (Git.Sha, Git.Sha))
updatedFromTree oldtree =
inRepo (Git.Ref.tree fullname) >>= \case
Just currtree | currtree /= oldtree ->
return $ Just (oldtree, currtree)
_ -> return Nothing