I was thinking that discardIncompleteAppend would make it strict, since
it looks at the end of the bytestring. But, it's applied lazily..
This probably fixes windows, which was failing:
git-annex.exe: .git\annex\journal\trust.log: DeleteFile "\\\\?\\C:\\Users\\runneradmin\\.t\\5\\tmprepo22\\.git\\annex\\journal\\trust.log": permission denied (The process cannot access the file because it is being used by another process.)
I would like for a new repo version to enable appends, but to do so
safely would need a v11 followed by a 1 year delay followed by a v12
that does it. Since a similar v9 and v10 transition is currently
happening, and is less than 6 months along in most repos, it does not
feel wise to stack up another year-long transition behind that. What if
I need to hurry up a new repo version for some other change?
Added todo so I remember to make this change at some time when a v11
and probably v12 repo version do make sense.
Sponsored-by: Dartmouth College's DANDI project
An append that is interrupted and writes part of a line is now dealt
with by subsequent reads and appends. This also handles a read that
happens at the same time as an append to the file.
Old versions of git-annex will still see a partially written line,
and could get confused. Since appends are currently done for url logs
and location logs, the confusion is limited to a substring of the actual
url or UUID of the remote being read. This will not affect writes, since
the journal file is locked when reading in preparation for writing.
However, the bad data can be output by git-annex and used by other
things, or could cause surprising behavior by git-annex. Including eg,
downloading the content of the wrong url.
So, something needs to be done to prevent old versions of git-annex from
running in a repository where this appending is being done..
Sponsored-by: Dartmouth College's DANDI project
This reverts commit 2e6e9876e3.
This is gonna be needed after all.. The append will only be atomic if
the journal is locked, because the file being appended will have to be
moved out of the way to avoid an old version of git-annex seeing an
incomplete write to it. When git-annex finds that the file is not in the
journal, and checks the append location, locking will be needed to avoid
a race causing it to miss it in the append location too due to it being
moved back to the journal.
Added annex.alwayscompact setting which can be unset to speed up writes to
the git-annex branch in some cases.
Sponsored-by: Dartmouth College's DANDI project
This reverts commit 47358a6f95.
This added overhead, and will not be needed, because appends are going
to have to be made atomic for other reasons than avoiding incomplete
reads of data being appended.
In particular, when git-annex is interrupted in the middle of an append,
it must not leave the file with a partially written line. So appending
has to somehow be made fully atomic.
Currently this is not an improvement, but it allows for optimising
appendJournalFile later. With an optimised appendJournalFile, this will
greatly speed up access patterns like git-annex addurl of a lot of urls
to the same key, where the log file can grow rather large. Appending
rather than re-writing the journal file for each line can save a lot of
disk writes.
It still has to read the current journal or branch file, to check
if it can append to it, and so when the journal file does not exist yet,
it can write the old content from the branch to it. Probably the re-reads
are better cached by the filesystem than repeated writes. (If the
re-reads turn out to keep performance bad, they could be eliminated, at
the cost of not being able to compact the log when replacing old
information in it. That could be enabled by a switch.)
While the immediate need is to affect addurl writes, it was implemented
at the level of presence logs, so will also perhaps speed up location logs.
The only added overhead is the call to isNewInfo, which only needs to
compare ByteStrings. Helping to balance that out, it avoids compactLog
when it's able to append.
Sponsored-by: Dartmouth College's DANDI project
This is not currently necessary; journal files are updated atomically.
However, for faster appends to large journal files, locking on read will
be needed, because appends are not atomic.
Sponsored-by: Dartmouth College's DANDI project
Fix a reversion that prevented --batch commands (and the assistant)
from noticing data written to the journal by other commands.
I have not identified which commit broke this for sure,
but probably it was aeca7c2207
--batch commands that wrote to the journal avoided the problem since
journalIgnorable sets unset on write. It's a little bit surprising that
nobody noticed that query --batch commands did not see data written by
other commands.
Sponsored-by: Dartmouth College's DANDI project
At this point I've checked all AnnexState values and these were all that
remained that could move.
Pity that Annex.repo can't move, but it gets modified sometimes..
A couple of AnnexState values are set by options and could be AnnexRead,
but happen to use Annex when being set.
Sponsored-by: Max Thoursie on Patreon
Some small wins, almost certianly swamped by the system calls, but still
worthwhile progress on the RawFilePath conversion.
Sponsored-by: Erik Bjäreholt on Patreon
The remaining callers all did not rely on it checking gitignore, so were
easy to convert.
They were susceptable to the same overwrite race as add and fix,
although less likely to have it and a narrower window than add's race.
Command.Rekey in passing got an unncessary call to removeFile deleted.
addSymlink handles deleting any existing worktree file.
Oddly, the second write did not cause it to lose the mtime inherited
from the file being added, although the mtime was not provided to that
write but only to the first. I don't quite know why that worked before!
In the unlikely case where git-annex add is run on an annex symlink that
is not already added, and while it's processing it, the annex symlink is
overwritten with something else, avoid git-annex overwriting that with
the symlink again.
Sponsored-by: Jack Hill on Patreon
This is not a complete fix for all such races, only the one where a
large file gets changed while adding and gets added to git rather than
to the annex.
addLink needs to go away, any caller of it is probably subject to the
same kind of race. (Also, addLink itself fails to check gitignore when
symlinks are not supported.)
ingestAdd no longer checks gitignore. (It didn't check it consistently
before either, since there were cases where it did not run git add!)
When git-annex import calls it, it's already checked gitignore itself
earlier. When git-annex add calls it, it's usually on files found
by withFilesNotInGit, which handles checking ignores.
There was one other case, when git-annex add --batch calls it. In that
case, old git-annex behaved rather badly, it would seem to add the file,
but git add would later fail, leaving the file as an unstaged annex symlink.
That behavior has also been fixed.
Sponsored-by: Brett Eisenberg on Patreon
Use cases include using git-annex init --no-autoenable and then going back
and enabling the special remotes that have autoenable configured. As well
as just querying to remember which ones have it enabled.
It lists all special remotes that have autoenable=yes whether currently
enabled or not. And it can be used with --json.
I pondered making this "git-annex info autoenable", but that seemed wrong
because then if the use has a directory named "autoenable", it's unclear
what they are asking for. (Although "git-annex info remote" may be
similarly unclear.) Making it an option does mean that it can't be provided
via --batch though.
Sponsored-by: Dartmouth College's Datalad project
This reverts windows-specific parts of 5a98f2d509
There were no code paths in common between windows and unix, so this
will return Windows to the old behavior.
The problem that the commit talks about has to do with multiple different
locations where git-annex can store annex object files, but that is not
too relevant to Windows anyway, because on windows the filesystem is always
treated as criplled and/or symlinks are not supported, so it will only
use one object location. It would need to be using a repo populated
in another OS to have the other object location in use probably.
Then a drop and get could possibly lead to a dangling lock file.
And, I was not able to actually reproduce that situation happening
before making that commit, even when I forced a race. So making these
changes on windows was just begging trouble..
I suspect that the change that caused the reversion is in
Annex/Content/Presence.hs. It checks if the content file exists,
and then called modifyContentDirWhenExists, which seems like it would
not fail, but if something deleted the content file at that point,
that call would fail. Which would result in an exception being thrown,
which should not normally happen from a call to inAnnexSafe. That was a
windows-specific change; the unix side did not have an equivilant
change.
Sponsored-by: Dartmouth College's Datalad project
Deal with git's recent changes to fix CVE-2022-24765, which prevent using
git in a repository owned by someone else.
That makes git config --list not list the repo's configs, only global
configs. So annex.uuid and annex.version are not visible to git-annex.
It displayed a message about that, which is not right for this situation.
Detect the situation and display a better message, similar to the one other
git commands display.
Also, git-annex init when run in that situation would overwrite annex.uuid
with a new one, since it couldn't see the old one. Add a check to prevent
it running too in this situation. It may be that this fix has security
implications, if a config set by the malicious user who owns the repo
causes git or git-annex to run code. I don't think any git-annex configs
get run by git-annex init. It may be that some git config of a command
does get run by one of the git commands that git-annex init runs. ("git
status" is the command that prompted the CVE-2022-24765, since
core.fsmonitor can cause it to run a command). Since I don't know how
to exploit this, I'm not treating it as a security fix for now.
Note that passing --git-dir makes git bypass the security check. git-annex
does pass --git-dir to most calls to git, which it does to avoid needing
chdir to the directory containing a git repository when accessing a remote.
So, it's possible that somewhere in git-annex it gets as far as running git
with --git-dir, and git reads some configs that are unsafe (what
CVE-2022-24765 is about). This seems unlikely, it would have to be part of
git-annex that runs in git repositories that have no (visible) annex.uuid,
and git-annex init is the only one that I can think of that then goes on to
run git, as discussed earlier. But I've not fully ruled out there being
others..
The git developers seem mostly worried about "git status" or a similar
command implicitly run by a shell prompt, not an explicit use of git in
such a repository. For example, Ævar Arnfjörð Bjarma wrote:
> * There are other bits of config that also point to executable things,
> e.g. core.editor, aliases etc, but nothing has been found yet that
> provides the "at a distance" effect that the core.fsmonitor vector
> does.
>
> I.e. a user is unlikely to go to /tmp/some-crap/here and run "git
> commit", but they (or their shell prompt) might run "git status", and
> if you have a /tmp/.git ...
Sponsored-by: Jarkko Kniivilä on Patreon
The purpose of this is to fix situations where the annex object file is
stored in a directory structure other than where annex symlinks point to.
But it will also move object files from the hashdirmixed back to
hashdirlower if the repo configuration makes that the normal location.
It would have been more work to avoid that than to let it do it.
Sponsored-by: Dartmouth College's Datalad project
It can be changed to 1 via a tuning, rather than the 2 this assumed. So
it would have tried to rmdir .git/annex/objects in that case, which
would not hurt anything, but is not what it is supposed to do.
Sponsored-by: Dartmouth College's Datalad project
If the content directory does not exist, then it does not make sense to
lock the content file, as it also does not exist, and so it's ok for the
lock operation to fail.
This avoids potential races where the content file exists but is then
deleted/renamed, while another process sees that it exists and goes to
lock it, resulting in a dangling lock file in an otherwise empty object
directory.
Also renamed modifyContent to modifyContentDir since it is not only
necessarily used for modifying content files, but also other files in
the content directory.
Sponsored-by: Dartmouth College's Datalad project
None of the special remotes do it yet, but this lays the groundwork.
Added MustFinishIncompleteVerify so that, when an incremental verify is
started but not complete, it can be forced to finish it. Otherwise, it
would have skipped doing it when verification is disabled, but
verification must always be done when retrievin from export remotes
since files can be modified during retrieval.
Note that retrieveExportWithContentIdentifier doesn't support incremental
verification yet. And I'm not sure if it can -- it doesn't know the Key
before it downloads the content. It seems a new API call would need to
be split out of that, which is provided with the key.
Sponsored-by: Dartmouth College's Datalad project
Ignore annex.numcopies set to 0 in gitattributes or git config, or by
git-annex numcopies or by --numcopies, since that configuration would make
git-annex easily lose data. Same for mincopies.
This is a continuation of the work to make data only be able to be lost
when --force is used. It earlier led to the --trust option being disabled,
and similar reasoning applies here.
Most numcopies configs had docs that strongly discouraged setting it to 0
anyway. And I can't imagine a use case for setting to 0. Not that there
might not be one, but it's just so far from the intended use case of
git-annex, of managing and storing your data, that it does not seem like
it makes sense to cater to such a hypothetical use case, where any
git-annex drop can lose your data at any time.
Using a smart constructor makes sure every place avoids 0. Note that this
does mean that NumCopies is for the configured desired values, and not the
actual existing number of copies, which of course can be 0. The name
configuredNumCopies is used to make that clear.
Sponsored-by: Brock Spratlen on Patreon
Removed vendored copy of http-client-restricted, and removed the
HttpClientRestricted build flag that avoided that dependency.
http-client-restricted is in Debian stable, and the i386ancient build also
uses it, so I think this vendored copy is no longer needed.
Sponsored-by: Noam Kremen on Patreon
It did nothing, since at this point the link is dangling. But when there
is a thaw hook, it would probably not be happy to be asked to run on a
symlink, or might do something unexpected.
Sponsored-by: Dartmouth College's Datalad project
When annex.freezecontent-command is set, and the filesystem does not
support removing write bits, avoid treating it as a crippled filesystem.
The hook may be enough to prevent writing on its own, and some filesystems
ignore attempts to remove write bits.
Sponsored-by: Dartmouth College's Datalad project
It will then proceed to add the file the same as if it were any other
file containing possibly annexable content. Usually the file is one that
was annexed before, so the new, probably corrupt content will also be added
to the annex. If the file was not annexed before, the content will be added
to git.
It's not possible for the smudge filter to throw an error here, because
git then just adds the file to git anyway.
Sponsored-by: Dartmouth College's Datalad project
This format is designed to detect accidental appends, while having some
room for future expansion.
Detect when an unlocked file whose content is not present has gotten some
other content appended to it, and avoid treating it as a pointer file, so
that appended content will not be checked into git, but will be annexed
like any other file.
Dropped the max size of a pointer file down to 32kb, it was around 80 kb,
but without any good reason and certianly there are no valid pointer files
anywhere that are larger than 8kb, because it's just been specified what it
means for a pointer file with additional data even looks like.
I assume 32kb will be good enough for anyone. ;-) Really though, it needs
to be some smallish number, because that much of a file in git gets read
into memory when eg, catting pointer files. And since we have no use cases
for the extra lines of a pointer file yet, except possibly to add
some human-visible explanation that it is a git-annex pointer file, 32k
seems as reasonable an arbitrary number as anything. Increasing it would be
possible, eg to 64k, as long as users of such jumbo pointer files didn't
mind upgrading all their git-annex installations to one that supports the
new larger size.
Sponsored-by: Dartmouth College's Datalad project
A few places were reading the max symlink size of a pointer file,
then passing tp parseLinkTargetOrPointer. Which is fine currently, but
to support pointer files with lines of data after the pointer, enough
has to be read that parseLinkTargetOrPointer can be assured of seeing
enough of that data to know if it's correctly formatted.
Sponsored-by: Dartmouth College's Datalad project
Only one place remained that differentiated between them.
It is the case that a symlink target that happens to contain a newline
somehow will be treated as a link to a key truncated at the newline.
This is super unlikely to happen, and since a key cannot actually
contain a newline, it's as good a behavior as any. Anyway, this commit
does not change the behavior there, although arguably it should be
changed. Note that getAnnexLinkTarget does prevent a symlink target
containing a newline.
Sponsored-by: Dartmouth College's Datalad project
File matching options like --include will be rejected in situations where
there is no filename to match against. (Or where there is a filename but
it's not relative to the cwd, or otherwise seemed too bothersome to match
against.)
The addition of listKeys' was necessary to avoid using more memory in the
common case of "git-annex info". Adding a filterM would have caused the
list to buffer in memory and not stream. This is an ugly hack, but listKeys
had previously run Annex operations inside unafeInterleaveIO (for direct
mode). And matching against a matcher should hopefully not change any Annex
state.
This does allow for eg `git-annex info somefile --include=*.ext`
although why someone would want to do that I don't really know. But it
seems to make sense to allow it.
But, consider: `git-annex info ./somefile --include=somefile`
This does not match, so will not display info about somefile.
If the user really wants to, they can `--include=./somefile`.
Using matching options like --copies or --in=remote seems likely to be
slower than git-annex find with those options, because unlike such
commands, info does not have optimised streaming through the matcher.
Note that `git-annex info remote` is not the same as
`git-annex info --in remote`. The former shows info about all files in
the remote. The latter shows local keys that are also in that remote.
The output should make that clear, but this still seems like a point
where users could get confused.
Sponsored-by: Jochen Bartl on Patreon
Implemented by making Git.Queue have a FlushAction, which can accumulate
along with another action on files, and runs only once the other action has
run.
This lets git-annex unlock queue up git update-index actions, without
conflicting with the restagePointerFiles FlushActions.
In a repository with filter-process enabled, git-annex unlock will
often not take any more time than before, though it may when the files are
large. Either way, it should always slow down less than git-annex status
speeds up.
When filter-process is not enabled, git-annex unlock will slow down as much
as git status speeds up.
Sponsored-by: Jochen Bartl on Patreon
autoUpgradeableVersions had latestVersion (10), but it did not make
sense for asking for old version 6 to get version 10, while asking for
version 8 got version 8. So use defaultVersion (8) instead.
Sponsored-by: Dartmouth College's Datalad project
The problem is that withContentLockFile, in a v8 repo, has to take a shared
lock of `.git/annex/content.lck`. But, in a readonly repository, if that
file does not yet exist, it cannot lock it. And while it will sometimes
work to `chmod +r .git/annex`, the repository might be readonly due to
being owned by another user, or due to being mounted readonly.
So, it seems that the only solution is to use some other file than
`.git/annex/content.lck` as the lock file. The inode sential file
was almost the only option that should always exist. (And if it somehow
does not exist, creating an empty one for locking will be ok.)
Wow, what a hack!
Sponsored-by: Dartmouth College's Datalad project
This has tradeoffs, but is generally a win, and users who it causes git add to
slow down unacceptably for can just disable it again.
It needed to happen in an upgrade, since there are git-annex versions
that do not support it, and using such an old version with a v8
repository with filter.annex.process set will cause bad behavior.
By enabling it in v9, it's guaranteed that any git-annex version that
can use the repository does support it. Although, this is not a perfect
protection against problems, since an old git-annex version, if it's
used with a v9 repository, will cause git add to try to run
git-annex filter-process, which will fail. But at least, the user is
unlikely to have an old git-annex in path if they are using a v9
repository, since it won't work in that repository.
Sponsored-by: Dartmouth College's Datalad project
Capstone of the v10 upgrade process.
Tested with a git-annex drop in a v8 repo that had a local v8 remote.
Upgrading the repo to v10 (with --force) immedaitely caused it to notice
and switch over to v10 locking. Upgrading the remote also caused it to
switch over when operating on the remote.
The InodeCache makes this fairly efficient, just an added stat call per
lock of an object file. After the v10 upgrade, there is no more
overhead.
Sponsored-by: Dartmouth College's Datalad project
Since it's easy to keep supporting v8, using it for a while (eg a few
months) will give users time to upgrade git-annex installations, before
it upgrades their repository to v9.
This commit should be reverted once ready to start upgrading
repositories by default.
Sponsored-by: Dartmouth College's Datalad project
The v10 upgrade should almost be safe now. What remains to be done is
notice when the v10 upgrade has occurred, while holding the shared lock,
and switch to using v10 lock files.
Sponsored-by: Dartmouth College's Datalad project
The upgrade from V9 uses this to avoid an automatic upgrade until 1 year
after the V9 update. It can also be used in future such situations.
Sponsored-by: Dartmouth College's Datalad project
v10 will run 1 year after the upgrade to v9, to give time for any v8
processes to die. Until that point, the v10 upgrade will be tried by
every process but deferred, so added support for deferring upgrades.
The upgrade prevention lock file that will be used by v10 is not yet
implemented, so it does not yet defer.
Sponsored-by: Dartmouth College's Datalad project
Upgrade the shared lock to an exclusive lock, and then delete the
lock file. If there is another process still holding the shared lock,
the first process will fail taking the exclusive lock, and not delete
the lock file; then the other process will later delete it.
Note that, in the time period where the exclusive lock is held, other
attempts to lock the content in place would fail. This is unlikely to be
a problem since it's a short period.
Other attempts to lock the content for removal would also fail in that
time period, but that's no different than a removal failing because
content is locked to prevent removal.
Sponsored-by: Dartmouth College's Datalad project
When dropping content, this was already done after deleting the content
file, but the lock file prevents deleting the directories. So, try the
deletion again.
This does mean there's a small added overhead of a failed rmdir().
Sponsored-by: Dartmouth College's Datalad project
This seems to be the best that can be done to avoid forever accumulating
the new content lock files, while being fully safe.
This is fixing code paths that have lingered unused since direct mode!
And direct mode seems to have been buggy in this area, since the content
lock file was deleted on unlock. But with a shared lock, there could be
another process that also had the lock file locked, and deleting it
invalidates that lock.
So, the lock file cannot be deleted after a shared lock. At least, not
wihout taking an exclusive lock first.. which I have not pursued yet but may.
After an exclusive lock, the lock file can be deleted. But there is
still a potential race, where the exclusive lock is held, and another
process gets the file open, just as the exclusive lock is dropped and
the lock file is deleted. That other process would be left with a file
handle it can take a shared lock of, but with no effect since the file
is deleted. Annex.Transfer also deletes lock files, and deals with this
same problem by using checkSaneLock, which is how I've dealt with it
here.
Sponsored-by: Dartmouth College's Datalad project
Now the content lock files are used in v9. However, I am not yet certian
they are correct. In particular, lockContentUsing deletes
the content lock file on unlock. But what if there's a shared lock
by another process? That seems like it would discard that lock too!
(Windows seems like it would not have the same problem, because as the
comment in there says, "Can't delete a locked file on Windows".
So if another process has a shared lock, removing it presumably fails.)
Sponsored-by: Dartmouth College's Datalad project
Seems to work ok. Unsure yet about the actual locking changes being
correct.
This is not the end of the story with upgrades, because it is unsafe for
this upgrade as implemented to run in a repository where an old
git-annex process is already running. The old process would use the old
locking method, and not notice files locked by the new, and this could
result in data loss. This problem will need to be dealt with before this
branch is suitable for merging.
Sponsored-by: Dartmouth College's Datalad project
Windows has always used a separate lock file, but on unix, the content
file itself was locked, and in v9 that changes to also use a separate
lock file.
This needs to be tested more. Eg, what happens after dropping a file;
does the the content lock file get deleted too, or linger around?
Sponsored-by: Dartmouth College's Datalad project
v9 will not need to write to annex content files in order to lock them,
so freezeContent removes the write bit in a shared repository, the same
as in any other repository.
checkContentWritePerm makes sure that the write perm is not set, which
will let git-annex fsck fix up the permissions. Upgrading to v9
will need to fix the permissions as well, but it seems likely there will
be situations where the user git-annex is running an upgrade as cannot,
so it will have to leave the write bit set. In such a case, git-annex
fsck can fix it later.
Sponsored-by: Dartmouth College's Datalad project
This is the start of v9, but it's currently identical to v8, and v8 is
not upgraded to it. git-annex upgrade will upgrade to v9 with this
change.
Sponsored-by: Dartmouth College's Datalad project
Recover from corrupted content being received from a git remote due eg to a
wire error, by deleting the temporary file when it fails to verify. This
prevents a retry from failing again.
Reversion introduced in version 8.20210903, when incremental verification
was added.
Only the git remote seems to be affected, although it is certianly
possible that other remotes could later have the same issue. This only
affects things passed to getViaTmp that return (False, UnVerified) due to
verification failing. As far as getViaTmp can tell, that could just as well
mean that the transfer failed in a way that would resume, so it cannot
delete the temp file itself. Remote.Git and P2P.Annex use getViaTmp internally,
while other remotes do not, which is why only it seems affected.
A better fix perhaps would be to improve the types of the callback
passed to getViaTmp, so that some other value could be used to indicate
the state where the transfer succeeded but verification failed.
Sponsored-by: Boyd Stephen Smith Jr.
Before it would pick one at random, though preferring ones that were not
dead over dead ones.
Now, if one is dead and the other not, it will use the non-dead one. But if
both are not dead, or both dead, it will error out, suggesting the user
clarify what they want to enable.
Sponsored-by: Luke Shumaker on Patreon
Capstone to this feature. Any transitions that have been performed on an
unmerged remote ref but not on the local git-annex branch, or vice-versa
have to be applied on the fly when reading files.
Sponsored-by: Dartmouth College's Datalad project
It would be difficult to make Annex.Branch.files query the unmerged
git-annex branches. Might be possible, similar to what was discussed in
7f6b2ca49c but again I decided to make it
not do anything in that situation to start with before adding such a
complicated thing.
git-annex info uses it when getting info about a repostory. The choices
were to make that fail with an error, or display the info it can, and
change the output slightly for the bits of info it cannot access. While
that is a behavior change, and I want to avoid any behavior changes due
to unmerged git-annex branches in a read-only repo, displaying a message
that is not a number seems unlikely to break anything that was consuming
a number, any worse than throwing an exception would. Probably.
Also git-annex unused --from origin is made to throw an error, but
it would fail later anyway when trying to write to the unused log files.
Sponsored-by: Dartmouth College's Datalad project
This makes --all error out in that situation. Which is better than
ignoring information from the branches.
To really handle the branches right, overBranchFileContents would need
to both query all the branches and union merge file contents
(or perhaps not provide any file content), as well as diffing between
branches to find files that are only present in the unmerged branches.
And also, it would need to handle transitions..
Sponsored-by: Dartmouth College's Datalad project
The way precaching works, it can't merge in information from those
branches efficiently, so just disable it and fall back to
Annex.Branch.get in order to get the correct information.
Sponsored-by: Dartmouth College's Datalad project
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
So that eg, addurl of several large files that take time to download will
update the index for each file, rather than deferring the index updates to
the end.
In cases like an add of many smallish files, where a new file is being
added every few seconds. In that case, the queue will still build up a
lot of changes which are flushed at once, for best performance. Since
the default queue size is 10240, often it only gets flushed once at the
end, same as before. (Notice that updateQueue updated _lastchanged
when adding a new item to the queue without flushing it; that is
necessary to avoid it flushing the queue every 5 minutes in this case.)
But, when it takes more than a 5 minutes to add a file, the overhead of
updating the index immediately is probably small, so do it after each
file. This avoids git-annex potentially taking a very very long time
indeed to stage newly added files, which can be annoying to the user who
would like to get on with doing something with the files it's already
added, eg using git mv to rename them to a better name.
This is only likely to cause a problem if it takes say, 30 seconds to
update the index; doing an extra 30 seconds of work after every 5
minute file add would be less optimal. Normally, updating the index takes
significantly less time than that. On a SSD with 100k files it takes
less than 1 second, and the index write time is bound by disk read and
write so is not too much worse on a hard drive. So I hope this will not
impact users, although if it does turn out to, the time limit could be
made configurable.
A perhaps better way to do it would be to have a background worker
thread that wakes up every 60 seconds or so and flushes the queue.
That is made somewhat difficult because the queue can contain Annex
actions and so this would add a new source of concurrency issues.
So I'm trying to avoid that approach if possible.
Sponsored-by: Erik Bjäreholt on Patreon
Was "failed to generate a key" when key generation did not fail
(it never does anymore) but the actual problem was it failed to stat
the source file, perhaps due to it being deleted while the key was being
generated.
A user reported this, in a comment I followed up on in
262400fe04, although I don't know
what they did to trigger the error message.
This fixes a FD leak when annex.pidlock is set and -J is used. Also, it
fixes bugs where the pid lock file got deleted because one thread was
done with it, while another thread was still holding it open.
The LockPool now has two distinct types of resources,
one is per-LockHandle and is used for file Handles, which get closed
when the associated LockHandle is closed. The other one is per lock
file, and gets closed when no more LockHandles use that lock file,
including other shared locks of the same file.
That latter kind is used for the pid lock file, so it's opened by the
first thread to use a lock, and closed when the last thread closes a lock.
In practice, this means that eg git-annex get of several files opens and
closes the pidlock file a few times per file. While with -J5 it will open
the pidlock file, process a number of files, until all the threads happen to
finish together, at which point the pidlock file gets closed, and then
that repeats. So in either case, another process still gets a chance to
take the pidlock.
registerPostRelease has a rather intricate dance, there are fine-grained
STM locks, a STM lock of the pidfile itself, and the actual pidlock file
on disk that are all resolved in stages by it.
Sponsored-by: Dartmouth College's Datalad project
This locking has been missing from the beginning of annex.pidlock.
It used to be possble, when two threads are doing conflicting things,
for both to run at the same time despite using locking. Seems likely
that nothing actually had a problem, but it was possible, and this
eliminates that possible source of failure.
Sponsored-by: Dartmouth College's Datalad project
This version of git -- or its new default "ort" resolver -- handles such
a conflict by staging two files, one with the original name and the other
named file~ref. Use unmergedSiblingFile when the latter is detected.
(It doesn't do that when the conflict is between a directory and a file
or symlink though, so see previous commit for how that case is handled.)
The sibling file has to be deleted separately, because cleanConflictCruft
may not delete it -- that only handles files that are annex links,
but the sibling file may be the non-annexed file side of the conflict.
The graftin code had assumed that, when the other side of a conclict
is a symlink, the file in the work tree will contain the non-annexed
content that we want it to contain. But that is not the case with the new
git; the file may be the annex link and needs to be replaced with the
content, while the annex link will be written as a -variant file.
(The weird doesDirectoryExist check in graftin turns out to still be
needed, test suite failed when I tried to remove it.)
Test suite passes with new git with ort resolver default. Have not tried it
with old git or other defaults.
Sponsored-by: Noam Kremen on Patreon
Bugfix: When -J was enabled, getting files leaked a ever-growing number of
git cat-file processes.
(Since commit dd39e9e255)
The leak happened when mergeState called stopNonConcurrentSafeCoProcesses.
While stopNonConcurrentSafeCoProcesses usually manages to stop everything,
there was a race condition where cat-file processes were leaked. Because
catFileStop modifies Annex.catfilehandles in a non-concurrency safe way,
and could clobber modifications made in between. Which should have been ok,
since originally catFileStop was only used at shutdown.
Note the comment on catFileStop saying it should only be used when nothing
else is using the handles. It would be possible to make catFileStop
race-safe, but it should just not be used in a situation where a race is
possible. So I didn't bother.
Instead, the fix is just not to stop any processes in mergeState. Because
in order for mergeState to be called, dupState must have been run, and it
enables concurrency mode, stops any non-concurrent processes, and so all
processes that are running are concurrency safea. So there is no need to
stop them when merging state. Indeed, stopping them would be extra work,
even if there was not this bug.
Sponsored-by: Dartmouth College's Datalad project
When non-concurrent git coprocesses have been started, setConcurrency
used to not stop them, and so could leak processes when enabling
concurrency, eg when forkState is called.
I do not think that ever actually happened, given where setConcurrency
is called. And it probably would only leak one of each process, since it
never downgrades from concurrent to non-concurrent.
Based on my earlier benchmark, I have a rough cost model for how
expensive it is for git-annex smudge to be run on a file, vs
how expensive it is for a gigabyte of a file's content to be read and
piped through to filter-process.
So, using that cost model, it can decide if using filter-process will
be more or less expensive than running the smudge filter on the files to
be restaged.
It turned out to be *really* annoying to temporarily disable
filter-process. I did find a way, but urk, this is horrible. Notice
that, if it's interrupted with it disabled, it will remain disabled
until the next time restagePointerFile runs. Which could be some time
later. If the user runs `git add` or `git checkout` on a lot of small
files before that, they will see slower than expected performance.
(This commit also deletes where I wrote down the benchmark results
earlier.)
Sponsored-by: Noam Kremen on Patreon
This reverts commit afe327ac49.
Unfortunately, disabling it by setting it to "" does not work, git
then ignores filter.annex.smudge/clean, and does not pass files through
git-annex at all.
I don't think there is a way to temporarily disable this git config
from the git command line. Which seems like a bug in git.
So, it may be more expensive than anticipated to enable
filter.annex.process, since git checkout etc will pipe all annexed files
being checked out through it.
This means git will run git-annex smudge --clean once per file that is
restaged, which can be slow. But probably *not* as slow as git feeding
all the content of annexed files you've gotten through a pipe to
git-annex filter-process.
The only time this is probably not ideal is after a drop of a bunch of
files, when filter-process would be faster.
This makes it be displayed in the error-messages field with
--json-error-messages. And with --quiet, it will let it be displayed,
which makes sense because it's telling the user why what they requested
to do has failed to happen.
This opens the potential for the object file to be in place but
git-annex is interrupted before it can freeze it. git-annex fsck already
fixes that situation, which can also occur when lockContentForRemoval
thaws content.
Also improve comment to not be Windows-specific.
This change better supports Windows ACL management using
annex.freezecontent-command and annex.thawcontent-command and matches
the behaviour of adding an unlocked file.
By calling freezeContent after the file has moved into the annex,
the file's delete permission can be denied. If the file's delete
permission is denied before moving into the annex, the file cannot
be moved or deleted. If the file's delete permission is not denied after
moving into the annex, it will likely inherit a grant for the delete
permission which allows it to be deleted irrespective of the permissions
of the parent directory.
Fix bug that caused stale git-annex branch information to read when
annex.private or remote.name.annex-private is set.
The private journal file should not prevent reading more current
information from the git-annex branch, but used to.
Note that, overBranchFileContents has to do additional work now, when
there's a private journal file, it reads from the branch redundantly
and more slowly.
Sponsored-by: Jack Hill on Patreon
This improves the borg special remote memory usage, by
letting it only load one archive's worth of filenames into memory at a
time, and building up a larger tree out of the chunks.
When a borg repository has many archives, git-annex could easily OOM
before. Now, it will use only memory proportional to the number of
annexed keys in an archive.
Minor implementation wart: Each new chunk re-opens the content
identifier database, and also a new vector clock is used for each chunk.
This is a minor innefficiency only; the use of continuations makes
it hard to avoid, although putting the database handle into a Reader
monad would be one way to fix it.
It may later be possible to extend the ImportableContentsChunkable
interface to remotes that are not third-party populated. However, that
would perhaps need an interface that does not use continuations.
The ImportableContentsChunkable interface currently does not allow
populating the top of the tree with anything other than subtrees. It
would be easy to extend it to allow putting files in that tree, but borg
doesn't need that so I left it out for now.
Sponsored-by: Noam Kremen on Patreon
This adds the overhead of a copy when serializing and deserializing keys.
I have not benchmarked much, but runtimes seem barely changed at all by that.
When a lot of keys are in memory, it improves memory use.
And, it prevents keys sometimes getting PINNED in memory and failing to GC,
which is a problem ByteString has sometimes. In particular, git-annex sync
from a borg special remote had that problem and this improved its memory
use by a large amount.
Sponsored-by: Shae Erisson on Patreon
Commit 4bf7940d6b introduced this
problem, but was otherwise doing a good thing. Problem being
that fileRef "/foo" used to return ":./foo", which was actually wrong,
but as long as there was no foo in the local repository, catKey
could operate on it without crashing. After that fix though, fileRef
would return eg "../../foo", resulting in fileRef returning
":./../../foo", which will make git cat-file crash since that's
not a valid path in the repo.
Fix is simply to make fileRef detect paths outside the repo and return
Nothing. Then catKey can be skipped. This needed several bugfixes to
dirContains as well, in previous commits.
In Command.Smudge, this led to needing to check for Nothing. That case
should actually never happen, because the fileoutsiderepo check will
detect it earlier.
Sponsored-by: Brock Spratlen on Patreon
This method avoids breaking test_readonly. Just check if the dest file
exists, and avoid CoW probing when it does, so when CoW probing fails,
it can resume where the previous non-CoW copy left off.
If CoW has been probed already to work, delete the dest file
since a CoW copy will presumably work. It seems like it would be almost
as good to just skip CoW copying in this case too, but consider that the
dest file might have started to be copied from some other remote, not
using CoW, but CoW has been probed to work to copy from the current
place.
Sponsored-by: Dartmouth College's Datalad project
commit 63d508e885 broke test_readonly.
When a local git remote is readonly, tryCopyCoW run to copy a file
from it failed at withOtherTmp.
Sponsored-by: Dartmouth College's Datalad project
This avoids starting one process when only the other one is needed.
Eg in git-annex smudge --clean, this reduces the total number of
cat-file processes that are started from 4 to 2.
The only performance penalty is that when both are needed, it has to do
twice as much work to maintain the two Maps. But both are very small,
consisting of 1 or 2 items, so that work is negligible.
Sponsored-by: Dartmouth College's Datalad project
RemoteGitConfig parsing looks for annex.bwlimit when a remote
does not have a per-remote config for it, so no need for a separate
gobal config.
Sponsored-by: Svenne Krap on Patreon
RemoteGitConfig parsing looks for annex.stalldetection when a remote
does not have a per-remote config for it, so no need for a separate
gobal config.
Sponsored-by: Noam Kremen on Patreon
Probably this fixes a reversion, but I don't know what version broke it.
This does use withOtherTmp for a temp file that could be quite large.
Though albeit a reflink copy that will not actually take up any space
as long as the file it was copied from still exists. So if the copy cow
succeeds but git-annex is interrupted just before that temp file gets
renamed into the usual .git/annex/tmp/ location, there is a risk that
the other temp directory ends up cluttered with a larger temp file than
later. It will eventually be cleaned up, and the changes of this being
a problem are small, so this seems like an acceptable thing to do.
Sponsored-by: Shae Erisson on Patreon
Added annex.bwlimit and remote.name.annex-bwlimit config that works for git
remotes and many but not all special remotes.
This nearly works, at least for a git remote on the same disk. With it set
to 100kb/1s, the meter displays an actual bandwidth of 128 kb/s, with
occasional spikes to 160 kb/s. So it needs to delay just a bit longer...
I'm unsure why.
However, at the beginning a lot of data flows before it determines the
right bandwidth limit. A granularity of less than 1s would probably improve
that.
And, I don't know yet if it makes sense to have it be 100ks/1s rather than
100kb/s. Is there a situation where the user would want a larger
granularity? Does granulatity need to be configurable at all? I only used that
format for the config really in order to reuse an existing parser.
This can't support for external special remotes, or for ones that
themselves shell out to an external command. (Well, it could, but it
would involve pausing and resuming the child process tree, which seems
very hard to implement and very strange besides.) There could also be some
built-in special remotes that it still doesn't work for, due to them not
having a progress meter whose displays blocks the bandwidth using thread.
But I don't think there are actually any that run a separate thread for
downloads than the thread that displays the progress meter.
Sponsored-by: Graham Spencer on Patreon
This should complete the fix started in
6329997ac4, fixing the actual cause of the
test suite failure this time.
Sponsored-by: Dartmouth College's Datalad project
* When downloading urls fail, explain which urls failed for which
reasons.
* web: Avoid displaying a warning when downloading one url failed
but another url later succeeded.
Some other uses of downloadUrl use urls that are effectively internal use,
and should not all be displayed to the user on failure. Eg, Remote.Git
tries different urls where content could be located depending on how the
remote repo is set up. Exposing those urls to the user would lead to wild
goose chases. So had to parameterize it to control whether it displays urls
or not.
A side effect of this change is that when there are some youtube urls
and some regular urls, it will try regular urls first, even if the
youtube urls are listed first. This seems like an improvement if
anything, but in any case there's no defined order of urls that it's
supposed to use.
Sponsored-by: Dartmouth College's Datalad project
This fixes a reversion caused by a99a84f342,
when git-annex init is run as root on a FAT filesystem mounted with
hdiutil on OSX. Such a mount point has file mode 777 for everything and
it cannot be changed. The existing crippled filesystem test tried to
write to a file after removing write bit, but that test does not run as
root (since root can write to unwritable files). So added a check of the
write permissions of the file, after attempting to remove them.
Sponsored-by: Dartmouth College's Datalad project
This is to track down what file in .git/annex/ is being written to via a
temp file when the repository is read-only.
Sponsored-by: Dartmouth College's Datalad project
And fail with an informative message.
I don't think ACLs can prevent removing the write bit, but I'm not sure,
so kept it mentioning them as a possibility.
Should git-annex lock also check if the write bits are able to be removed?
Maybe, but the case I know about with xattrs involves cp -a copying NFS
xattrs, and it's the copy of the file that is the problem. So when locking
a file, I guess it will not be the copy.
Sponsored-by: Dartmouth College's Datalad project
It would be better if the Arbitrary instance avoided generating impossible
filenames like "foo/c:bar", but proably this is the only place that splits
the file from the directory and then uses the file without the directory..
At least on the quickcheck properties.
Sponsored-by: Svenne Krap on Patreon
Except when configuration makes curl be used. It did not seem worth
trying to tail the file when curl is downloading.
But when an interrupted download is resumed, it does not read the whole
existing file to hash it. Same reason discussed in
commit 7eb3742e4b76d1d7a487c2c53bf25cda4ee5df43; that could take a long
time with no progress being displayed. And also there's an open http
request, which needs to be consumed; taking a long time to hash the file
might cause it to time out.
Also in passing implemented it for git and external special remotes when
downloading from the web. Several others like S3 are within striking
distance now as well.
Sponsored-by: Dartmouth College's DANDI project
IncrementalVerifier moved to Utility.Hash, which will let Utility.Url
use it later.
It's perhaps not really specific to hashing, but making a separate
module just for the data type seemed unncessary.
Sponsored-by: Dartmouth College's DANDI project
This fixes the recent reversion that annex.verify is not honored,
because retrieveChunks was passed RemoteVerify baser, but baser
did not have export/import set up.
Sponsored-by: Dartmouth College's DANDI project
Added fileRetriever', which will let the remaining special remotes
eventually also support incremental verify.
Sponsored-by: Dartmouth College's DANDI project
As happens when using the directory special remote, gitlfs, webdav, and
S3. But not external, adb, gcrypt, hook, or rsync.
Sponsored-by: Dartmouth College's DANDI project
Now it's run in VerifyStage.
I thought about keeping the file handle open, and resuming reading where
tailVerify left off. But that risks leaking open file handles, until the
GC closes them, if the deferred verification does not get resumed. Since
that could perhaps happen if there's an exception somewhere, I decided
that was too unsafe.
Instead, re-open the file, seek, and resume.
Sponsored-by: Dartmouth College's DANDI project
Wait for the file to get modified, not only opened. This way, if a
remote does not support resuming, and opens a new file over top of the
existing file, it will wait until that remote starts writing, and open
the file it's writing to, not the old file.
Sponsored-by: Dartmouth College's DANDI project
I saw this:
.git/annex/tmp/SHA256E-s1234376--5ba8e06e0163b217663907482bbed57684d7188024155ddc81da0710dfd2687d: openBinaryFile: resource busy (file is locked)
guess catching IO exceptions did not catch that one.
Not yet used, but this will let all remotes verify incrementally if it's
acceptable to pay the performance price. See comment for details of when
it will perform badly. I anticipate using this for all special remotes
that use fileRetriever. Except perhaps for a few like GitLFS that could
feed the incremental verifier themselves despite using that.
Sponsored-by: Dartmouth College's DANDI project
This eliminates the distinction between decodeBS and decodeBS', encodeBS
and encodeBS', etc. The old implementation truncated at NUL, and the
primed versions had to do extra work to avoid that problem. The new
implementation does not truncate at NUL, and is also a lot faster.
(Benchmarked at 2x faster for decodeBS and 3x for encodeBS; more for the
primed versions.)
Note that filepath-bytestring 1.4.2.1.8 contains the same optimisation,
and upgrading to it will speed up to/fromRawFilePath.
AFAIK, nothing relied on the old behavior of truncating at NUL. Some
code used the faster versions in places where I was sure there would not
be a NUL. So this change is unlikely to break anything.
Also, moved s2w8 and w82s out of the module, as they do not involve
filesystem encoding really.
Sponsored-by: Shae Erisson on Patreon
* Deal with clock skew, both forwards and backwards, when logging
information to the git-annex branch.
* GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1)
rather than needing to be advanced each time a new change is made.
* Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex.
When changing a file in the git-annex branch, the vector clock to use is now
determined by first looking at the current time (or GIT_ANNEX_VECTOR_CLOCK
when set), and comparing it to the newest vector clock already in use in
that file. If a newer time stamp was already in use, advance it forward by
a second instead.
When the clock is set to a time in the past, this avoids logging with
an old timestamp, which would risk that log line later being ignored in favor
of "newer" line that is really not newer.
When a log entry has been made with a clock that was set far ahead in the
future, this avoids newer information being logged with an older timestamp
and so being ignored in favor of that future-timestamped information.
Once all clocks get fixed, this will result in the vector clocks being
incremented, until finally enough time has passed that time gets back ahead
of the vector clock value, and then it will return to usual operation.
(This latter situation is not ideal, but it seems the best that can be done.
The issue with it is, since all writers will be incrementing the last
vector clock they saw, there's no way to tell when one writer made a write
significantly later in time than another, so the earlier write might
arbitrarily be picked when merging. This problem is why git-annex uses
timestamps in the first place, rather than pure vector clocks.)
Advancing forward by 1 second is somewhat arbitrary. setDead
advances a timestamp by just 1 picosecond, and the vector clock could
too. But then it would interfere with setDead, which wants to be
overrulled by any change. So it could use 2 picoseconds or something,
but that seems weird. It could just as well advance it forward by a
minute or whatever, but then it would be harder for real time to catch
up with the vector clock when forward clock slew had happened.
A complication is that many log files contain several different peices of
information, and it may be best to only use vector clocks for the same peice
of information. For example, a key's location log file contains
InfoPresent/InfoMissing for each UUID, and it only looks at the vector
clocks for the UUID that is being changed, and not other UUIDs.
Although exactly where the dividing line is can be hard to determine.
Consider metadata logs, where a field "tag" can have multiple values set
at different times. Should it advance forward past the last tag?
Probably. What about when a different field is set, should it look at
the clocks of other fields? Perhaps not, but currently it does, and
this does not seems like it will cause any problems.
Another one I'm not entirely sure about is the export log, which is
keyed by (fromuuid, touuid). So if multiple repos are exporting to the
same remote, different vector clocks can be used for that remote.
It looks like that's probably ok, because it does not try to determine
what order things occurred when there was an export conflict.
Sponsored-by: Jochen Bartl on Patreon
