Check just before running update-index if the worktree file's content is
still the same, don't update it when it's been modified. This narrows
the race window a lot, from possibly minutes or hours, to seconds or
less.
(Use replaceFile so that the worktree update happens atomically,
allowing the InodeCache of the new worktree file to itself be gathered
w/o any other race.)
This doesn't eliminate the race; it can still occur in the window before
update-index runs. When annex.queue is large, a lot of files will be
statted by the checks, and so the window may still be large enough to be a
problem.
When only a few files are being processed, the window is as small as it
is in the race where a modification gets overwritten by git-annex when
it updates the worktree. Or maybe as small as whatever race git
checkout/pull/merge may have when the worktree gets modified during it.
Still, I've kept a todo about this race.
This commit was supported by the NSF-funded DataLad project.
Use git update-index --refresh, since it's a little bit more
efficient and the user can be told to run it if a locked index prevents
git-annex from running it.
This also fixes the problem where an annexed file was deleted in the index
and a get of another file that uses the same key caused the index update to
add back the deleted file. update-index will not add back the deleted file.
Documented in tips/unlocked_files.mdwn the gotcha that the index update
may conflict with other operations. I can't see any way to possibly avoid
that conflict.
One new todo about a race that causes a modification to be accidentially
staged.
Note that the assistant only flushes the git command queue when it
commits a modification. I have not tested the assistant with v6 unlocked
files, but assume most users of the assistant won't care if the index
shows a file as modified for a while.
This commit was supported by the NSF-funded DataLad project.
After updating the worktree for an add/drop, update git's index, so git
status will not show the files as modified.
What actually happens is that the index update removes the inode
information from the index. The next git status (or similar) run
then has to do some work. It runs the clean filter.
So, this depends on the clean filter being reasonably fast and on git
not leaking memory when running it. Both problems were fixed in
a96972015d, but only for git 2.5. Anyone
using an older git will see very expensive git status after an add/drop.
This uses the same git update-index queue as other parts of git-annex, so
the actual index update is fairly efficient. Of course, updating the index
does still have some overhead. The annex.queuesize config will control how
often the index gets updated when working on a lot of files.
This is an imperfect workaround... Added several todos about new
problems this workaround causes. Still, this seems a lot better than the
old behavior.
This commit was supported by the NSF-funded DataLad project.
v6 add: Take advantage of improved SIGPIPE handler in git 2.5 to speed up
the clean filter by not reading the file content from the pipe. This also
avoids git buffering the whole file content in memory.
When built with an older git, still consumes stdin. If built with a newer
git and used with an older one, it breaks, but that's acceptable --
checking the git version every time would make repeated smudge runs slow.
This commit was supported by the NSF-funded DataLad project.
When --batch is used with matching options like --in, --metadata, etc, only
operate on the provided files when they match those options. Otherwise, a
blank line is output in the batch protocol.
Affected commands: find, add, whereis, drop, copy, move, get
In the case of find, the documentation for --batch already said it honored
the matching options. The docs for the rest didn't, but it makes sense to
have them honor them. While this is a behavior change, why specify the
matching options with --batch if you didn't want them to apply?
Note that the batch output for all of the affected commands could
already output a blank line in other cases, so batch users should
already be prepared to deal with it.
git-annex metadata didn't seem worth making support the matching options,
since all it does is output metadata or set metadata, the use cases for
using it in combination with the martching options seem small. Made it
refuse to run when they're combined, leaving open the possibility for later
support if a use case develops.
This commit was sponsored by Brett Eisenberg on Patreon.
Added getStaged, to get the versions of git-annex branch files staged in its
index, and use during transitions so the result of merging sibling branches
is used.
The catFileStop in performTransitionsLocked is absolutely necessary,
without that the bug still occurred, because git cat-file was already
running and was looking at the old index file.
Note that getLocal still has cat-file look at the git-annex branch, not the
index. It might be faster if it looked at the index, but probably only
marginally so, and I've not benchmarked it to see if it's faster at all. I
didn't want to change unrelated behavior as part of this bug fix. And as
the need for catFileStop shows, using the index file has added
complications.
Anyway, it still seems fine for getLocal to look at the git-annex branch,
because normally the index file is updated just before the git-annex branch
is committed, and so they'll contain the same information. It's only during
a transition that the two diverge.
This commit was sponsored by Paul Walmsley in honor of Mark Phillips.
It was sorting by uuid, rather than cost!
Avoid future bugs of this kind by changing the Ord to primarily compare
by cost, with uuid only used when the cost is the same.
This commit was supported by the NSF-funded DataLad project.
Added annex.commitmessage config that can specify a commit message for the
git-annex branch instead of the usual "update".
This commit was supported by the NSF-funded DataLad project.
Useful for dropping old objects from cache repositories.
But also, quite a genrally useful thing to have..
Rather than imitiating find's -atime and other options, all of which are
pretty horrible to use, I made this match files accessed within a time
period, using the same duration format used by git-annex schedule and
--limit-time
In passing, changed the --limit-time option parser to parse the
duration, instead of having it later throw an error.
This commit was supported by the NSF-funded DataLad project.
Added remote.name.annex-speculate-present config that can be used to
make cache remotes.
Implemented it in Remote.keyPossibilities, which is used by the
get/move/copy/mirror commands, and nothing else. This way, things like
whereis will not show content that's speculatively present.
The assistant and sync --content were not using Remote.keyPossibilities,
and were changed to use it.
The efficiency hit should be small; Remote.keyPossibilities is only
used before transferring a file, which is the expensive operation.
And, it's only doing one lookup of the remoteList and a very cheap
filter over it.
Note that, git-annex still updates the location log when copying content
to a remote with annex-speculate-present set. In this case, the location
tracking will indicate that content is present in the remote. This may
not be wanted for caches, or may not be a real problem for them. TBD.
This commit was supported by the NSF-funded DataLad project.
