There was no good reason for it to be using annexLocationsNonBare,
and exporttree=yes annexobjects=yes is going to use annexLocationsBare,
so this should as well for consistency.
Since all returned ExportLocations are tried when retrieving objects,
this won't break backwards compatability.
Added Maybe POSIXTime to SafeDropProof, which gets set when the proof is
based on a LockedCopy. If there are several LockedCopies, it uses the
closest expiry time. That is not optimal, it may be that the proof
expires based on one LockedCopy but another one has not expired. But
that seems unlikely to really happen, and anyway the user can just
re-run a drop if it fails due to expiry.
Pass the SafeDropProof to removeKey, which is responsible for checking
it for expiry in situations where that could be a problem. Which really
only means in Remote.Git.
Made Remote.Git check expiry when dropping from a local remote.
Checking expiry when dropping from a P2P remote is not yet implemented.
P2P.Protocol.remove has SafeDropProof plumbed through to it for that
purpose.
Fixing the remaining 2 build warnings should complete this work.
Note that the use of a POSIXTime here means that if the clock gets set
forward while git-annex is in the middle of a drop, it may say that
dropping took too long. That seems ok. Less ok is that if the clock gets
turned back a sufficient amount (eg 5 minutes), proof expiry won't be
noticed. It might be better to use the Monotonic clock, but that doesn't
advance when a laptop is suspended, and while there is the linux
Boottime clock, that is not available on other systems. Perhaps a
combination of POSIXTime and the Monotonic clock could detect laptop
suspension and also detect clock being turned back?
There is a potential future flag day where
p2pDefaultLockContentRetentionDuration is not assumed, but is probed
using the P2P protocol, and peers that don't support it can no longer
produce a LockedCopy. Until that happens, when git-annex is
communicating with older peers there is a risk of data loss when
a ssh connection closes during LOCKCONTENT.
This is to avoid inserting a cluster uuid into the location log when
only dead nodes in the cluster contain the content of a key.
One reason why this is necessary is Remote.keyLocations, which excludes
dead repositories from the list. But there are probably many more.
Implementing this was challenging, because Logs.Location importing
Logs.Cluster which imports Logs.Trust which imports Remote.List resulted
in an import cycle through several other modules.
Resorted to making Logs.Location not import Logs.Cluster, and instead
it assumes that Annex.clusters gets populated when necessary before it's
called.
That's done in Annex.Startup, which is run by the git-annex command
(but not other commands) at early startup in initialized repos. Or,
is run after initialization.
Note that is Remote.Git, it is unable to import Annex.Startup, because
Remote.Git importing Logs.Cluster leads the the same import cycle.
So ensureInitialized is not passed annexStartup in there.
Other commands, like git-annex-shell currently don't run annexStartup
either.
So there are cases where Logs.Location will not see clusters. So it won't add
any cluster UUIDs when loading the log. That's ok, the only reason to do
that is to make display of where objects are located include clusters,
and to make commands like git-annex get --from treat keys as being located
in a cluster. git-annex-shell certainly does not do anything like that,
and I'm pretty sure Remote.Git (and callers to Remote.Git.onLocalRepo)
don't either.
Using the usual url download machinery even allows these urls to need
http basic auth, which is prompted for with git-credential. Which opens
the possibility for urls that contain a secret to be used, eg the cipher
for encryption=shared. Although the user is currently on their own
constructing such an url, I do think it would work.
Limited to httpalso for now, for security reasons. Since both httpalso
(and retrieving this very url) is limited by the usual
annex.security.allowed-ip-addresses configs, it's not possible for an
attacker to make one of these urls that sets up a httpalso url that
opens the garage door. Which is one class of attacks to keep in mind
with this thing.
It seems that there could be either a git-config that allows other types
of special remotes to be set up this way, or special remotes could
indicate when they are safe. I do worry that the git-config would
encourage users to set it without thinking through the security
implications. One remote config might be safe to access this way, but
another config, for one with the same type, might not be. This will need
further thought, and real-world examples to decide what to do.
This is just a good idea, I think. But it fixes this specific bug:
With buggy git version 2.45.1, git clone from an annex:: url, which has
a git-annex branch in it. Then in the repository, git fetch. That
left .git/annex/objects/ populated with bundles, since it did not clean
up. So later using git-annex failed to autoinit.
And use it to set annex-config-uuid in git config. This makes
using the origin special remote work after cloning.
Without the added Logs.Remote.configSet, instantiating the remote will
look at the annex-config-uuid's config in the remote log, which will be
empty, and so it will fail to find a special remote.
The added deletion of files in the alternatejournaldir is just to make
100% sure they don't get committed to the git-annex branch. Now that
they contain things that definitely should not be committed.
With the directory special remote, manifest objects uploaded by
git-remote-annex were mode 600. This prevented accessing them
from a httpalso special remote, for example.
The directory special remote just copies the file perms. Which is fine
except in this case the file perms were wrong.
An incremental push that gets converted to a full push due to this
config results in the inManifest having just one bundle in it, and the
outManifest listing every other bundle. So it actually takes up more
space on the special remote. But, it speeds up clone and fetch to not
have to download a long series of bundles for incremental pushes.
cleanupInitialization gets run when an exception is thrown, so needs to
avoid throwing exceptions itself, as that would hide the error message
that the user needs to see.
When exporttree=yes is also set. Probably it would also be possible to
support ones with only importtree=yes, by enabling exporttree=yes for
the remote only when using git-remote-annex, but let's keep this
simple... I'm not sure what gets recorded in .git/annex/ state
differently in the two cases that might cause a problem when doing that.
Note that the full annex:: urls generated and displayed for such a
remote omit the importree=yes. Which is ok, cloning from such an url
uses an exporttree=remote, but the git-annex branch doesn't get written
by this program, so once the real config is available from the git-annex
branch, it will still function as an importree=yes remote.
This git bug also broke git-lfs, and I am confident it will be reverted
in the next release.
For now, cloning from an annex:: url wastes some bandwidth on the next
pull by not caching bundles locally.
If git doesn't fix this in the next version, I'd be tempted to rethink
whether bundle objects need to be cached locally. It would be possible to
instead remember which bundles have been seen and their heads, and
respond to the list command with the heads, and avoid unbundling them
agian in fetch. This might even be a useful performance improvement in
the latter case. It would be quite a complication to a currently simple
implementation though.
This fixes pushing a new ref that is the same as something already
pushed. In findotherprereq, it compares two shas, which didn't work when
one is actually not a sha but a ref.
This is one of those cases where Sha being an alias for Ref makes it
hard to catch mistakes. One of these days those need to be
differentiated at the type level, but not today..
Locally record the manifest before uploading it or any bundles,
and read it on the next push. Any bundles from the push that are
not included in the currently being pushed manifest will get added
to the outManifest, and so eventually get deleted.
This deals with an interrupted push that is not resumed and instead
something else is pushed. And it deals with a push race that overwrites
the manifest.
Of course, this can't help if one of those situations is followed by
the local repo being deleted. But that's equivilant to doing a git-annex
copy of a new annexed file to a special remote and then deleting the
special repo w/o pushing. In either case the special remote ends up with
a object in it that git-annex doesn't know about.
This avoids some apparently otherwise unsolveable problems involving
races that resulted in the manifest listing bundles that were deleted.
Removed the annex-max-git-bundles config because it can't actually
result in deleting old bundles. It would still be possible to have a
config that controls how often to do a full push, which would avoid
needing to download too many bundles on clone, as well as needing to
checkpresent too many bundles in verifyManifest. But it would need a
different name and description.
Added a backup manifest key, which is used if the main manifest key is
not present. When uploading a new Manifest, it makes sure that it never
drops one key except when the other key is present.
It's entirely possible for the two manifest keys to get out of sync, due
to races. The main one wins when it's present, it is possible for the
main one being dropped to expose the backup one, which has a different
push recorded.
On push, first try to drop all outManifest keys listed in the current
manifest file, which resumes from an interrupted push that didn't
get a chance to delete those keys.
The new manifest gets its outManifest populated with the keys that were
in the old manifest, plus any of the keys that were unable to be
dropped.
Note that it would be possible for uploadManifest to skip dropping old
keys at all. The old keys would get dropped on the next push. But it
seems better to delete stuff immediately rather than waiting. And the
extra work is limited to push and typically is small.
A remote where dropKey always fails will result in an outManifest that
grows longer and longer. It would be possible to check if the remote
has appendonly = True and avoid populating the outManifest. Of course,
an appendonly remote will grow with every git push anyway. And currently
only Remote.GitLFS sets that, which can't be used as a git-remote-annex
remote anyway.
Implemented alternateJournal, which git-remote-annex
uses to avoid any writes to the git-annex branch while setting up
a special remote from an annex:: url.
That prevents the remote.log from being overwritten with the special
remote configuration from the url, which might not be 100% the same as
the existing special remote configuration.
And it prevents an overwrite deleting of other stuff that was
already in the remote.log.
Also, when the branch was created by git-remote-annex, only delete it
at the end if nothing else has been written to it by another command.
This fixes the race condition described in
797f27ab05, where git-remote-annex
set up the branch and git-annex init and other commands were
run at the same time and their writes to the branch were lost.
I hope to support importtree=yes eventually, but it does not currently
work.
Added remote.<name>.allow-encrypted-gitrepo that needs to be set to
allow using it with encrypted git repos.
Note that even encryption=pubkey uses a cipher stored in the git repo
to encrypt the keys stored in the remote. While it would be possible to
not encrypt the GITBUNDLE and GITMANIFEST keys, and then allow using
encryption=pubkey, it doesn't currently work, and that would be a
complication that I doubt is worth it.
Updating the remote list needs the config to be written to the git-annex
branch, which was not done for good reasons. While it would be possible
to instead use Remote.List.remoteGen without writing to the branch, I
already have a plan to discard git-annex branch writes made by
git-remote-annex, so the simplest fix is to write the config to the
branch.
Sponsored-by: k0ld on Patreon
Put the annex objects in .git/annex/objects/ inside the export remote.
This way, when importing from the remote, they will be filtered out.
Note that, when importtree=yes, content identifiers are used, and this
means that pushing to a remote updates the git-annex branch. Urk.
Will need to try to prevent that later, but I already had a todo about
that for other reasons.
Untested!
Sponsored-By: Brock Spratlen on Patreon
Otherwise, it can be confusing to clone from a wrong url, since it fails
to download a manifest and so appears as if the remote exists but is empty.
Sponsored-by: Jack Hill on Patreon
This will eventually be used to recover from an interrupted fullPush
and drop the old bundle keys it was unable to delete.
Sponsored-by: Luke T. Shumaker on Patreon
Such as annex::?type=foo&...
I accidentially left out the uuid when creating one,
and the result is it appears to clone an empty repository.
So let's guard against that mistake.
Full pushing will probably work, but is untested.
Incremental pushing is not implemented yet.
While a fairly straightforward port of the shell prototype, the details
of exactly how to get the objects to the remote were tricky. And the
prototype did not consider how to deal with partial failures and
interruptions.
I've taken considerable care to make sure it always leaves things in a
consistent state when interrupted or when it loses access to a remote in
the middle of a push.
Sponsored-by: Leon Schuermann on Patreon
It did not seem possible to avoid creating a git-annex branch while
git-remote-annex is running. Special remotes can even store their own
state in it. So instead, if it didn't exist before git-remote-annex
created it, it deletes it at the end.
This does possibly allow a race condition, where git-annex init and
perhaps other git-annex writing commands are run, that writes to the
git-annex branch, at the same time a git-remote-annex process is being
run by git fetch/push with a full annex:: url. Those writes would be
lost. If the repository has already been initialized before
git-remote-annex, that race won't happen. So it's pretty unlikely.
Sponsored-by: Graham Spencer on Patreon
Also support using annex:: urls that specify the whole special remote
config.
Both of these cases need a special remote to be initialized enough to
use it, which means writing to .git/config but not to the git-annex
branch. When cloning, the remote is left set up in .git/config,
so further use of it, by git-annex or git-remote-annex will work. When
using git with an annex:: url, a temporary remote is written to
.git/config, but then removed at the end.
While that's a little bit ugly, the fact is that the Remote interface
expects that it's ok to set git configs of the remote that is being
initialized. And it's nowhere near as ugly as the alternative of making
a temporary git repository and initializing the special remote in there.
Cloning from a repository that does not contain a git-annex branch and
then later running git-annex init is currently broken, although I've
gotten most of the way there to supporting it.
See cleanupInitialization FIXME.
Special shout out to git clone for running gitremote-helpers with
GIT_DIR set, but not in the git repository and with GIT_WORK_TREE not
set. Resulting in needing the fixupRepo hack.
Sponsored-by: unqueued on Patreon
Tested using a manually populated directory special remote.
Pushing is still to be done. So is fetching from special remotes
configured via the annex:: url.
Sponsored-by: Brock Spratlen on Patreon
Not quite there yet.
Also, changed the format of GITBUNDLE keys to use only one '-'
after the UUID. A sha256 does not contain that character, so can just
split at the last one.
Amusingly, the sha256 will probably not actually be verified. A git
bundle contains its own checksums that git uses to verify it. And if
someone wanted to replace the content of a GITBUNDLE object, they
could just edit the manifest to use a new one whose sha256 does verify.
Sponsored-by: Nicholas Golder-Manning
