A bit tricky since Proto doesn't support threads. Rather than adding
threading support to it, ended up using a callback that waits for both
data on a Handle, and incoming messages at the same time.
This commit was sponsored by Denis Dzyubenko on Patreon.
Is content locking needed in the P2P protocol? Based on re-reading
bugs/concurrent_drop--from_presence_checking_failures.mdwn,
I think so: Peers can form cycles, and multiple peers can all be trying
to drop the same content.
So, added content locking to the protocol, with some difficulty.
The implementation is fine as far as it goes, but note the warning
comment for lockContentWhile -- if the connection to the peer is dropped
unexpectedly, the peer will then unlock the content, and yet the local
side will still think it's locked.
To be honest I'm not sure if Remote.Git's lockKey for ssh remotes
doesn't have the same problem. It checks that the
"ssh remote git-annex-shell lockcontent"
process has not exited, but if the connection closes afer that check,
the lockcontent command will unlock it, and yet the local side will
still think it's locked.
Probably this needs to be fixed by eg, making lockcontent catch any
execptions due to the connection closing, and in that case, wait a
significantly long time before dropping the lock.
This commit was sponsored by Anthony DeRobertis on Patreon.
For use with tor hidden services, and perhaps other transports later.
Based on Utility.SimpleProtocol, it's a line-based protocol,
interspersed with transfers of bytestrings of a specified size.
Implementation of the local and remote sides of the protocol is done
using a free monad. This lets monadic code be included here, without
tying it to any particular way to get bytes peer-to-peer.
This adds a dependency on the haskell package "free", although that
was probably pulled in transitively from other dependencies already.
This commit was sponsored by Jeff Goeke-Smith on Patreon.
ghc 8 added backtraces on uncaught errors. This is great, but git-annex was
using error in many places for a error message targeted at the user, in
some known problem case. A backtrace only confuses such a message, so omit it.
Notably, commands like git annex drop that failed due to eg, numcopies,
used to use error, so had a backtrace.
This commit was sponsored by Ethan Aubin.
* S3: Support the special case endpoint needed for the cn-north-1 region.
* Webapp: Don't list the Frankfurt region, as this (and some other new
regions) need V4 authorization which the aws library does not yet use.
This commit was sponsored by Nick Daly on Patreon.
If a transfer fails for some reason, but some data managed to be sent, the
transfer will be retried. (The assistant already did this.)
Possible impacts:
* More ssh prompts if ssh needs to prompt for a password to connect to a
host, or is prompting about some other problem like a ssh key mismatch.
* More data transfer due to retrying, epecially when a remote does not
support resuming a transfer.
In the worst case, a lot of data will be transferred but it fails before
the end, and then all that data gets transferred again plus one byte more;
repeat until it manages to get the whole file.
Multiple external special remote processes for the same remote will be
started as needed when using -J.
This should not beak any existing external special remotes, because running
multiple git-annex commands at the same time could already start multiple
processes for the same external special remotes.
Note that get --from foo --failed will get things that a previous get --from bar
tried and failed to get, etc. I considered making --failed only retry
transfers from the same remote, but it was easier, and seems more useful,
to not have the same remote requirement.
Noisy due to some refactoring into Types/
Ignore exceptions when getting the cost and availability for the remote,
and return sane defaults. These defaults are not cached, so if a special
remote program has a transient problem, it will re-query it later.
Removed the instance LensGpgEncParams RemoteConfig because it encouraged
code that does not take the RemoteGitConfig into account.
RemoteType's setup was changed to take a RemoteGitConfig,
although the only place that is able to provide a non-empty one is
enableremote, when it's changing an existing remote. This led to several
folow-on changes, and got RemoteGitConfig plumbed through.
This is useful for makking a special remote that anyone with a clone of the
repo and your public keys can upload files to, but only you can decrypt the
files stored in it.
The naming is unofrtunately not consistent, but the gnupg-options
were only used for encrypting, and it's too late to change that.
It would be nice to have a third setting that is always passed to gnupg,
but ~/.gnupg/options can be used to specify such global options when really
needed.
Since git-annex unsets these when started, they have to be explicitly
propigated. Also, this makes --git-dir and --work-tree settings be
reflected in the environment.
The need for this came up in
https://github.com/DanielDent/git-annex-remote-rclone/issues/3
Fix hang when dropping content needs to lock the content on a ssh remote,
which occurred when the remote has git-annex version 5.20151019 or newer.
Analysis: `race` runs 2 threads at once, and the hGetLine finishes first.
So, it tries to cancel the waitForProcess, but unfortunately that is making
a foreign call and so cannot be canceled. The remote git-annex-shell
is waiting for a line on stdin before it will exit. Deadlock.
This only occurred sometimes; I reproduced it going from darkstar to
elephant, but not from darkstar to darkstar. Not sure how that fits into
the above analysis -- perhaps a race condition is also involved?
Fixed by not using `race`; now the hGetLine will fail with an exception
if the remote git-annex-shell exits without any output.
sshOptions is now designed for working out ssh options only, and may
insert the extra options it is given to the middle. So it is incorrect
to call it with the remote parameters at the end. Instead, append them
to its return value.
This half regressed in 5be7ba7, and presumably regressed fully when
sshOptions was changed some time later.
That trailing slash is needed for legacy chunked mode, because it puts the
chunks in a subdir under the key. But, outside legacy chunked mode, it's BS
and it's amazing it worked at all with some webdav servers.
* Removed the webapp-secure build flag, rolling it into the webapp build
flag.
* Removed the quvi and tahoe build flags, which only adds aeson to
the core dependencies.
* Removed the feed build flag, which only adds feed to the core
dependencies.
Build flags have cost in both code complexity and also make Setup configure
have to work harder to find a usable set of build flags when some
dependencies are missing.
In copyFromRemote, it used to check isDirect, but that was not needed;
the remote is sending the file, so it doesn't matter if the local,
receiving repository is in direct mode or not. And, since the content is not
present, yet, it's certianly not unlocked. Note that, the remote may indeed
be sending an unlocked file, but sendkey uses sendAnnex, which will detect
if the file is modified before or during transfer, and will exit nonzero,
aborting the upload. So, the receiver doesn't need any checks.
In copyToRemote, it forces recvkey to verify content whenever it's being
sent from a v6 repository. recvkey is almost always going to verify content
anyway, unless annex.verify is not set. So, this doesn't make it any more
expensive, except for in that unusual configuration. The alternative would
be to change the recvkey interface, so that the sender checks afterwards if
what it was sending changed, and the receiver then throws out the bad
transfer. That would be less expensive for the reciever, as it would not
need to do a checksum verification. But, it would mean another network
round trip, and since rsync closes the connection, it would need to open
another ssh connection to do this. Even with connction caching, that would
add latency to uploads. It would also complicate the interface, especially
because an older git-annex-shell would not have the new interface
available. For these reasons, I prefer punting on that at this time, and
instead someone might set annex.verify=false and be unhappy that it still
verifies..
(One other gotcha not dealt with is that a v5 repo could be upgraded to v6
while an upload is in progress, and a file unlocked and modified.)
(Also, I double-checked Remote.GCrypt's calls to rsyncParamsRemote, and
they're fine. When a file is being uploaded to gcrypt, or any other special
repository, it is mediated by sendAnnex, so changes will be detected at
that level and the special remote implementation doesn't need to worry
about them.)
The direct flag is also set when sending unlocked content, to support old
versions of git-annex-shell. At some point, the direct flag will be
removed, and only the unlocked flag will be used.
Had everything available, just didn't combine the progress meter with the
other places progress is sent to update it. (And to a remote repo already
did show progress.)
Most special remotes should already display progress meters with -J,
same as without it. One exception to this is the web, since it relies on
wget/curl progress display without -J. Still todo..
* Fix failure to build with aws-0.13.0.
* When built with aws-0.13.0, the S3 special remote can be used to create
google nearline buckets, by setting storageclass=NEARLINE.
Instead, only display transport error if the configlist output doesn't
include an annex.uuid line, even an empty one.
A recent change made git-annex init try to get all the remote uuids, and so
the transport error would be displayed by it. It was also displayed when
eg, copying files to a remote that had no uuid yet.
sideAction is for things not generally related to the current action being
performed. And, it adds a newline after the side action. This was not the
right thing to use for stuff like "checksum", where doing a checksum is
part of the git annex get process, and indeed we want it to display
"(checksum...) ok"
/dev/null stderr; ssh is still able to display a password prompt
despite this
Show some messages so the user knows it's locking a remote, and
knows if that locking failed.
Also, rename lockContent to lockContentExclusive
inAnnexSafe should perhaps be eliminated, and instead use
`lockContentShared inAnnex`. However, I'm waiting on that, as there are
only 2 call sites for inAnnexSafe and it's fiddly.
In c6632ee5c8, it actually only handled
uploading objects to a shared repository. To avoid verification when
downloading objects from a shared repository, was a lot harder.
On the plus side, if the process of downloading a file from a remote
is able to verify its content on the side, the remote can indicate this
now, and avoid the extra post-download verification.
As of yet, I don't have any remotes (except Git) using this ability.
Some more work would be needed to support it in special remotes.
It would make sense for tahoe to implicitly verify things downloaded from it;
as long as you trust your tahoe server (which typically runs locally),
there's cryptographic integrity. OTOH, despite bup being based on shas,
a bup repo under an attacker's control could have the git ref used for an
object changed, and so a bup repo shouldn't implicitly verify. Indeed,
tahoe seems unique in being trustworthy enough to implicitly verify.
* When annex objects are received into git repositories, their checksums are
verified then too.
* To get the old, faster, behavior of not verifying checksums, set
annex.verify=false, or remote.<name>.annex-verify=false.
* setkey, rekey: These commands also now verify that the provided file
matches the key, unless annex.verify=false.
* reinject: Already verified content; this can now be disabled by
setting annex.verify=false.
recvkey and reinject already did verification, so removed now duplicate
code from them. fsck still does its own verification, which is ok since it
does not use getViaTmp, so verification doesn't happen twice when using fsck
--from.
Since I want git-annex to keep building on debian stable, I need to still
support the old http-client, which required explicit calls to
closeManager, or use of withManager to get Managers to close at appropriate
times. This is not needed in the new version, and so they added a
deprecation warning. IMHO much too early, because look at the mess I had to
go through to avoid that deprecation warning while supporting both
versions..
Added support for storageclass=STANDARD_IA to use Amazon's
new Infrequently Accessed storage.
Also allows using storageclass=NEARLINE to use Google's NearLine storage.
The necessary changes to aws to support this are in
https://github.com/aristidb/aws/pull/176
When gpg.program is configured, it's used to get the command to run for
gpg. Useful on systems that have only a gpg2 command or want to use it
instead of the gpg command.
This only makes sense for public repos, that are not chunked, so
that there's a 1:1 from Key in the git-annex repo to file on the remote.
Rather than making every remote implementation deal with that, just disable
whereisKey when it doesn't make sense.
Note that, if an url is added to the web log for such a remote, it's not
distinguishable from another url that might be added for the web remote.
(Because the web log doesn't distinguish which remote owns a plain url.
Urls with a downloader set are distinguishable, but we're not using them
here.)
This seems ok-ish.. In such a case, both remotes will try to use both
urls, and both remotes should be able to.
The only issue I see is that dropping a file from the web remote will
remove both urls in this case. This is not often done, and could even
be considered a feature, I suppose.
Note that I had one in Annex.Action.startup too, but it resulted in a weird
message printed by ssh, "channel 2: bad ext data". I don't know why, but
it only happened when transferinfo was run, so I wonder
if 983a95f021 introduced a fragility somehow.
