Having the git-annex-p2p-<netname> command output the socket filename
left git-annex scrambling to listen to it in order to not miss incoming
connections. And if the command uses something like socat UNIX-CONNECT,
that expects the socket to be accepting connections and errors out when
it's not, that would be a problem.
Rather than complicating the protocol with git-annex needing to send
back a message when it's listening to the socket, simplified it by
having git-annex provide the socket path to the command.
This does mean that, if a P2P network has its own place it expects to
find a socket file, the git-annex-p2p-<netname> command would need to
somehow arrange for it to use the git-annex socket path. A symlink would
be one way to handle that situation.
Added git-remote-p2p-annex, which allows git pull and push to P2P networks
provided by external commands.
This is a refactor of git-remote-tor-annex, and should just work. Except
possibly for quirks with the address parsing. I've checked that the address
parsing basically works.
One thing I don't understand is why git-remote-tor-annex removes "/*" from
the end of the address. The git history does not provide any hints. So I
didn't make git-remote-p2p-annex do the same. Maybe that is needed in some
situation? But, a P2P address could contain "/", so removing it would be a
problem. I can't see anything in gitremote-helpers(7) about why the url
might get such a thing added to the end of it. My guess is that is not
needed for tor either (but does no harm there since onion addresses never
contain "/").
At this point, the implementation of generic P2P transports needs only
remotedaemon support.
These are another sort of external addon process, and this makes several
things work including shell scripts on windows. And it makes for nicer
error messages when the command is not in the path.
Note that the refactored startExternalAddonProcess used by this
does not use propGitEnv to set git environment variables in the
environment. Unlike startExternalAddonProcessProtocol which does.
This is because it runs in IO and does not have access to that
information. But also, I don't think that P2P.Generic processes need
that.
p2p: Added --enable option, which can be used to enable P2P networks
provided by external commands git-annex-p2p-<netname>
Made git-annex p2p --enable tor behave the same as git-annex enable-tor,
to make tor a bit less of a special case. However, it canot be run as root,
since it cannot take the user id parameter.
This is probably enough to support accessing remotes using p2p-annex:: urls.
Not tested yet of course since there is not yet support for serving the
other side of such a connection, or for setting up such a connection.
P2P.Generic has an implementation of the whole interface to the
git-annex-p2p-<netname> commands.
When using the new generic P2P transport to open an outgoing connection
to a peer, this will hold the pid of the git-annex-p2p-<netname>
command.
closeConnection simply waits for it. Rather than relying on garbage
collection of the closed handles to close it.
In Remote.Helper.Ssh, connProcess is set to Nothing, even though there
is a similar process being used there. That code stores the pid in
OpenConnection instead, and handles waiting for it itself. A bit ugly,
but not worth cleaning up at this point, maybe later.
This is for p2p-annex:: urls that will use the new generic P2P
transport.
In addressCredsFile, threw in an url encoding of any non-alphanumeric
characters that are in the address. This is to avoid any possible path
traversal attacks via a p2p-annex:: url, since the address part of it
could contain any characters. And, went ahead and did the same url
encoding of tor-annex:: urls, even though tor onion addresses are all
alphanumerics, on the off chance that might avoid a similar problem.
(It does not seem likely enough to treat it as a security hole.)
Improved workaround for git 2.50 bug, avoding an occasional test suite
failure, as well as some situations where an unlocked file did not get
populated when adding another file to the repository with the same content.
This uses the alternate code path that was already using when there was
a conflict. Since that code path is not able to record its work,
it will redo the same work next time. If the only way reconcileStaged
is getting run is via the smudge clean filter, that could result in
more and more changes getting processed redundantly each time. Once
some other git-annex command runs and calls reconcileStaged, it
will stop redoing that work. I don't think the extra work will be a
problem.
To match content that is recorded as present in an url.
Note that, this cannot ask remotes to provide an url using whereisKey, like
whereis does. Because preferred content expressions need to match the same
from multiple perspectives, and the remote would not always be available.
That's why the docs say "recorded as present", but still this may be
surprising to some who see an url in whereis output and are surprised they
cannot match on it.
The use of getDownloader is to strip the downloader prefix from urls like
"yt:". Note that, when OtherDownloader is used, this strips the ":" prefix,
and allows matching on those urls too.
This cleans up after the bug that was fixed in commit
6a9e923c74
Object files that were stored in the wrong location are rescued,
and after that any wrong location logs will be fixed by the usual fsck.
Fix symlinks generated to annexed content when in adjusted unlocked branch in a
linked worktree on a filesystem not supporting symlinks.
Before this fix, the symlink generated by eg git-annex sync in an adjusted
unlocked branch contained ".git/worktrees/". There was a special case
for the similar problem with submodules, so just use it to handle this case as
well. To detect this case, rely on fixupUnusualRepos setting mainWorkTreePath.
This code dates back 10 years to commit
e322826e33. But as far as I can tell,
it was never necessary. Notice that the comment added in that
commit doesn't match the code -- it says it adjusts the Repo when
the filesystem doesn't support symlinks, but it actually only adjusts
the Repo when the filesystem *does* support symlinks.
Testing in a submodule after this change, annex symlinks still point to
.git/annex/objects/.
(Note that gitAnnexLink contains a special case for submodules on filesystem
not supporting symlinks. I have verified that special case still works.
Without that special case, the annex links look like eg
"../.git/modules/foo/annex/objects", and with them, they look like
".git/annex/objects"
Fix bug in handling of linked worktrees on filesystems not supporting
symlinks, that caused annexed file content to be stored in the wrong
location inside the git directory, and also caused pointer files to not get
populated.
This parameterizes functions in Annex.Locations with a GitLocationMaker.
The uses of standardGitLocationMaker are in cases where the path returned
by a function should not change when in a linked worktree. For example,
gitAnnexLink uses standardGitLocationMaker because symlink targets should
always be to ".git/annex/objects" paths, even when in a linked worktree.
Hopefully I have gotten all uses of standardGitLocationMaker right.
This also assumes that all path construction to the annex directory
is done via the functions in Annex.Locations, and there is no other,
ad-hoc construction elsewhere. Thankfully, Annex.Locations has been around
since the beginning, and has been used consistently. I think.
---
In fixupUnusualRepos, when symlinks are supported, the .git file is replaced
with a symlink to the linked worktree git directory. And in that directory,
an "annex" symlink points to the main annex directory. In that case,
it's not necessary to set mainWorkTreePath. It would be ok to set it,
but not setting it in that case allows an optimisation of avoiding reading
the "commondir" file.
The change to make fixupUnusualRepos set mainWorkTreePath when the
repository is not initialized yet is done in case the initialization itself
writes to the annex directory. If that were the case, without setting
mainWorkTreePath, the annex symlink would not be set up yet, and so
it might have created the annex directory in the wrong place. Currently
that didn't happen, but now that mainWorkTreePath is available, using it
here avoids any such later problem.
---
This commit does not deal with the mess of a worktree that has
experienced this bug before. In particular, if `git-annex get` were
run in such a worktree, it would have stored the object files in the
linked worktree's git directory, rather than in the main git directory.
Such misplaced object files need to be dealt with; the plan is to make
git-annex fsck notice and fix them.
A worktree that has experienced this bug before will contain unpopulated
pointer files. Those may eventually get fixed up in regular usage of
git-annex, but git-annex fsck will also fix them up.
---
Finally, this has me pondering if all of git-annex's state files should
really be stored in one common place across all linked worktrees. Should
perhaps state files that are specific to the worktree be stored per-worktree?
That has not been the case when using git-annex on filesystems supporting
symlinks, but it *has* been the case on filesystems not supporting
symlinks. Perhaps this leads to some other buggy behavior in some cases.
Or perhaps to extra work being done.
For example, the keys database has an associated files table. Which depends
on the worktree. But reconcileStaged updates that table, so when git-annex
is used first in one worktree and then in another one, reconcileStaged will
update the table to reflect the current worktree. Which is extra work each
time a different worktree is used. But also, what if two git-annex
processes are running at the same time, in separate worktrees? Probably
this needs more thought and investigation.
So there is a risk that this commit exposes such buggy behavior in a
situation where it didn't happen before, due to the filesystem not
supporting symlinks. But, given how much this bug crippled using linked
worktrees in such a situation, I doubt that many people have been doing
that.
