Support VERSION 2 in the external special remote protocol, which is
identical to VERSION 1, but avoids external remote programs neededing to
work around the above bug. External remote program that support
exporttree=yes are recommended to be updated to send VERSION 2.
Sponsored-by: Kevin Mueller on Patreon
Remote.Directory makes a temp file, then calls this, and since the temp
file exists, it prevented probing if CoW works.
Note that deleting the empty file does mean there's a small window for a
race. If another process is also exporting to the remote, that could let it
make the same temp file. However, the temp filename actually has the
processes's pid in it, which avoids that being a problem.
This may have been a reversion caused by commits around
63d508e885, but I haven't gone back and
tested to be sure. The directory special remote had supposedly supported
CoW for this going back to about half a year before that.
Sponsored-by: Graham Spencer on Patreon
The temporary URL key used for the download, before the real key is
generated, was blocked by annex.securehashesonly.
Fixed by passing the Backend that will be used for the final key into
runTransfer. When a Backend is provided, have preCheckSecureHashes
check that, rather than the key being transferred.
Sponsored-by: unqueued on Patreon
I don't know of scenarios where that can happen (besides the bug
fixed by the parent commit), but there probably are some.
Sponsored-by: Boyd Stephen Smith Jr. on Patreon
Avoid failure to update adjusted branch --unlock-present after git-annex
drop when annex.adjustedbranchrefresh=1
At higher values, it did flush the queue, which ran restagePointerFiles.
But at 1, adjustedBranchRefreshFull gets added to the queue, and while
restagePointerFiles is also in the queue, it runs after that.
Sponsored-by: Brock Spratlen on Patreon
Such an url is not valid; parseURI will fail on it. But git-annex doesn't
actually need to parse the url, because all it needs to do to support
syncing with it is know that it's not a local path, and use git pull and
push.
(Note that there is no good reason for the user to use such an url. An
absolute url is valid and I patched git-remote-gcrypt to support them
years ago. Still, users gonna do anything that tools allow, and
git-remote-gcrypt still supports them.)
Sponsored-by: Jack Hill on Patreon
copy: When --from and --to are combined and the content is already present
on the destination remote, update location tracking as necessary.
Sponsored-by: Dartmouth College's DANDI project
A repository can have a newline in its description due to being in a
directory containing a newline, or due to git-annex describe being
passed a string with a newline in it for some reason. Putting that
newline in uuid.log breaks its format.
So, escape the newline when it enters uuid.log, to \n
This is a one-way escaping, it is not converted back to a newline
when reading the log. If it were, commands like git-annex info and
whereis would display a multi-line description, which could be confusing
to read.
And, implementing roundtripping would necessarily cause problems if an
old version of git-annex were used to set a description that contained
whatever special character is used to escape the \n. Eg, a \ or if
it used the ! prefix before base64 data that is used in some other logs,
the ! character. Then the description set by the old git-annex would not
roundtrip.
There just doesn't seem to be any benefit of roundtripping newlines through,
so why bother? And, git often displays \n for newline when a filename
contains a newline, so git-annex doing it in this case seems sorta ok
by analogy to git.
(Some other git-annex logs can also have newlines put into them if the
user really wants to break git-annex. For example:
git-annex config annex.largefiles "foo
bar"
The full list is probably config.log, remote.log, group.log,
preferred-content.log, required-content.log,
group-preferred-content.log, schedule.log. Probably there is no
good reason to use a newline in any of these, and the breakage is
probably limited to the bad data the user put in not coming back out.
And users can write any garbage to log files themselves manually in any
case. So, I am not going to address all of those at this time. If a
problem such as this one with the newline in the repository path comes
up, it can be dealt with on a case by case basis.)
Sponsored-by: Dartmouth College's Datalad project
git hash-object --stdin-paths is a newline protocol so it cannot
support them. It would help to not use absPath, when the problem
is that the repository itself is in a path with a newline. But,
there's a reason it used absPath, which is that
git hash-object --stdin-paths actually chdirs to the top of the
repository on startup! That is not documented, and I think is a bug
in git.
I considered making the path relative to the top of the repo, but
then what if this is a git bug and gets fixed? git-annex would break
horribly.
So instead, keep the absPath, but when the path contains a newline,
fall back to running git hash-object once per file, which avoids
the problem with newlines and --stdin-paths. It will be slower,
but this is an edge case. (Similar slow code paths are already used
elsewhere when dealing with filenames with newlines and other parts
of git that use line-based protocols.)
Sponsored-by: Dartmouth College's Datalad project
This reverts commit 648e59cac2.
Failed to build on windows, because
In the dependencies for haskeline-0.8.2:
Win32-2.11.1.0 from Stack configuration does not match >=2.1 && <2.10 || >=2.12 (latest
matching version is 2.13.4.0)
jkniiv did find a solution that builds:
-- Win32-2.11.1.0
+- Win32-2.9.0.0
+- Cabal-3.6.3.0
+- directory-1.3.7.1
+- process-1.6.17.0
+- time-1.11.1.2
But that is a quite old version of Win32 and risks bugs from it, and bumping
Cabal and directory to newer than lts-19.33 has seems also likely to be risky.
So, I've given up. aws-0.24 won't be able to be in the stack build until
there's a stackage lts (or nightly) that has filepath (>=1.4.100.0),
which will not happen until sometime after the next ghc release.
info: Fix reversion in last release involving handling of unsupported input
by continuing to handle any other inputs, before exiting nonzero at the
end.
Sponsored-by: Dartmouth College's Datalad project
path to a bare repo when git config is not allowed to list the configs
due to the CVE-2022-24765 fix.
That resulted in a confusing error message, and prevented the nice
message that explains how to mark the repo as safe to use.
Made isBare a tristate so that the case where core.bare is not returned can
be handled.
The handling in updateLocation is to check if the directory
contains config and objects and if so assume it's bare.
Note that if that heuristic is somehow wrong, it would construct a repo
that thinks it's bare but is not. That could cause follow-on problems,
but since git-annex then checks checkRepoConfigInaccessible, and skips
using the repo anyway, a wrong guess should not be a problem.
Sponsored-by: Luke Shumaker on Patreon
Used to fail with a bad error message, indicating there was no
repository with the specified name, or something like that. Now, suggest
they use the uuid to disambiguate.
* info, enableremotemote, renameremote: Avoid a confusing message when more
than one repository matches the user provided name.
* info: Exit nonzero when the input is not supported.
Sponsored-by: Kevin Mueller on Patreon
sync: Fix a bug that caused files to be removed from an importtree=yes
exporttree=yes special remote when the remote's annex-tracking-branch was
not the currently checked out branch.
Sponsored-by: Max Thoursie on Patreon
S3: Support a region= configuration useful for some non-Amazon S3
implementations. This feature needs git-annex to be built with aws-0.24.
datacenter= sets both the AWS hostname and region in one setting, which is
easy when using AWS, but not useful for other hosts. So kept datacenter
as-is, but added this additional config.
Sponsored-By: Brett Eisenberg on Patreon
Remove closed bugs and todos that were last edited or commented before 2022.
Except for ones tagged projects/* since projects like datalad want to keep
around records of old deleted bugs longer.
Command line used:
for f in $(grep -l '|done\]\]' -- ./*.mdwn); do if ! grep -q "projects/" "$f"; then d="$(echo "$f" | sed 's/.mdwn$//')"; if [ -z "$(git log --since=01-01-2022 --pretty=oneline -- "$f")" -a -z "$(git log --since=01-01-2022 --pretty=oneline -- "$d")" ]; then git rm -- "./$f" ; git rm -rf "./$d"; fi; fi; done
for f in $(grep -l '\[\[done\]\]' -- ./*.mdwn); do if ! grep -q "projects/" "$f"; then d="$(echo "$f" | sed 's/.mdwn$//')"; if [ -z "$(git log --since=01-01-2022 --pretty=oneline -- "$f")" -a -z "$(git log --since=01-01-2022 --pretty=oneline -- "$d")" ]; then git rm -- "./$f" ; git rm -rf "./$d"; fi; fi; done
This is to cut down on the number of files in bugs/, which makes it slow
to file new bug reports or update active bug reports. These old bugs
were about 1/3rd of the files in there. These projects want lists of
their old bugs to still be accessible, and have the lists on their
project pages, which will still list the old bugs.
Commands used:
for f in $(git grep -l '\[\[!tag projects/dandi\]\]'); do if grep -q 'done\]\]' "$f"; then git mv "$f" ../projects/dandi/bugs-done; g=$(echo "$f" | sed 's/.mdwn//'); if [ -d "$g" ]; then git mv "$g" ../projects/dandi/bugs-done; fi; fi; done
for f in $(git grep -l '\[\[!tag projects/repronim\]\]'); do if grep -q 'done\]\]' "$f"; then git mv "$f" ../projects/repronim/bugs-done; g=$(echo "$f" | sed 's/.mdwn//'); if [ -d "$g" ]; then git mv "$g" ../projects/repronim/bugs-done; fi; fi; done
for f in $(git grep -l '\[\[!tag projects/datalad\]\]'); do if grep -q 'done\]\]' "$f"; then git mv "$f" ../projects/datalad/bugs-done; g=$(echo "$f" | sed 's/.mdwn//'); if [ -d "$g" ]; then git mv "$g" ../projects/datalad/bugs-done; fi; fi; done
That assumes that bugs are not tagged by multiple projects at the same
time. Of the ones I moved, I've checked and none are.
Could do the same with todo/ but there are only 370 files in there, and
less than 84 of them could be moved this way, which does not seem likely
to produce a sizeable speedup.
Sponsored-by: Dartmouth College's Datalad project
In Makefile, listed additional deps of Build/Standalone. Without that,
it does not get updated for the change to Utility/LinuxMkLibs.hs when
compiling incrementally.
Sponsored-by: Dartmouth College's DANDI project
Fix a hang that occasionally occurred during commands such as move.
(A bug introduced in 10.20220927, in
commit 6a3bd283b8)
The restage.log was kept locked while running a complex index refresh
action. In an unusual situation, that action could need to write to the
restage log, which caused a deadlock.
The solution is a two-stage process. First the restage.log is moved to a
work file, which is done with the lock held. Then the content of the work
file is read and processed, which happens without the lock being held.
This is all done in a crash-safe manner.
Note that streamRestageLog may not be fully safe to run concurrently
with itself. That's ok, because restagePointerFiles uses it with the
index lock held, so only one can be run at a time.
streamRestageLog does delete the restage.old file at the end without
locking. If a calcRestageLog is run concurrently, it will either see the
file content before it was deleted, or will see it's missing. Either is
ok, because at most this will cause calcRestageLog to report more
work remains to be done than there is.
Sponsored-by: Dartmouth College's Datalad project
The splitting of the tests into parts for parallelism made --pattern
do extra work, because init tests have to be run for each part, but
many of the parts are empty.
For example, git-annex test --pattern '/move (ssh remote)/'
took 12 seconds to run before. This improves the runtime to 4 seconds.
Sponsored-by: Dartmouth College's Datalad project
init: Avoid scanning for annexed files, which can be lengthy in a
large repository. Instead that scan is done on demand. This lets git-annex
init be run and some query commands be used in a repository without
waiting.
Note that autoinit already behaved this way, so while this will mean some
commands like git-annex get/unlock/add will do the scan the first time run,
that is not really a significant behavior change.
And, it's really better to have a consistent behavior. The reason for
the inconsistency was a strange bug discussed in
b3c4579c79. Avoiding reconcileStaged in
init will keep avoiding whatever that was.
Sponsored-by: Dartmouth College's DANDI project
Increasing the size of the queue 10x makes git-annex init 7% faster in a
repository with 86000 annexed files.
The memory use goes up, from 70876 kb to 85376 kb.
Avoids database querying overhead when the database is newly created.
In the large repository where git-annex init took 24 seconds, this sped it
up to 20.47 seconds, a speedup of around 15%.
Sponsored-by: Dartmouth College's DANDI project
export: Fix a bug that left a file on a special remote when two files with
the same content were both deleted in the exported tree.
Case of the wrong data structure leading to the wrong result.
The DiffMap now contains all the old filenames, and all the new filenames.
Note that, when 2 files with the same content are both renamed,
it only renames the first, but deletes and re-exports the second.
Improving that is possible, but it would need to use a different temporary
filename. Anyway, that is an unusual case, and there are known to be other
unusual cases where export does not rename with maximum efficiency, IIRC.
(Or maybe this is the case that I remember?)
Sponsored-by: Dartmouth College's OpenNeuro project
