An application can choose to invoke setAdvertisedRefs multiple times,
for example several AdvertiseRefsHook installed in a chain. Each of
these invocations populates the advertisedHaves collection with the
unique set of ObjectIds.
This can lead to a server over-advertising with ".have" lines if the
first hook pushes in a lot of references, and the second hook filters
this to a subset. ReceivePack will advertise the unique objects from
the first hook using ".have" lines, which may lead to a huge
advertisement sent to the client.
This can also contribute to a very slow connectivity check after the
pack is parsed as ReceivePack calls markUninteresting on every commit
in advertisedHaves. This may require expanding a lot of subtrees to
mark all trees as uninteresting as well. On a very big repository
this can lead to a many-second stall.
Clear the advertisedHaves collection any time the refs are updated.
Add a test to verify the correct set of objects was sent.
Change-Id: I97f6998d0597251444a2e846a3ea1f461bae96f9
The tests:
- testCheckBlobNotCorrupt
- testCheckBlobCorrupt
create instances of ObjectChecker that are the same.
The tests:
- testCheckBlobWithBlobObjectCheckerNotCorrupt
- testCheckBlobWithBlobObjectCheckerCorrupt
also create instances of ObjectChecker that are the same.
Factor these instances out to constants instead of creating them
in the tests.
The `checker` member is still created anew in each test, since some
of the tests change its state.
Change-Id: I2d90263829d01d208632185b1ec2f678ae1a3f4c
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Add tests for "true" and "false" matching to "YES" and "NO".
Change-Id: I58223855022871ac4b21bd34ff6a9cd00fce30a1
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
If a ReftableConfig has been supplied by the caller, write out a
reftable as a sibling of the the GC pack, alongside the heads.
To bootstrap from a non-reftable system, the refs are read from the
DfsRefDatabase if no GC reftables are present. Its assumed the
references are fully current, and do not need to be merged with any
other reftables. Any non-GC reftables will be pruned at the end of
the GC cycle, just like any packs that were replaced.
If a GC reftable is present, all existing reftables are compacted, and
references from DfsRefDatabase are only used to seed the packer. Its
assumed these are consistent with each other.
Change-Id: Ie397eb58aaaefb6865c816d9b39de3ac12998019
ServerSocket.accept() is not interruptible: a thread busy in accept()
may not react to Thread.interrupt() and may not return from accept()
via an InterruptedException. Close the socket instead to make the
daemon's listener thread terminate.
* Close the listening socket to get the listening thread to exit
instead of interrupting it.
* Add a stopAndWait() method that stops the listening thread and
then waits until it has indeed finished.
* Set SO_REUSE_ADDRESS on the listening socket.
Bug: 376369
Change-Id: I9d6014103e6dcb0173daea134feb44dc52c5c69a
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
While parsing .gitmodules, the name of the submodule subsection is
purely arbitrary: it frequently is the path of the submodule, but
there's no requirement for it to be. By building a map of paths to
the section name in .gitmodules, we can more accurately return
the submodule URL.
Bug: 508801
Change-Id: I8399ccada1834d4cc5d023344b97dcf8d5869b16
Also-by: Doug Kelly <dougk.ff7@gmail.com>
Signed-off-by: Doug Kelly <dougk.ff7@gmail.com>
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Some tests call out to external cgit. Those tests all failed for me
locally on Mac. Turned out that the reason was that the system git
config used by the git in the bazel run contained paths with ~/ but
somehow $HOME was not set. As a result the external git returned
with exit code 128.
Fix this by passing along $HOME explicitly. Also improve assertions
to make sure we do get the stderr of the external command in the
test log.
I hadn't noticed that until now because apparently the maven build
does pass along $HOME.
Change-Id: I7069676d5cc7b23a71e79a4866fe8acab5a405f4
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Make jsch visible to the test bundle and add the dependency.
Change-Id: I0c49ee9b8f64fe8a8c74d2f08865917eb33069b4
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Do not automatically organize imports using a save action since this
seems to be buggy and removed some annotations org.eclipse.jgit.pgm
needs to use args4j.
Change-Id: I5a91292c3b9241ce2dde3e4ecce14ad460097129
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Revert the following save actions which were introduced in c0ad77d8:
- always use braces around blocks
- remove unused imports
Other than I expected save actions are run globally on edited files -
and not only on edited code lines only.
Hence revert the save action "Convert control statement bodies to
blocks" which would affect a large number of code lines not affected by
the change editing some small part of a class. This would generate a
large number of changes which may lead to many unnecessary conflicts.
Total number of affected lines across jgit would be around 10k lines.
Also revert "Remove unused imports" since it erroneously removes imports
of some annotations needed by pgm classes using args4j.
Change-Id: I879a47f68e664129e6124cf25c1ae1f6a2d7a5aa
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Add the following Eclipse save actions executed when saving modified
lines. This should help to reduce manual work needed to maintain a clean
and consistent code style:
- organize imports
- always use braces around blocks
- add missing annotations
- @Override including implementation of interface methods
- @Deprecated
- remove
- unused imports
- unnecessary $NON-NLS$ tags
- redundant type arguments
Also add default values for new settings that were introduced in recent
Eclipse versions up to Neon since we updated save rules the last time.
Change-Id: Idc90b249df044d0552f04edf01a5f607c4846f50
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Some repositories can have a policy that do not accept certain blobs. To
check if the incoming pack file contains such blobs, ObjectChecker can
be used. However, this ObjectChecker is not called by PackParser if the
blob is stored as a whole. This is because the object can be so large
that it doesn't fit in memory.
This change introduces BlobObjectChecker. This interface takes chunks of
a blob instead of the entire object. ObjectChecker can optionally return
a BlobObjectChecker. This won't change existing ObjectChecker
implementation; existing implementation continues to receive deltified
blob objects only.
Change-Id: Ic33a92c2de42bd7a89786a4da26b7a648b25218d
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
** matching always tries the empty match first. If a mismatch occurs
later, the ** must be extended by exactly one segment and matching must
resume with the matcher following the ** matcher.
Bug: 520920
Change-Id: Id019ad1c773bd645ae92e398021952f8e961f45c
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Path pattern matching for attribute rules is different than matching
for excluded files.
The first difference concerns patterns without slashes. For
gitattributes those must match on the last component only, not on
any earlier segment. This is true also for directory-only patterns.
The second difference concerns directory-only patterns. Those also
must not match on a prefix or segment except the last one. They do
not apply recursively to all files beneath.
And third, matches only on a prefix must match for gitattributes
only if the last matcher was "/**".
Add a new parameter for such path matching to IMatcher.matches() and
pass it through as appropriate (false for gitignore, true for
gitattributes). As far as gitignore is concerned, there is no change.
New tests have been added, and some existing attribute matching tests
have been fixed since they operated on wrong assumptions.
Bug: 508568
Change-Id: Ie825dc2cac8a85a72a7eeb0abb888f3193d21dd2
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
These tests verify that JGit matches the same as C git, for
both attribute matching (.gitattributes) and file exclusion matching
(.gitignore). These tests work by setting up a test repository and
test rules, and then determine excluded files or attributes both with
JGit and with the native C git, and then compare the results.
For .gitignore tests, we run
git ls-files --ignored --exclude-standard -o
and for attribute tests we use
git check-attr --stdin --all
and pass the list of all files in the repository via stdin.
Change-Id: I5b40946e04ff4a97456be7dffe09374323b7c89d
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
The Eclipse compiler raises errors for unthrown exceptions declared to
be thrown by test methods introduced in 88e45399.
Change-Id: I0d91c89e1b20ceff52c38b759abf906cc94e9902
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Except for %p and %r and partially %C, we can do token substitutions
as defined by OpenSSH inside the config file parser. %p and %r can
be replaced only if specified in the config; if not, it would be the
caller's responsibility to replace them with values obtained from the
URI to connect to.
Jsch doesn't know about token substitutions at all. By doing the
replacements as good as we can in the config file parser, we can
make Jsch support most of these tokens.
%i is not handled at all as Java has no concept of a "user ID".
Includes unit tests.
Bug: 496170
Change-Id: If9d324090707de5d50c740b0d4455aefa8db46ee
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Ensure the Jsch instance used knows about ~/.ssh/config. This
enables Jsch to honor more user configurations (see
com.jcraft.jsch.Session.applyConfig()), in particular also the
UserKnownHostsFile configuration, or additional identities given
via multiple IdentityFile entries.
Turn JGit's OpenSshConfig into a full parser that can be a
Jsch-compliant ConfigRepository. This avoids a few bugs
in Jsch's OpenSSHConfig and keeps the JGit-facing interface
unchanged. At the same time we can supply a JGit OpenSshConfig
instance as a ConfigRepository to Jsch. And since they'll both
work from the same object, we can also be sure that the parsing
behavior is identical.
The parser does not handle the "Match" and "Include" keys, and it
doesn't do %-token substitutions (yet).
Note that Jsch doesn't handle multi-valued UserKnownHostFile
entries as known by modern OpenSSH.[1]
[1] http://man.openbsd.org/OpenBSD-current/man5/ssh_config.5
Additional tests for new features are provided in OpenSshConfigTest.
Bug: 490939
Change-Id: Ic683bd412fa8c5632142aebba4a07fad4c64c637
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Add an update_index to every reference in a reftable, storing the
exact transaction that last modified the reference. This is necessary
to fix some merge race conditions.
Consider updates at T1, T3 are present in two reftables. Compacting
these will create a table with range [T1,T3]. If T2 arrives during
or after the compaction its impossible for readers to know how to
merge the [T1,T3] table with the T2 table.
With an explicit update_index per reference, MergedReftable is able to
individually sort each reference, merging individual entries at T3
from [T1,T3] ahead of identically named entries appearing in T2.
Change-Id: Ie4065d4176a5a0207dcab9696ae05d086e042140
resolve(Ref) helps callers recursively chase symbolic references and
is a useful function when wrapping a Reftable inside a RefDatabase, as
RefCursor does not resolve symbolic references during iteration.
Change-Id: I1ba143f403773497972e225dc92c35ecb989e154
A compaction of reftables is just copying the results of a
MergedReftable into a ReftableWriter. Wrap this up into a utility.
Change-Id: I6f5677d923e9628993a2d8b4b007a9b8662c9045
MergedReftable combines multiple reference tables together in a stack,
allowing higher/later tables to shadow earlier/lower tables. This
forms the basis of a transaction system, where each transaction writes
a new reftable containing only the modified references, and readers
perform a merge on the fly to get the latest value.
Change-Id: Ic2cb750141e8c61a8b2726b2eb95195acb6ddc83
Add additional test cases for looking up entries within a namespace
such as refs/heads/ or refs/tags/, where the seek is passed a name
that ends with '/'.
Change-Id: I5f944de7518cd0090374bddba48d4dd3955a8d72
Add more test cases that cover larger collections of
references, verifying every reference is accessible
both by scan and by seek.
Change-Id: Icada59fdcfc92a4634f6df61baaebb1c37b75d98
This set of tests covers primitive storage of an empty
file, and each type of supported reference.
Change-Id: I3bdff35cae8ae27283051932f20608b3ac353559
Currently there is no way to determine the precise changes done
to the working tree by a JGit command. Only the CheckoutCommand
actually provides access to the lists of modified, deleted, and
to-be-deleted files, but those lists may be inaccurate (since they
are determined up-front before the working tree is modified) if
the actual checkout then fails halfway through. Moreover, other
JGit commands that modify the working tree do not offer any way to
figure out which files were changed.
This poses problems for EGit, which may need to refresh parts of the
Eclipse workspace when JGit has done java.io file operations.
Provide the foundations for better file change tracking: the working
tree is modified exclusively in DirCacheCheckout. Make it emit a new
type of RepositoryEvent that lists all files that were modified or
deleted, even if the checkout failed halfway through. We update the
'updated' and 'removed' lists determined up-front in case of file
system problems to reflect the actual state of changes made.
EGit thus can register a listener for these events and then knows
exactly which parts of the Eclipse workspace may need to be refreshed.
Two commands manage checking out individual DirCacheEntries themselves:
checkout specific paths, and applying a stash with untracked files.
Make those two also emit such a new WorkingTreeModifiedEvent.
Furthermore, merges may modify files, and clean, rm, and stash create
may delete files.
CQ: 13969
Bug: 500106
Change-Id: I7a100aee315791fa1201f43bbad61fbae60b35cb
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
When creating a new PackFile instance it is specified whether this pack
has an associated bitmap index file or not. This information is cached
and the public method getBitmapIndex() will always assume a bitmap index
file must exist if the cached data tells so. But it may happen that the
packfiles are repacked during a gc in a different process causing the
packfile, bitmap-index and index file to be deleted. Since JGit still
has an open FileHandle on the packfile this file is not really deleted
and can still be accessed. But index and bitmap index file are deleted.
Fix getBitmapIndex() to invalidate the cached packfile instance if such
a situation occurs.
This problem showed up when a gerrit server was serving repositories
which where garbage collected with native git regularly. Fetch and
clone commands for certain repositories failed permanently after a
native git gc had deleted old bitmap index files.
Change-Id: I8e620bec74dd3f310ba42024f9a657062f868f0e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Per the git config documentation[1], pushInsteadOf is ignored when
a remote has explicit pushUris.
Implement this, and adapt tests.
Up to now JGit mistakenly applied pushInsteadOf also to existing
pushUris. If some repositories had relied on this mis-feature,
pushes may newly suddenly fail (the uncritical case; the config
just needs to be fixed) or even still succeed but push to unexpected
places, namely to the non-rewritten pushUrls (the critical case).
The release notes should point out this change.
[1] https://git-scm.com/docs/git-config
Bug: 393170
Change-Id: I38c83204d2ac74f88f3d22d0550bf5ff7ee86daf
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
According to [1], pushInsteadOf is
1. applied to the uris, not to the pushUris
2. ignored if a remote has an explicit pushUri
JGit applied it only to the pushUris. As a result, pushInsteadOf was
ignored for remotes having only a uri, but no pushUri.
This commit implements (1) if there are no pushUris. I did not dare
implement (2) because:
* there are explicit tests for it that expect that pushInsteadOf gets
applied to existing pushUrls, and
* people may actually use and rely on this JGit behavior.
[1] https://git-scm.com/docs/git-config
Bug: 393170
Change-Id: I6dacbf1768a105190c2a8c5272e7880c1c9c943a
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
This matches the proposal that has been discussed at length on
git-core mailing list and seems to be the accepted convention.
Change-Id: I9f6ab15144826893d1e2a4b48a2d657d6dd445ec
Otherwise fancy combinations of attributes (binary or -text in
combination with crlf or eol) may result in the corruption of binary
data.
Bug: 520910
Change-Id: I3ffc666c13d1b9d2ed987b69a67bfc7f42ccdbfc
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Attribute rules must match against the entry path relative to the
attribute node containing the rule. The global entry path is to be
used only for the init and the global node (and of course the root
node).
Bug: 520677
Change-Id: I80389a2dc272a72312729ccd5358d7c75e1ea20a
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
This avoids executing mergeAlgorithm.merge on binary data, which is
unlikely to be useful.
Arguably, binary data should not make it to
ResolveMerger#contentMerge, but this approach has the following
advantages:
* binary detection is exact, since it doesn't only look at the start
of the blob.
* it is cheap, as we have to iterate over the bytes anyway to find
'\n'.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Change-Id: I424295df1dc60a719859d9d7c599067891b15792
Very short abbreviations that are under 8 hex digits do not
have values in w2. Use w1 as the Java hashCode() instead, so
that the prefix of the abbreviation is always included in the
hashing function used by any java.util.Collection type.
Change-Id: Idaf69f86b62630ba4a022d31b4c293c6d138f557
In a server scenario such as Gerrit Code Review, there may be many
atomic BatchRefUpdates contending for locks on both the packed-refs file
and some subset of loose refs. We already retry lock acquisition to
improve this situation slightly, but we can do better by using an
in-process lock. This way, instead of retrying and potentially exceeding
their timeout, different threads sharing the same Repository instance
can wait on a fair lock without having to touch the disk lock. Since a
server is probably already using RepositoryCache anyway, there is a high
likelihood of reusing the Repository instance.
Change-Id: If5dd1dc58f0ce62f26131fd5965a0e21a80e8bd3
If a repo frequently uses PackedBatchRefUpdates, there is likely to be
contention on the packed-refs file, so it's not appropriate to fail
immediately the first time we fail to acquire a lock. Add some logic to
RefDirectory to support general retrying of lock acquisition.
Currently, there is a hard-coded wait starting at 100ms and backing off
exponentially to 1600ms, for about 3s of total wait. This is no worse
than the hard-coded backoff that JGit does elsewhere, e.g. in
FileUtils#delete. One can imagine a scheme that uses per-repository
configuration of backoff, and the current interface would support this
without changing any callers.
Change-Id: I4764e11270d9336882483eb698f67a78a401c251
Shawn Pearce
David Pursehouse
Thomas Wolf
Matthias Sohn
Masaya Suzuki
Joan Goyeau
Christian Halstrick
Han-Wen Nienhuys
Dave Borowitz
Zhen Chen