There was a bug when carrying over flags from a merge commit to its
non-first parents. The first parent of a merge commit was handled
differently and correct but the non-first parents are handled by a
recursive algorithm. Flags should be copied from the root merge commit
to parent-2, to grandparent-2, ... up to the limit of STACK_DEPTH==500
parents-levels. But the recursive algorithm was always copying only to
the direct parents of the merge commit and not the grand*-parents.

This seems to be no problem when commits are handled in a strict date
order because then copying only one level is no problem if children are
handled before parents. But when commits are not seperated anymore by
distinctive correct dates (e.g. because all commits have the same date)
then it may happen that a merge-parent is handled before the merge
commit and when dealing later with the merge commit one has to copy
flags down to more than one level

Bug: 501211
Change-Id: I2d79a7cf1e3bce21a490905ccd9d5e502d7b8421

BranchConfig treated this config property as a boolean, but git also
allows the values "preserve" and "interactive". Config property
pull.rebase also allows the same values.

Replace private enum PullCommand.PullRebaseMode by new public enum
BranchConfig.BranchRebaseMode and adapt all uses. Add a new setter to
PullCommand.

Note: PullCommand will treat "interactive" like "true", i.e., as a
non-interactive rebase. Not sure how "interactive" should be handled.
At least it won't balk on it.

Bug: 499482
Change-Id: I7309360f5662b2c2efa1bd8ea6f112c63cf064af
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>

Bug: 500967
Change-Id: I7fabc2654af97011c62f46d5c30ee992341e45e2
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>

Change-Id: I6691b454404dd4db3c690ecfc7515de765bc2ef7
Signed-off-by: Martin Goellnitz <m.goellnitz@outlook.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

- enhance FS.readPipe to throw an exception if the external command
fails to enable the caller to handle the command failure
- reduce log level to warning if system git config does not exist
- improve log message 

Bug: 476639
Change-Id: I94ae3caec22150dde81f1ea8e1e665df55290d42
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

Without the bridge JGit tests don't show log output in Eclipse console.

Change-Id: I7acce1f1787960b5ca98377cb5c7f599a8a220b5
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

Since 4.0 JGit does no longer support Java versions older than Java 7 so
there is no need anymore to mention Java 7 in the class name.

Change-Id: Ic46c9d89a7e919ae4a69487fa06de0478d2b21f0
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

Guide implementors which exception to throw in case of errors.

Change-Id: I74fb76cdf6b7cdef513f3fe8c144572e869cc533
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

Change-Id: I438456b76aefd361384729686271288186d3be3b
Signed-off-by: Ned Twigg <ned.twigg@diffplug.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

Change-Id: I05d5392789b5b64e6ee44f678556cf25dc30d7ba
Signed-off-by: Ned Twigg <ned.twigg@diffplug.com>

Instead of returning null, LfsProtocolServlet#getLargeFileRepository
should throw LfsUnavailable.

If null is returned, throw a generic LfsException.

Handle LfsException as an internal server error and return HTTP 500.

Change-Id: I33e2a19fcc0fde8aaf0f703860c8fa8ce2de2db5
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Change-Id: I4967b5f7a9a7e9488d20f60aaa949efe97442429
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

This allows the same try/catch to handle parsing the command list,
push certificate and push options. Any errors will be caught and
handled by the same catch block, as the client is in the same state.

Change-Id: I13a66f9100e2dc8ca8f72cd701a5bd44d093ec84

Checking if the instance allows push options before returning the
collection or null is a bit overkill. Just return the collection
or return null.

Change-Id: Icdc3755194373966e5819284aeb9bfe8dd34de82

Some embeddings of JGit require creating a ReceivePack instance in
another process from the one that handled the network socket with the
client.  Similar to the PushCertificate add a setter to allow the
option list to be supplied.

Change-Id: I303a30e54942ad067c79251eff8b53329c406628

Noticed by error-prone (http://errorprone.info/bugpattern/ClassCanBeStatic).

Change-Id: Ie271e9e789a5bc764a287f7bd2a8c8af297d7d41

Refactor all of the push option support code to allocate the list
immediately before parsing the options section off the stream.

Move option support down to ReceivePack instead of BaseReceivePack.
Push options are specific to the ReceivePack protocol and are not
likely to appear in the 4 year old subscription proposal.  These
changes are OK before JGit 4.5 ships as no consumer should be relying
on these new APIs.

Change-Id: Ib07d18c877628aba07da07cd91875f918d509c49

Initialize pushOptions when we decide to use them, instead of when we
advertise them.

In the case of HTTP the advertisement is in a different network
request, hence in a different instance of the BaseReceivePack.

Change-Id: I094c60942e04de82cb6d8433c9cd43a46ffae332
Signed-off-by: Stefan Beller <sbeller@google.com>

Do not open an OBJ_TREE if the caller is expecting an OBJ_BLOB or
OBJ_COMMIT; instead throw IncorrectObjectTypeException.  This better
matches behavior of WindowCursor, the ObjectReader implementation of
the local file based object store.

Change-Id: I3fb0e77f54895b123679a405e1b6ba5b95752ff0

DfsRefDatabase#compareAndPut had a vague semantics for reference
matching. Because of this, an operation to make a symbolic
reference had been broken for some DFS implementations even if they
followed the contract of compareAndPut. The clarified semantics
requires the implementations to satisfy the followings:

* Matching references should be both symbolic references or both
  object ID references.
* If both are symbolic references, both should have the same target
  name.
* If both are object ID references, both should have the same object
  ID.

This semantics is defined based on
https://git.eclipse.org/r/#/c/77416/. Before this commit,
DfsRefDatabase couldn't see the target of symbolic references.

InMemoryRepository is changed to comply with the new semantics. This
semantics change can affect the existing DFS implementations that only
checks object IDs. This commit adds two tests that the previous
InMemoryRepository couldn't pass.

Change-Id: I6c6b5d3cc8241a81f4a37782381c88e8a59fdf15
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>

The RepositoryTestCase hierarchy no longer comes from TestCase, so all
test methods must have @Test.

Fix one test that was broken but never run; fortunately this was just
a typo in the test code.

Change-Id: I3ac8ccdab5e2d5539c63d7b0a88d8bdb0c5ff66e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

A JUnit TestRule which enables to run the same JUnit test repeatedly.
This may help to identify the root cause why a flaky tests which succeed
most often does fail sometimes.

Add the RepeatRule to the test class containing the test to be repeated:

public class MyTest {
	@Rule
	public RepeatRule repeatRule = new RepeatRule();
	...
}
 
and annotate the test to be repeated with the @Repeat(n=<repetitions>)
annotation:
 
	@Test
	@Repeat(n = 100)
	public void test() {
		...
	}
 
then this test will be repeated 100 times. If any test execution fails
test repetition will be stopped.

Change-Id: I7c49ccebe1cb00bcde6b002b522d95c13fd3a35e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

When doing a detaching operation, JGit fakes a SymbolicRef as an
ObjectIdRef. This is because RefUpdate#updateImpl dereferences the
SymbolicRef when updating it. For example, assume that HEAD is
pointing to refs/heads/master. If I try to make a detached HEAD
pointing to a commit c0ffee, RefUpdate dereferences HEAD as
refs/heads/master first and changes refs/heads/master to c0ffee. The
detach argument of RefDatabase#newUpdate avoids this dereference by
faking HEAD as ObjectIdRef.

This faking is problematic for the linking operation of
DfsRefDatabase. It does a compare-and-swap operation on every
reference change because of its distributed systems nature. If a
SymbolicRef is faked as an ObjectRef, it thinks that there is a
racing change in the reference and rejects the update. Because of
this, DFS based repositories cannot change the link target of symbolic
refs. This has not been a problem for file-based repositories because
they have a file-lock based semantics instead of the CAS based one.
The reference implementation, InMemoryRepository, is not affected
because it only compares ObjectIds.

When [1] introduced this faking code, there was no way for RefUpdate
to distinguish the detaching operation. When [2] fixed the detaching
operation, it introduced a detachingSymbolicRef flag. This commit uses
this flag to control whether it needs to dereference the symbolic refs
by calling Ref#getLeaf. The same flag is used in the reflog update
operation.

This commit does not affect any operation that succeeds currently. In
some DFS repository implementations, this fixes a ref linking
operation, which is currently failing.

[1]: 01b5392cdb
[2]: 3a86868c08

Change-Id: I118f85f0414dbfad02250944e28d74dddd59469b
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>

Since [1], the git-lfs specification allows the server to return
HTTP 507 if there is insufficient storage for the uploaded object(s).

Add a new exception class, which implementations may throw from the
getRepository() method, causing HTTP 507 to be returned to the client.

[1] https://github.com/github/git-lfs/pull/1473

Change-Id: If5bc0a35fcf870d4216af6ca2f7c8924689ef9c5
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Imitate the packet tracing feature from C Git v1.7.5-rc0~58^2~1 (add
packet tracing debug code, 2011-02-24).  Unlike C Git, use the log4j
log level setting instead of the GIT_TRACE_PACKET environment variable
to enable tracing.

Tested as follows:

 1. Enable tracing by adding the lines

	log4j.logger.org.eclipse.jgit.transport=DEBUG, stderr
	log4j.additivity.org.eclipse.jgit.transport=false

    to org.eclipse.jgit.pgm/resources/log4j.properties.

 2. mvn package

 3. org.eclipse.jgit.pgm/target/jgit \
	ls-remote git://git.kernel.org/pub/scm/git/git 2>&1 |less

Then the output provides a trace of packets sent and received over
the wire:

  2016-08-24 16:36:42 DEBUG PacketLineOut:145 - git> git-upload-pack /pub/scm/git/git^@host=git.kernel.org^@
  2016-08-24 16:36:42 DEBUG PacketLineIn:165 - git< 2632c897f7 HEAD^@multi_ack thin-pack side-band side-band-64k ofs-delta shallow no-progress include-tag multi_ack_detailed symref=HEAD:refs/heads/master agent=git/2.8.4
  2016-08-24 16:36:42 DEBUG PacketLineIn:165 - git< e0c1ceafc5 refs/heads/maint

Change-Id: I5028c064f3ac090510386057cb4e6d30d4eae232
Signed-off-by: Dan Wang <dwwang@google.com>

Unless the user passed --push-option, the client does not intend to
pass push options to the server.

Without this change, all pushes to servers without push option support
fail.

Not enabling the feature (instead of enabling it and sending an empty
list of options) in this case is more intuitive and matches the
behavior of C git push's --push-option parameter better.

Bug: 500149
Change-Id: Ia4f13840cc54d8ba54e99b1432108f1c43022c53
Signed-off-by: Stefan Beller <sbeller@google.com>

HttpClientConnection uses a TemporaryBufferEntity which uses
TemporaryBuffer.LocalFile to buffer an HttpEntity. It was leaking
temporary files if the buffered entities were larger than 1MB since it
failed to destroy the TemporaryBuffer.LocalFile.

Bug: 500079
Change-Id: Ib963e04efc252bdd0420a5c69b1a19181e9e6169
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

This fixes the tests failed in JDK8.

FS uses java.nio API to get file attributes. The timestamps obtained
from that API are more precise than the ones from
java.io.File#lastModified() since Java8.

This difference accidentally makes JGit detect newly added files as
smudged. Use the precised timestamp to avoid this false positive.

Bug: 500058
Change-Id: I9e587583c85cb6efa7562ad6c5f26577869a2e7c
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>

The metadata comparison of submodules is not reliable because of the
last modified timestamp and directory length.

Bug: 498759
Change-Id: If5db69ef3868e475ac477d3e8a7750b268799b0c

This avoids symlink test errors on Linux

Change-Id: Id8193524c40394a90b8315ab0b8256670d618cb5
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>

The exception can be thrown in a various reason, and sometimes 403
Forbidden is not appropriate. Make the HTTP status code customizable.

Change-Id: If2ef6f454f7479158a4e28a12909837db483521c
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>

The git-lfs specification [1] describes the following optional status codes
that may be returned:

429 - The user has hit a rate limit with the server. Though the API does
      not specify any rate limits, implementors are encouraged to set some
      for availability reasons.

509 - Returned if the bandwidth limit for the user or repository has been
      exceeded. The API does not specify any bandwidth limit, but implementors
      may track usage.

Add two new exception classes to represent these cases. Implementations may
throw these from #getLargeFileRepository(), causing the corresponding HTTP
status codes to be returned to the client.

[1] https://github.com/github/git-lfs/blob/master/docs/api/v1/http-v1-batch.md

Change-Id: I7b93f3cf90f7344c90b1587e07927fdeb167097e
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

If the message is not sent, the client shows:

  Unable to parse HTTP response for POST http://admin@localhost:8080/test-project/info/lfs/objects/batch

Change-Id: I8b72d1aded2bcd41b7389676e2373034625a1379
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Instead of hard-coding the message strings, define them in a properties
file. This will allow them to be translated.

Change-Id: I77556881579e66b2c13d187759c7efdddfee87ae
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Instead of using hard-coded HTTP status codes, use the enums
which makes it a bit easier to see what's expected.

Change-Id: I2da5d25632f374b8625d64da4df70d1c9c406bb1
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

AddCommandTest is flaky because IOException is thrown sometimes.

Caused by: java.io.IOException: Stream closed
        at java.lang.ProcessBuilder$NullOutputStream.write(ProcessBuilder.java:433)
        at java.io.OutputStream.write(OutputStream.java:116)
        at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:82)
        at java.io.BufferedOutputStream.flush(BufferedOutputStream.java:140)
        at java.io.FilterOutputStream.close(FilterOutputStream.java:158)
        at org.eclipse.jgit.util.FS.runProcess(FS.java:993)
        at org.eclipse.jgit.util.FS.execute(FS.java:1102)
        at org.eclipse.jgit.treewalk.WorkingTreeIterator.filterClean(WorkingTreeIterator.java:470)
        ... 22 more

OpenJDK replaces the underlying OutputStream with NullOutputStream when
the process exits. This throws IOException for all write operation. When
it exits before JGit writes the input to the pipe buffer, the input
stays in BufferedOutputStream. The close method tries to write it again,
and IOException is thrown.

Since we ignore IOException in StreamGobbler, we also ignore it when
we close the stream.

Fixes Bug 499633.

Change-Id: I30c7ac78e05b00bd0152f697848f4d17d53efd17
Signed-off-by: Masaya Suzuki <draftcode@gmail.com>

If a reference was updated more recently than a pack was written
(typical) the PackList was perpetually dirty until the next GC
was completed for the repository.

Detect this condition by observing no changes to the PackList
membership and resetting the dirty bit.

Change-Id: Ie2133aca1f8083307c73b6a26358175864f100ef

This will be used by EGit for implementing commit amend in the staging
view (see Idcd1efeeee8b3065bae36e285bfc0af24ab1e88f).

Change-Id: Ice9ebbb1c0c3314c679f4db40cdd3664f61c27c3
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>

If the Content-Type is not set on error responses, the git-lfs client
does not read the body which contains the error message, and instead
just displays a generic error message.

Also set the charset on the Content-Type header.

Change-Id: I88e6f07f20b622a670e7c5063145dffb8b630aee
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Pretty printing is only used for outputting json content, which is
interpreted by the client and does not need to be pretty printed.

Change-Id: I48e0280241b6b0f5706300ae0f4c9bc461a89110
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Change-Id: I5c6790719991931d615d821c900bfd90a20e540b
Signed-off-by: Dan Wang <dwwang@google.com>

Change-Id: I167f9e4746b8aa814b361e884c3c9a25e95b9d54
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Trying to open a new writer on the response causes an illegal state
exception and the response is not sent.

Change-Id: Ic718d23cfb3e74f5691cc2aea7283003af7df207
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>

Reported-by: David Pursehouse <david.pursehouse@gmail.com>
Change-Id: I9e9b021d335bda4d58b6bcc30f59b81ac5b37724
Signed-off-by: Jonathan Nieder <jrn@google.com>