]> source.dussan.org Git - jgit.git/log
jgit.git
6 years agoLFS: Better SSH authentication token timeout handling 88/124288/2
Markus Duft [Thu, 12 Apr 2018 08:11:18 +0000 (10:11 +0200)]
LFS: Better SSH authentication token timeout handling

 * Larger eager timeout to compensate for high-latency lines
 * Respect eager timeout in case the server uses "expiresIn"

Change-Id: Id87da1eea874e70b69eaccf35c84af4c3bb50770
Signed-off-by: Markus Duft <markus.duft@ssi-schaefer.com>
6 years agoValidate branch names on branch creation 72/124272/7
Matthias Sohn [Fri, 8 Jun 2018 15:45:00 +0000 (17:45 +0200)]
Validate branch names on branch creation

Since v2.16.0-rc0~89^2~1 (branch: correctly reject
refs/heads/{-dash,HEAD}, 2017-11-14),
native git does not allow branch names
- refs/heads/HEAD
- starting with '-'

Bug: 535655
Change-Id: Ib1c4ec9ea844073901a4ebe6a29ff6cc8ae58e93
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoUse constant for ".lock" 71/124271/3
Matthias Sohn [Fri, 8 Jun 2018 15:22:54 +0000 (17:22 +0200)]
Use constant for ".lock"

Change-Id: Id65dc94c970ffd3ca3d3d4a5d57123c95d29e8af
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoSimplify locking of FileRepository's index snapshot 35/124235/3
Matthias Sohn [Fri, 8 Jun 2018 07:50:39 +0000 (09:50 +0200)]
Simplify locking of FileRepository's index snapshot

synchronize on simple Object monitor instead of using ReentrantLock

Change-Id: I897020ab35786336b51b0fef76ea6071aff8aefa
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoRefactor FileRepository.detectIndexChange() 34/124234/1
Matthias Sohn [Fri, 8 Jun 2018 07:42:13 +0000 (09:42 +0200)]
Refactor FileRepository.detectIndexChange()

Change-Id: I50c751e2e90f685dc27043c569da2eb210d4611b
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge "Ensure index change event is fired when index snapshot changed" into stable-5.0
Jonathan Nieder [Fri, 8 Jun 2018 05:34:12 +0000 (01:34 -0400)]
Merge "Ensure index change event is fired when index snapshot changed" into stable-5.0

6 years agoUpdate to latest Photon Orbit R20180606145124 91/124191/1
Matthias Sohn [Thu, 7 Jun 2018 12:44:40 +0000 (14:44 +0200)]
Update to latest Photon Orbit R20180606145124

also add missing bundle org.hamcrest

Change-Id: I88ad7769b8fda491fe3641ddaf0964ca55ffc0f6
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoUpdate maven plugins to fix Zip Slip vulnerability 56/124056/1
David Pursehouse [Wed, 6 Jun 2018 04:34:34 +0000 (13:34 +0900)]
Update maven plugins to fix Zip Slip vulnerability

Zip Slip [1] is an arbitrary file write generic vulnerability, that can
be achieved using a specially crafted zip (or bzip2, gzip, tar, xz, war)
archive, that holds path traversal filenames.

According to Maven's announcement [2] several plugins use plexus-archiver to
unpack dependencies to disk and have been identified as potential triggers
for exposing the vulnerability.

Of those, JGit uses the maven-dependency-plugin and the maven-javadoc-plugin.

Update them to the fixed versions reported in [2].

See the corresponding issues for the maven-dependency-plugin [3] and the
maven-javadoc-plugin [4] for details.

[1] https://snyk.io/research/zip-slip-vulnerability
[2] https://maven.apache.org/security-plexus-archiver.html
[3] https://issues.apache.org/jira/browse/MDEP-611
[4] https://issues.apache.org/jira/browse/MJAVADOC-520

Change-Id: Id3ab2d6161db240f2ab8f82298fa3ecd7a930a43
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoEnsure index change event is fired when index snapshot changed 92/123892/6
Matthias Sohn [Sun, 3 Jun 2018 11:32:51 +0000 (13:32 +0200)]
Ensure index change event is fired when index snapshot changed

Ensure that notifyIndexChanged is called every time we call
FileSnapshot.save, except the first.

Change-Id: I5a4e9826e791f518787366ae7c3a0ef3d416d2c1
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoPrepare 5.0.0-SNAPSHOT builds 17/124017/1
Matthias Sohn [Tue, 5 Jun 2018 13:31:07 +0000 (15:31 +0200)]
Prepare 5.0.0-SNAPSHOT builds

Change-Id: I9acca328146aa7100fbc1b457fcb1f8fd87cf52a
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoJGit v5.0.0.201806050710-rc3 90/123990/1 v5.0.0.201806050710-rc3
Matthias Sohn [Tue, 5 Jun 2018 11:10:10 +0000 (13:10 +0200)]
JGit v5.0.0.201806050710-rc3

Change-Id: Ic7a6a87dd3a21351e60fbf6cd330456adf1f6800
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoTeach UploadPack "filter" in protocol v2 fetch 61/123961/2
Jonathan Tan [Mon, 30 Apr 2018 20:21:43 +0000 (13:21 -0700)]
Teach UploadPack "filter" in protocol v2 fetch

If the configuration variable uploadpack.allowfilter is true, advertise
that "filter" is supported, and support it if the client sends such an
argument.

Change-Id: I7de66c0a0ada46ff71c5ba124d4ffa7c47254c3b
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoRefactor test of capabilities output 60/123960/2
Jonathan Tan [Wed, 2 May 2018 23:35:48 +0000 (16:35 -0700)]
Refactor test of capabilities output

A subsequent patch will dynamically generate the capability
advertisement, so the capability advertisements produced are not always
the same. Separate the checking of the advertisements into its own test
method.

Change-Id: I768d14b9d1a244d5d886c42ffd62ef3957b518fb
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoRefactor v2 advertisement into own function 59/123959/2
Jonathan Tan [Wed, 2 May 2018 22:43:50 +0000 (15:43 -0700)]
Refactor v2 advertisement into own function

A subsequent patch needs dynamic generation of this advertisement
depending on a configuration variable in the underlying repository, so
refactor it into a function instead of using a constant list.

Change-Id: Ie00584add1fb56c9e88c7b57f75703981ea5bb85
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoRefactor parsing of "filter" into its own method 58/123958/2
Jonathan Tan [Mon, 30 Apr 2018 19:49:02 +0000 (12:49 -0700)]
Refactor parsing of "filter" into its own method

The implementation of protocol v2 will also need to parse the "filter"
option, so refactor it into its own method.

Change-Id: I751f6e6ca63fab873298594653a3885202297a2e
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoDisallow unknown args to "fetch" in protocol v2 57/123957/2
Jonathan Tan [Thu, 3 May 2018 00:17:04 +0000 (17:17 -0700)]
Disallow unknown args to "fetch" in protocol v2

JGit's implementation of the fetch command of protocol v2, unlike its
implementation of ls-refs, currently tolerates unknown arguments.
Tighten fetch to not allow unrecognized arguments and add tests to
verify this behavior for both ls-refs and fetch.

Change-Id: I321161d568bd638252fab1a47b06b924d472a669
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack shallow fetch in protocol v2 56/123956/2
Jonathan Tan [Thu, 15 Mar 2018 22:56:50 +0000 (15:56 -0700)]
Teach UploadPack shallow fetch in protocol v2

Add support for the "shallow" and "deepen" parameters in the "fetch"
command in the fetch-pack/upload-pack protocol v2. Advertise support for
this in the capability advertisement.

TODO: implement deepen-relative, deepen-since, deepen-not

Change-Id: I7ffd80d6c38872f9d713ac7d6e0412106b3766d7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoRefactor unshallowCommits to local variable 55/123955/2
Jonathan Tan [Tue, 22 May 2018 22:19:04 +0000 (15:19 -0700)]
Refactor unshallowCommits to local variable

This reduces the amount of state held as instance variables in
UploadPack, and makes it easier for a future patch to contain a clearer
version of UploadPack#processShallow.

Change-Id: I6df80b42f9e5118fda1420692e02e417670cced3
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoAdd protocol v2 support in http 54/123954/2
Jonathan Tan [Wed, 28 Feb 2018 22:36:44 +0000 (14:36 -0800)]
Add protocol v2 support in http

Teach UploadPack to support protocol v2 with non-bidirectional pipes,
and add support to the HTTP protocol for v2. This is only activated if
the repository's config has "protocol.version" equal to 2.

Change-Id: I093a14acd2c3850b8b98e14936a716958f35a848
Helped-by: Matthias Sohn <matthias.sohn@sap.com>
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoGive info/refs services more control over response 53/123953/2
Jonathan Tan [Wed, 25 Apr 2018 20:59:52 +0000 (13:59 -0700)]
Give info/refs services more control over response

Currently, SmartServiceInfoRefs always prints "# service=serviceName"
followed by a flush packet in response to an info/refs request, and then
hands it off to the specific service class. Printing of "#
service=serviceName" is mandated for protocol v0, but not v2.

Therefore, the existing code works for protocol v0, but whenever a
service that supports protocol v2 receives an info/refs request, it must
first determine which protocol version is to be used (depending on, for
example, the request and any relevant configuration variables), and then
decide if "# service=serviceName" needs to be printed.

Create a new method that v2-supporting service classes can override,
covering the printing of both "# service=serviceName" and everything
that the #advertise method prints. This will be used in a subsequent
commit in which UploadPackServlet (and the other classes it uses) is
updated to support protocol v2.

Change-Id: Ia026b06e96a6b15937514096babd024ef77df1ea
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoChain RemoteAddCommand methods 62/123762/2
Mincong HUANG [Wed, 30 May 2018 20:24:24 +0000 (22:24 +0200)]
Chain RemoteAddCommand methods

Change-Id: I1cf64c90a5cd2220034810b0d65d59582b420c4e
Signed-off-by: Mincong Huang <mincong.h@gmail.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoUpdate target platforms to Orbit R20180531190352-Photon 33/123833/1
Matthias Sohn [Fri, 1 Jun 2018 12:18:29 +0000 (14:18 +0200)]
Update target platforms to Orbit R20180531190352-Photon

Use the final Photon Orbit release

Change-Id: Ie07eb3dedb5ccb6a2a7b0ad2c90593a07fdd240e
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoPrepare 5.0.0-SNAPSHOT builds 03/123703/1
Matthias Sohn [Wed, 30 May 2018 21:16:23 +0000 (23:16 +0200)]
Prepare 5.0.0-SNAPSHOT builds

Change-Id: I9c4a0465f8e85041d24da97f4b4c7ad5f5eed3e4
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoJGit v5.0.0.201805301535-rc2 91/123691/1 v5.0.0.201805301535-rc2
Matthias Sohn [Wed, 30 May 2018 19:33:49 +0000 (21:33 +0200)]
JGit v5.0.0.201805301535-rc2

Change-Id: Ifad678794525c01838b03e52bada66e1c089c5ae
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoFix description property in feature.properties of pgm source feature 80/123580/1
Matthias Sohn [Tue, 29 May 2018 23:14:14 +0000 (01:14 +0200)]
Fix description property in feature.properties of pgm source feature

Change-Id: I29c7bf1dc6aaf71542e2529883d44dcd995e17f7
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoUpdate SUA 31/123531/4
Matthias Sohn [Tue, 29 May 2018 11:45:06 +0000 (13:45 +0200)]
Update SUA

Following [1] update SUA in license.html [2] and feature.properties [3]

[1] https://www.eclipse.org/projects/handbook/#legaldoc-plugins
[2] https://www.eclipse.org/legal/epl/notice.html
[3] https://www.eclipse.org/legal/epl/feature.properties.txt

Bug: 532265
Change-Id: Ia3929808154510ad09ceeead263a3fe0dba9a34a
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoDon't prune symbolic refs when fetch.prune = true 25/123325/1
Thomas Wolf [Thu, 24 May 2018 14:26:36 +0000 (16:26 +0200)]
Don't prune symbolic refs when fetch.prune = true

The canonical implementation also doesn't. Compare current
code in remote.c, function get_stale_heads_cb.[1] Not handling
symrefs in this case was introduced in canonical git in [2]
in 2008.

[1] https://github.com/git/git/blob/v2.17.0/remote.c#L2259
[2] https://github.com/git/git/commit/740fdd27f0

Bug: 533549
Change-Id: If348d56bb4a96b8aa7141f7e7b5a0d3dd4e7808b
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
6 years agoPrepare 5.0.0-SNAPSHOT builds 14/123114/1
Matthias Sohn [Tue, 22 May 2018 23:54:32 +0000 (01:54 +0200)]
Prepare 5.0.0-SNAPSHOT builds

Change-Id: Ie343ccf37f46168041046500a2e19acc80814cfe
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoJGit v5.0.0.201805221745-rc1 08/123108/1 v5.0.0.201805221745-rc1
Matthias Sohn [Tue, 22 May 2018 21:43:43 +0000 (23:43 +0200)]
JGit v5.0.0.201805221745-rc1

Change-Id: Ie2c35fab87f294b00f9754b07b60a848bf256b10
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge branch 'master' into stable-5.0 04/123104/1
Matthias Sohn [Tue, 22 May 2018 21:00:37 +0000 (23:00 +0200)]
Merge branch 'master' into stable-5.0

* master:
  DescribeCommand: Refactor to not use deprecated Repository#peel
  Repository: Deprecate #peel method
  Repository: Make #exactRef and #findRef final
  Skip ignored directories in FileTreeIterator
  Repository: Deprecate getTags method
  InfoRefsServlet: Refactor to not use deprecated methods
  RefAdvertiser: Add send(Collection<Ref>) and deprecate send(Map<String, Ref>)
  Remove deprecated Repository#notifyIndexChanged
  Implementors should override Repository#notifyIndexChanged(boolean)
  Revive Repository#notifyIndexChanged()
  Remove further unnecessary 'final' keywords
  Execute AdvertiseRefsHook only for protocol v0 and v1
  Add protocol v2 support in "jgit daemon"
  Teach UploadPack "ofs-delta" in "fetch"
  Teach UploadPack "include-tag" in "fetch"
  Avoid using #refs in UploadPack#sendPack
  FileRepository: Don't use deprecated RefDatabase#getRefs(String)
  BatchRefUpdate: Don't use deprecated RefDatabase#getRefs(String)

Change-Id: I16c5da62d09262d3f4070aa0f466dd6c8352b5ea
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoDescribeCommand: Refactor to not use deprecated Repository#peel 36/123036/1
David Pursehouse [Tue, 22 May 2018 03:08:08 +0000 (12:08 +0900)]
DescribeCommand: Refactor to not use deprecated Repository#peel

Change-Id: I76073ad62d1bc4fc21d8a1f5fc7eb92060a73baa
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRepository: Deprecate #peel method 35/123035/2
David Pursehouse [Tue, 22 May 2018 02:27:30 +0000 (11:27 +0900)]
Repository: Deprecate #peel method

Callers should use getRefDatabase().peel(ref) instead since it
doesn't swallow the IOException.

Adapt all trivial callers to user the alternative.

DescribeCommand still uses the deprecated method and is not adapted in
this change since it will require more refactoring to add handling of
the IOException.

Change-Id: I14d4a95a5e0570548753b9fc5c03d024dc3ff832
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRepository: Make #exactRef and #findRef final 34/123034/3
David Pursehouse [Tue, 22 May 2018 00:47:16 +0000 (09:47 +0900)]
Repository: Make #exactRef and #findRef final

This means less cognitive overhead for both implementors and callers,
since this way we can guarantee that they are always synonyms for
RefDatabase#exactRef and RefDatabase#findRef, respectively.

Change-Id: Ic8aeb52fc7ed65672f3f6cd1da39a66908d88baa
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoSkip ignored directories in FileTreeIterator 37/120337/3
Thomas Wolf [Tue, 27 Mar 2018 20:22:09 +0000 (22:22 +0200)]
Skip ignored directories in FileTreeIterator

Make FileTreeIterator not enter ignored directories by default. We
only need to enter ignored directories if we do some operation against
git, and there is at least one tracked file underneath an ignored
directory.

Walking ignored directories should be avoided as much as possible as
it is a potential performance bottleneck. Some projects have a lot of
files or very deep hierarchies in ignored directories; walking those
may be costly (especially so on Windows). See for instance also bug
500106.

Provide a FileTreeIterator.setWalkIgnoredDirectories() operation to
force the iterator to iterate also through otherwise ignored
directories. Useful for tests (IgnoreNodeTest, CGitIgnoreTest), or
to implement things like "git ls-files --ignored".

Add tests in DirCacheCheckoutTest, and amend IndexDiffTest to test a
little bit more.

Bug: 388582
Change-Id: I6ff584a42c55a07120a4369fd308409431bdb94a
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
6 years agoRepository: Deprecate getTags method 17/123017/3
David Pursehouse [Mon, 21 May 2018 23:34:53 +0000 (08:34 +0900)]
Repository: Deprecate getTags method

Callers should use getRefDatabase().getRefsByPrefix(R_TAGS) instead.

Adjust the tests accordingly.

Bug: 534731
Change-Id: Ib28ae365e42720268996ff46e34cae1745ad545c
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoMerge changes Ibf6ee80d,I721db0f4,I7d014890
David Pursehouse [Mon, 21 May 2018 23:15:22 +0000 (19:15 -0400)]
Merge changes Ibf6ee80d,I721db0f4,I7d014890

* changes:
  Remove deprecated Repository#notifyIndexChanged
  Implementors should override Repository#notifyIndexChanged(boolean)
  Revive Repository#notifyIndexChanged()

6 years agoInfoRefsServlet: Refactor to not use deprecated methods 28/122728/4
David Pursehouse [Wed, 16 May 2018 03:07:34 +0000 (12:07 +0900)]
InfoRefsServlet: Refactor to not use deprecated methods

- Replace RefDatabase#getRefs(String) with #getRefsByPrefix(String)
- Replace RefAdvertiser#send(Map<String, Ref>) with #send(Collection<Ref>)

Bug: 534731
Change-Id: I25b617c3b0c54793cf4ab5b62f002e17745a5377
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRefAdvertiser: Add send(Collection<Ref>) and deprecate send(Map<String, Ref>) 27/122727/4
David Pursehouse [Wed, 16 May 2018 03:03:02 +0000 (12:03 +0900)]
RefAdvertiser: Add send(Collection<Ref>) and deprecate send(Map<String, Ref>)

Bug: 534731
Change-Id: If15032a34dc62f420569e2b2b6d8e14e2dfed522
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRemove deprecated Repository#notifyIndexChanged 72/122972/1
Jonathan Nieder [Fri, 18 May 2018 16:23:03 +0000 (09:23 -0700)]
Remove deprecated Repository#notifyIndexChanged

It is no longer used.

Change-Id: Ibf6ee80d5b6cd554295f4e657e087302770621b6

6 years agoImplementors should override Repository#notifyIndexChanged(boolean) 71/122971/1
Jonathan Nieder [Fri, 18 May 2018 16:16:16 +0000 (09:16 -0700)]
Implementors should override Repository#notifyIndexChanged(boolean)

Declare Repository#notifyIndexChanged() final and
Repository#notifyIndexChanged(boolean) abstract to force implementors
to switch to overriding the latter method.  This makes Repository less
error-prone to extend since implementors no longer need to remember to
override one of those two methods.

Change-Id: I721db0f4a4865db3b35212ee0a2045d5b31c96af
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoRevive Repository#notifyIndexChanged() 70/122970/1
Jonathan Nieder [Fri, 18 May 2018 16:15:30 +0000 (09:15 -0700)]
Revive Repository#notifyIndexChanged()

e9e150fdd24d (Store in IndexChangedEvent if it was caused by JGit
itself, 2018-05-13) modified Repository#notifyIndexChanged to take a
boolean argument to indicate whether the index change happened under
the current process's control or externally, for use by EGit.  In
other words, the function signature changed from

public abstract void notifyIndexChanged();

to

public abstract void notifyIndexChanged(boolean internal);

Callers outside JGit itself notifying a Repository about index changes
are expected to be rare, so this is not very disruptive to them.  In
most cases they would be notifying about changes that they made
themselves, so treating their notifyIndexChanged() calls as
notifyIndexChanged(true) should be relatively safe.

Implementors have the opposite problem: adding the new "abstract void
notifyIndexChanged(boolean)" method means they are obligated to
override it.  Add a default implementation that calls their existing
override of notifyIndexChanged() to make their migration easier.

The main downside is that authors of new Repository subclasses that
do not realize they need to override notifyIndexChanged would end up
with a default implementation which calls notifyIndexChanged(true),
in turn calling notifyIndexChanged() again and so on, resulting in
StackOverflowException.  Add an implementors' note to the class
Javadoc to avoid this issue.  A followup commit will force
implementors to adapt to the new API by changing the methods to

@Deprecated
public final void notifyIndexChanged() {
notifyIndexChanged(true);
}

public abstract void notifyIndexChanged(boolean internal);

Change-Id: I7d014890ee19abf283ea824d9baa9044bfdde130
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoRemove further unnecessary 'final' keywords 88/122788/2
Han-Wen Nienhuys [Wed, 16 May 2018 14:34:16 +0000 (16:34 +0200)]
Remove further unnecessary 'final' keywords

Remove it from

 * package private functions.

 * try blocks

 * for loops

this was done with the following python script:

$ cat f.py
import sys
import re
import os

def replaceFinal(m):
  return m.group(1) + "(" +  m.group(2).replace('final ', '') + ")"

methodDecl = re.compile(r"^([\t ]*[a-zA-Z_ ]+)\(([^)]*)\)")

def subst(fn):
  input = open(fn)
  os.rename(fn, fn + "~")

  dest = open(fn, 'w')
  for l in input:
    l = methodDecl.sub(replaceFinal, l)
    dest.write(l)
  dest.close()

for root, dirs, files in os.walk(".", topdown=False):
  for f in files:
    if not f.endswith('.java'):
      continue

    full = os.path.join(root, f)
    print full
    subst(full)

Change-Id: If533a75a417594fc893e7c669d2c1f0f6caeb7ca
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
6 years agoExecute AdvertiseRefsHook only for protocol v0 and v1 26/122726/3
Masaya Suzuki [Mon, 14 May 2018 21:13:11 +0000 (14:13 -0700)]
Execute AdvertiseRefsHook only for protocol v0 and v1

Refs are not advertised as part of the protocol v2 capability
advertisement. Don't call AdvertiseRefsHook.

Noticed because many implementations of AdvertiseRefsHook read all
refs in order to call UploadPack#setAdvertisedRefs, causing the
capability advertisement to be as slow as a v0 ref advertisement with
some RefDatabase implementations.

Such an AdvertiseRefsHook is of dubious utility (a better place to
determine which refs are advertised is in the RefDatabase
implementation itself, as in Gerrit), but at any rate since it's not
bringing about any benefit here, we can skip the hook call.

TODO:
- call an appropriate hook instead (https://bugs.eclipse.org/534847)
- add tests

[jn: fleshed out commit message; added TODO notes]

Change-Id: I6eb60ccfb251a45432954467a9ae9c1079a8c8b5
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoAdd protocol v2 support in "jgit daemon" 54/119454/14
Jonathan Tan [Sat, 24 Feb 2018 00:55:03 +0000 (16:55 -0800)]
Add protocol v2 support in "jgit daemon"

With this patch, a server spawned by "jgit daemon" can be accessed using
protocol v2 from a Git client that supports it (for example, "git" with
the appropriate patches). This is only activated if the repository's
config has "protocol.version" be 2.

This required a change to the package-private #execute methods in
DaemonService to allow passing of extra parameters.

This has been tested with a patched Git.

Change-Id: Icf043efec7ce956d72b075fc6dc7a87d5a2da82a
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack "ofs-delta" in "fetch" 53/119453/14
Jonathan Tan [Tue, 13 Mar 2018 20:29:46 +0000 (13:29 -0700)]
Teach UploadPack "ofs-delta" in "fetch"

Add support for the "ofs-delta" parameter in the "fetch" command in
the fetch-pack/upload-pack protocol v2.

Change-Id: I728cf986082fce4ddeb6a6435897692e15e60cc7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack "include-tag" in "fetch" 52/119452/14
Jonathan Tan [Tue, 13 Mar 2018 18:07:36 +0000 (11:07 -0700)]
Teach UploadPack "include-tag" in "fetch"

Add support for the "include-tag" parameter in the "fetch" command in
the fetch-pack/upload-pack protocol v2.

In order to determine which tags to include, only objects pointed to by
refs starting with "refs/tags/" are checked. This restriction is for
performance reasons and to match the behavior of Git (see add_ref_tag()
in builtin/pack-objects.c).

Change-Id: I7d70aa09bcc8a525218ff1559e286c2a610258ca
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoAvoid using #refs in UploadPack#sendPack 04/121604/7
Jonathan Tan [Mon, 23 Apr 2018 20:03:04 +0000 (13:03 -0700)]
Avoid using #refs in UploadPack#sendPack

When OPTION_INCLUDE_TAG is set, UploadPack#sendPack uses the #refs
instance variable as a source of information of tags. A subsequent patch
will need to supply this information to #sendPack without
modifying #refs, so refactor #sendPack to take in this information
through a parameter instead.

Note that prior to this patch, #refs was used twice in #sendPack: once
to generate the argument to PackWriter#setTagTargets, and once to
determine if any tags need to be included in the packfile. This patch
only updates the latter use, since the former is meant not only for
"true" tag targets but any object that should be hoisted earlier during
packing (see the documentation of PackWriter#setTagTargets).

This patch does not introduce any functionality change.

Change-Id: I70ed65a1041334abeda8d4bac98cce7cae7efcdf
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoPrepare 5.0.0-SNAPSHOT builds 24/122824/1
Matthias Sohn [Wed, 16 May 2018 20:23:51 +0000 (22:23 +0200)]
Prepare 5.0.0-SNAPSHOT builds

Change-Id: I65d7d74c6124507aa9cb2856c74db6a1a6935bab
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoJGit v5.0.0.201805151920-m7 23/122723/1 v5.0.0.201805151920-m7
Matthias Sohn [Tue, 15 May 2018 23:20:14 +0000 (01:20 +0200)]
JGit v5.0.0.201805151920-m7

Change-Id: I42eb046c8ac8537672010aabe21459293cb23838
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoFileRepository: Don't use deprecated RefDatabase#getRefs(String) 28/122628/2
David Pursehouse [Tue, 15 May 2018 04:41:15 +0000 (13:41 +0900)]
FileRepository: Don't use deprecated RefDatabase#getRefs(String)

Change-Id: Iec58c973537ddbe0f4e6b8b62fcda5cecc961661
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoBatchRefUpdate: Don't use deprecated RefDatabase#getRefs(String) 25/122625/3
David Pursehouse [Tue, 15 May 2018 00:56:32 +0000 (09:56 +0900)]
BatchRefUpdate: Don't use deprecated RefDatabase#getRefs(String)

Change-Id: I672c9cfe221ddc4acbde7a8040bd6ba83b16626e
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRemove 'final' in parameter lists 05/122605/3
Han-Wen Nienhuys [Mon, 14 May 2018 16:42:02 +0000 (18:42 +0200)]
Remove 'final' in parameter lists

Change-Id: Id924f79c8b2c720297ebc49bf9c5d4ddd6d52547
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
6 years agoFixup javadoc and formatting in RawText and RawParseUtils 97/122597/6
Han-Wen Nienhuys [Mon, 14 May 2018 15:21:39 +0000 (17:21 +0200)]
Fixup javadoc and formatting in RawText and RawParseUtils

Change-Id: I9d6002941a33ec204d29e4fd920dde965387bb24
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
6 years agoRawParseUtils#lineMap: Simplify by using null sentinel internally 00/122600/2
Jonathan Nieder [Mon, 14 May 2018 15:54:01 +0000 (08:54 -0700)]
RawParseUtils#lineMap: Simplify by using null sentinel internally

Add an internal lineMapOrNull helper that returns null when the file
is binary.

This is simpler than using an exception for control flow and avoids
having to override fillInStackTrace to avoid a performance regression.

Change-Id: Ib8bb8df6a6bbd60c62cfb3b4c484a962a98b7507

6 years agoStore in IndexChangedEvent if it was caused by JGit itself 48/122548/2
Matthias Sohn [Sun, 13 May 2018 19:37:21 +0000 (21:37 +0200)]
Store in IndexChangedEvent if it was caused by JGit itself

This allows to differentiate if index was changed by an external git
command or by JGit itself.

Change-Id: Iae692ba7d9bf01a288b3fb2dc2d07aec9891c712
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoIntroduce new RawText constructor and RawParseUtils.lineMapOrBinary 82/121382/6
Han-Wen Nienhuys [Wed, 18 Apr 2018 21:37:25 +0000 (23:37 +0200)]
Introduce new RawText constructor and RawParseUtils.lineMapOrBinary

This makes binary detection exact in ResolveMerger and DiffFormatter

This has the same intention as
Id4342a199628d9406bfa04af1b023c27a47d4014 but preserves backward
compatibility of the signature of RawParseUtils.lineMap.

Change-Id: Ia24a4e716592bab3363ae24e3a46315a7511154f
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoUse a secure random generator to seed nonce for digest authentication 76/122476/1
Matthias Sohn [Fri, 11 May 2018 12:06:53 +0000 (14:06 +0200)]
Use a secure random generator to seed nonce for digest authentication

https://tools.ietf.org/html/rfc7616 says:

5.12.  Parameter Randomness

The security of this protocol is critically dependent on the
randomness of the randomly chosen parameters, such as client and
server nonces.  These should be generated by a strong random or
properly seeded pseudorandom source (see [RFC4086]).

Change-Id: I4da5316cb1eb3f59ae06c070ce1c3335e9ee87d6
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge branch 'stable-4.11' 94/122394/3
Matthias Sohn [Thu, 10 May 2018 11:41:45 +0000 (13:41 +0200)]
Merge branch 'stable-4.11'

* stable-4.11:
  Retry stale file handles on .git/config file

Change-Id: I4fe6152c3c40dde9cb88913cc9706852de0fd712
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge branch 'stable-4.10' into stable-4.11 11/122411/1
Matthias Sohn [Thu, 10 May 2018 10:55:30 +0000 (12:55 +0200)]
Merge branch 'stable-4.10' into stable-4.11

* stable-4.10:
  Retry stale file handles on .git/config file

Change-Id: Ice5c8ae8c2992243a81da77e166406bc1930fe0e

6 years agoMerge branch 'stable-4.9' into stable-4.10 01/122401/1
Matthias Sohn [Thu, 10 May 2018 10:51:57 +0000 (12:51 +0200)]
Merge branch 'stable-4.9' into stable-4.10

* stable-4.9:
  Retry stale file handles on .git/config file

Change-Id: I6db7256dbd1c71b23e1231809642ca21e996e066
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge branch 'stable-4.8' into stable-4.9 00/122400/1
Matthias Sohn [Thu, 10 May 2018 10:30:05 +0000 (12:30 +0200)]
Merge branch 'stable-4.8' into stable-4.9

* stable-4.8:
  Retry stale file handles on .git/config file

Change-Id: Ib029b5536c038190626e7a7ff43b70f0a5673721

6 years agoMerge branch 'stable-4.7' into stable-4.8 99/122399/1
Matthias Sohn [Thu, 10 May 2018 10:08:46 +0000 (12:08 +0200)]
Merge branch 'stable-4.7' into stable-4.8

* stable-4.7:
  Retry stale file handles on .git/config file

Change-Id: Ib665c094b28eefb8236752bb273de0c2d9bf9578

6 years agoMerge branch 'stable-4.6' into stable-4.7 98/122398/1
Matthias Sohn [Thu, 10 May 2018 09:59:56 +0000 (11:59 +0200)]
Merge branch 'stable-4.6' into stable-4.7

* stable-4.6:
  Retry stale file handles on .git/config file

Change-Id: If5a21d38224528edfc551b3216daca6a2582e3ac

6 years agoMerge branch 'stable-4.5' into stable-4.6 95/122395/1
Matthias Sohn [Thu, 10 May 2018 09:39:52 +0000 (11:39 +0200)]
Merge branch 'stable-4.5' into stable-4.6

* stable-4.5:
  Retry stale file handles on .git/config file

Change-Id: Ib6e6ec0846c3ef261ec1016bfa6d26d2eadc3f26

6 years agoRetry stale file handles on .git/config file 73/120973/4
Nasser Grainawi [Fri, 6 Apr 2018 20:05:37 +0000 (14:05 -0600)]
Retry stale file handles on .git/config file

On a local non-NFS filesystem the .git/config file will be orphaned if
it is replaced by a new process while the current process is reading the
old file. The current process successfully continues to read the
orphaned file until it closes the file handle.

Since NFS servers do not keep track of open files, instead of orphaning
the old .git/config file, such a replacement on an NFS filesystem will
instead cause the old file to be garbage collected (deleted).  A stale
file handle exception will be raised on NFS clients if the file is
garbage collected (deleted) on the server while it is being read.  Since
we no longer have access to the old file in these cases, the previous
code would just fail. However, in these cases, reopening the file and
rereading it will succeed (since it will open the new replacement file).
Since retrying the read is a viable strategy to deal with stale file
handles on the .git/config file, implement such a strategy.

Since it is possible that the .git/config file could be replaced again
while rereading it, loop on stale file handle exceptions, up to 5 extra
times, trying to read the .git/config file again, until we either read
the new file, or find that the file no longer exists. The limit of 5 is
arbitrary, and provides a safe upper bounds to prevent infinite loops
consuming resources in a potential unforeseen persistent error
condition.

Change-Id: I6901157b9dfdbd3013360ebe3eb40af147a8c626
Signed-off-by: Nasser Grainawi <nasser@codeaurora.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMark CrissCrossMergeTest as flaky 63/122063/4
Jonathan Nieder [Wed, 2 May 2018 22:23:31 +0000 (15:23 -0700)]
Mark CrissCrossMergeTest as flaky

It often fails on my machine, both in maven and bazel.

This patch marks the test flaky[1] in bazel so that "bazel test" can
run it a few times before declaring failure.

[1] https://docs.bazel.build/versions/master/be/common-definitions.html#test.flaky

Bug: 534285
Change-Id: Ibe5414fefbffe4e8f86af7047608d51cf5df5c47

6 years agoMerge "Add API filter for "non-API type FileRepository" in tests"
Matthias Sohn [Sat, 5 May 2018 23:44:26 +0000 (19:44 -0400)]
Merge "Add API filter for "non-API type FileRepository" in tests"

6 years agoAdd API filter for "non-API type FileRepository" in tests 35/121035/3
David Pursehouse [Thu, 12 Apr 2018 01:53:29 +0000 (10:53 +0900)]
Add API filter for "non-API type FileRepository" in tests

Change-Id: If805ad4a962e48dd16fbc7eff915fd6539839933
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoReplace http://errorprone.info with https://errorprone.info 13/122213/1
Jonathan Nieder [Fri, 4 May 2018 22:09:22 +0000 (15:09 -0700)]
Replace http://errorprone.info with https://errorprone.info

That site serves from https now.

Reported-by: Nicholas Glorioso <glorioso@google.com>
Change-Id: I2150a18425a1fe3ab5a022882ffe06ccbde17f16
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoBazel: Add a target to build jgit commandline tool 58/122058/2
Jonathan Nieder [Wed, 2 May 2018 21:34:21 +0000 (14:34 -0700)]
Bazel: Add a target to build jgit commandline tool

This works well enough for basic testing, especially the "jgit daemon"
command:

 $ bazel build org.eclipse.jgit.pgm:jgit
 INFO: Analysed target //org.eclipse.jgit.pgm:jgit (39 packages loaded).
 INFO: Found 1 target...
 Target //org.eclipse.jgit.pgm:jgit up-to-date:
   bazel-bin/org.eclipse.jgit.pgm/jgit.jar
   bazel-bin/org.eclipse.jgit.pgm/jgit
 INFO: Elapsed time: 15.862s, Critical Path: 14.74s
 INFO: 42 processes: 32 linux-sandbox, 1 local, 9 worker.
 INFO: Build completed successfully, 60 total actions
 $ bazel-bin/org.eclipse.jgit.pgm/jgit rev-list HEAD^..HEAD 2>/dev/null
 e5ed926e02dd29b5e970d421b63b6921cc1800f5

Still to do:

 - suppress "Failed to load class "org.slf4j.impl.StaticLoggerBinder"
   output
 - add org.eclipse.jgit.pgm.test tests to bazel build
 - supply package information so that "jgit version" can work

Change-Id: Ie89872839fa5aae974d5d39b6776b96332b0e83f

6 years agossh: Kill the external process when we're done instead of waiting forever 02/115802/7
Carsten Pfeiffer [Wed, 2 May 2018 15:40:30 +0000 (17:40 +0200)]
ssh: Kill the external process when we're done instead of waiting forever

Bug: 529463
Change-Id: Iaf6fe20a1c759ac5e91a2393d7bc40a94f859e84
Signed-off-by: Carsten Pfeiffer <carsten.pfeiffer@gebit.de>
6 years agoRename RefDatabase#getAllRefs to getRefs 16/121916/2
Jonathan Nieder [Mon, 30 Apr 2018 00:15:18 +0000 (17:15 -0700)]
Rename RefDatabase#getAllRefs to getRefs

This is easier to type and makes it clearer that it only returns refs
and not the pseudo-refs returned by getAdditionalRefs. It also puts us
in a better position to add a method to the Repository class later
that delegates to this one without colliding with the existing
Repository#getAllRefs method that returns a Map<String, Ref>.

While at it, clarify the javadoc of getRefs and hasRefs to make the
same point.

Suggested-by: David Pursehouse <david.pursehouse@gmail.com>
Change-Id: I23497c66ac7b5e0c987b91efbc9e9cc29924ca66
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoRefDatabase: add hasRefs convenience method 02/121902/2
David Pursehouse [Fri, 27 Apr 2018 12:49:08 +0000 (21:49 +0900)]
RefDatabase: add hasRefs convenience method

Callers can now say:

 db.getRefDatabase().hasRefs()

rather than the more verbose:

 !db.getRefDatabase().getAllRefs().isEmpty()

The default implementation simply uses getAllRefs().isEmpty(), but a
derived class could possibly override the method with a more efficient
implementation.

Change-Id: I5244520708a1a7d9adb351f10e43fc39d98e22a1
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoFix comparison order in AnyObjectId 99/121899/1
David Turner [Fri, 27 Apr 2018 19:04:18 +0000 (15:04 -0400)]
Fix comparison order in AnyObjectId

The previous version suggested testing w2 first because w1 was used
for hashing, but in fact, hashCode returns w2.  The order (w3, w4, w5,
w1, w2) might be better on 64-bit processors too, since it allows
comparing 64 bits at a time, although perhaps on a modern SIMD
processor, the entire 160 bytes would be compared at once anyway.

Change-Id: Ieb69606d3c1456aeff36bffe99a71587ea76e977
Signed-off-by: David Turner <dturner@twosigma.com>
6 years agoRefDatabase: Introduce getAllRefs method 27/121827/2
David Pursehouse [Fri, 27 Apr 2018 01:42:53 +0000 (10:42 +0900)]
RefDatabase: Introduce getAllRefs method

Currently to get all refs, callers must use:

  getRefsByPrefix(ALL)

Introduce getAllRefs, which does this, and migrate all existing
callers of getRefsByPrefix(ALL).

Change-Id: I7b1687c162c8ae836dc7db3ccc7ac847863f691d
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoRefDatabase: Update Javadoc for ALL constant 26/121826/1
David Pursehouse [Fri, 27 Apr 2018 01:12:32 +0000 (10:12 +0900)]
RefDatabase: Update Javadoc for ALL constant

The Javadoc refers to the deprecated getRefs method. Update it to refer
to getRefsByPrefix which is the recommended replacement of getRefs.

Change-Id: I61f2abcf1a3794f40a1746317dbc18aa0beb87a7
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoLargePackedWholeObject#openStream: Suppress resource warning 25/121825/1
David Pursehouse [Thu, 26 Apr 2018 23:27:30 +0000 (08:27 +0900)]
LargePackedWholeObject#openStream: Suppress resource warning

Eclipse warns that DfsReader should be managed by try-with-resource.

As described in 1484d6e (LargePackedWholeObject: Do not reuse released
inflater, 2018-04-26), the DfsReader is owned and closed by the
PackInputStream or explicitly closed in the try block's finally.

Suppress the warning with a brief explanatory comment.

Change-Id: I4187c935742072f3ee7f2d3551a6a98d40fc2702
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoLargePackedWholeObject: Do not reuse released inflater 24/121824/2
Jonathan Nieder [Thu, 26 Apr 2018 22:14:50 +0000 (15:14 -0700)]
LargePackedWholeObject: Do not reuse released inflater

LargePackedWholeObject.openStream produces a stream that allows
reading a large object.  This stream holds a DfsReader that takes care
of caching delta bases etc and in particular holds zlib Inflater for
use while reading the each delta in the packfile.

At DfsReader creation time, the Inflater is acquired from a global
InflaterCache to avoid initialization overhead in case there is an
existing Inflater available for reuse.  When done with the Inflater,
the DfsReader is responsible for returning it to the cache for reuse.
The DfsReader is AutoClosable to remind the caller to close it and
release the Inflater when finished with it.

b0ac5f9c8907a4034612543a92eb465e88a9c6f2 (LargePackedWholeObject:
Refactor to open DfsReader in try-with-resource, 2018-04-11) tried to
clarify the lifetime of the DfsReader but was too aggressive: when
this function returns, PackInputStream owns the DfsReader and is
already going to release it.  Worse, the returned InflaterInputStream
holds a reference to the DfsReader's inflater, making releasing the
DfsReader not only unnecessary but unsafe.

The Inflater gets released into the InflaterCache's pool, to be
acquired by another caller that uses it concurrently with the
InflaterInputStream.  This results in errors, such as

 java.util.zip.ZipException: incorrect header check
        at java.util.zip.InflaterInputStream.read(InflaterInputStream.java:164)
        at java.util.zip.InflaterInputStream.skip(InflaterInputStream.java:208)
        at java.io.BufferedInputStream.skip(BufferedInputStream.java:377)

and

 java.util.zip.DataFormatException: incorrect header check
        at java.util.zip.Inflater.inflateBytes(Native Method)
        at java.util.zip.Inflater.inflate(Inflater.java:259)
        at org.eclipse.jgit.internal.storage.dfs.DfsReader.inflate(DfsReader.java:783)
        at org.eclipse.jgit.internal.storage.dfs.DfsPackFile.decompress(DfsPackFile.java:420)
        at org.eclipse.jgit.internal.storage.dfs.DfsPackFile.load(DfsPackFile.java:767)

and

 Caused by: java.util.zip.ZipException: incorrect header check
        at java.util.zip.InflaterInputStream.read(InflaterInputStream.java:164)
        at java.io.BufferedInputStream.fill(BufferedInputStream.java:246)
        at java.io.BufferedInputStream.read1(BufferedInputStream.java:286)
        at java.io.BufferedInputStream.read(BufferedInputStream.java:345)
        at org.eclipse.jgit.lib.ObjectStream$Filter.read(ObjectStream.java:219)
        at org.eclipse.jgit.util.IO.readFully(IO.java:233)
        at org.eclipse.jgit.transport.PackParser.checkObjectCollision(PackParser.java:1173)

Verified in production.  It should be possible to make a
straightforward unit test for this using the InflaterCache state but
that can wait for a followup commit.

Change-Id: Iaf1d6fd368b64f76c520d215fd270a6098a1f236

6 years agoUse eclipse compiler in Maven build 00/121700/5
Matthias Sohn [Wed, 25 Apr 2018 01:09:05 +0000 (03:09 +0200)]
Use eclipse compiler in Maven build

Found instructions for configuring maven-compiler-plugin with ecj in
[1]. Verified that ecj run in this way raises build errors when executed
on commit d3ef5213.

Define profiles "ecj" for using Eclipse compiler and "javac" for using
javac including errorprone. By default ecj will be used.

use ecj:
$ mvn -Pecj clean install

use javac:
$ mvn -Pjavac clean install

TODO: find out how to run ecj with errorprone from Maven.

[1] https://stackoverflow.com/questions/33164976/using-eclipse-java-compiler-ecj-in-maven-builds

Change-Id: I716b603b57612b953e603387c82fd01eb1b5ca97
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoRemove trivial cases of using deprecated RefDatabase.getRefs() 99/121699/2
Matthias Sohn [Tue, 24 Apr 2018 23:44:43 +0000 (01:44 +0200)]
Remove trivial cases of using deprecated RefDatabase.getRefs()

Change-Id: I2d3e426a3391923f8a690ac68fcc33851f3eb419
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoSuppress boxing warning 98/121698/1
Matthias Sohn [Tue, 24 Apr 2018 23:08:58 +0000 (01:08 +0200)]
Suppress boxing warning

Change-Id: I90226674320841774ad691c84609e3fe71bf7852
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoRemove unnecessary semicolon 97/121697/1
Matthias Sohn [Tue, 24 Apr 2018 23:05:33 +0000 (01:05 +0200)]
Remove unnecessary semicolon

Change-Id: I3677ecb106ade209e756cec2c6f4232a416bfb44
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoFix non-externalized String warnings 96/121696/1
Matthias Sohn [Tue, 24 Apr 2018 23:05:13 +0000 (01:05 +0200)]
Fix non-externalized String warnings

- suppress warning for non-translatable texts
- externalize error messages

Change-Id: Ieba42219b2c0b51a288f5a60438a6cc4f9dfe641
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoRemove unused code and link to deprecated code 88/121688/3
Jonathan Tan [Tue, 24 Apr 2018 20:05:16 +0000 (13:05 -0700)]
Remove unused code and link to deprecated code

Eclipse reports these as errors, so remove them.

Change-Id: Ic53d8003f9faef38fe776af5a73794e7bb1dfc49
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoFile compile and API errors in JGit 02/121602/3
Michael Keppler [Mon, 23 Apr 2018 20:00:27 +0000 (22:00 +0200)]
File compile and API errors in JGit

* Photon throws null analysis errors on the repeated invocation of those
previously null checked methods. Extract them to a local variable to
avoid this. (the null analysis is configured in project properties)
* setUseProtocolV2() misses @since tag. Problem was introduced with
332bc611249d21f9b604f2c0207bf0bdfbfc3a78. Might be caused by the long
delay of 2 months from creation to merging.

Change-Id: Ibbb1a1580b604b8e7cd4bf7edc4643e292b6b4a8
Signed-off-by: Michael Keppler <Michael.Keppler@gmx.de>
6 years agoMerge "Upgrade error_prone_core to 2.3.1"
Matthias Sohn [Mon, 23 Apr 2018 23:18:34 +0000 (19:18 -0400)]
Merge "Upgrade error_prone_core to 2.3.1"

6 years agoTeach UploadPack "no-progress" in "fetch" 51/119451/10
Jonathan Tan [Thu, 1 Mar 2018 23:45:19 +0000 (15:45 -0800)]
Teach UploadPack "no-progress" in "fetch"

Add support for the "no-progress" parameter in the "fetch" command in
the fetch-pack/upload-pack protocol v2.

Change-Id: I6a6d6b1534f44845254b81d0e1f5c4ba2ac3d10b
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack "thin-pack" in "fetch" 50/119450/10
Jonathan Tan [Thu, 1 Mar 2018 22:24:16 +0000 (14:24 -0800)]
Teach UploadPack "thin-pack" in "fetch"

Add support for the "thin-pack" parameter in the "fetch" command in
the fetch-pack/upload-pack protocol v2.

Change-Id: I39a37b2b66a16929137d35c718a3acf2afb6b0b5
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack basic "fetch" command 49/119449/11
Jonathan Tan [Fri, 23 Feb 2018 22:07:02 +0000 (14:07 -0800)]
Teach UploadPack basic "fetch" command

Add basic support for the "fetch" command in the fetch-pack/upload-pack
protocol v2. This patch teaches "have" and "done".

The protocol specification (Documentation/technical/protocol-v2.txt in
the Git project) states:

    want <oid>
Indicates to the server an object which the client wants to
retrieve.  Wants can be anything and are not limited to
advertised objects.

It is unspecified whether the server should respect the
uploadpack.allowtipsha1inwant option etc. when serving packfiles. This
patch is conservative in that the server respects them.

Change-Id: I3dbec172239712ef9286a15b8407e86b87ea7863
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoTeach UploadPack "ref-prefix" in "ls-refs" 48/119448/10
Jonathan Tan [Thu, 22 Feb 2018 21:58:29 +0000 (13:58 -0800)]
Teach UploadPack "ref-prefix" in "ls-refs"

Add support for the "ref-prefix" parameter in the "ls-refs" command in
the fetch-pack/upload-pack protocol v2.

Change-Id: If9cf93b2646f75d50a11b5f482594f014d59a836
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
6 years agoUploadPack: Don't buffer ls-refs output 09/121509/2
Jonathan Nieder [Fri, 20 Apr 2018 23:34:15 +0000 (16:34 -0700)]
UploadPack: Don't buffer ls-refs output

Once we have read the user's entire command, there is no more need to
buffer our response --- even the strictest servlet engine allows
writing output once the input has been consumed.  Noticed when the
analogous code in the "fetch" command (introduced in a later patch)
overflowed its buffer:

 java.lang.OutOfMemoryError
        at java.io.ByteArrayOutputStream.hugeCapacity(ByteArrayOutputStream.java:123)
 [...]
        at org.eclipse.jgit.transport.UploadPack.sendPack(UploadPack.java:1905)
        at org.eclipse.jgit.transport.UploadPack.sendPack(UploadPack.java:1741)
        at org.eclipse.jgit.transport.UploadPack.fetchV2(UploadPack.java:1001)
        at org.eclipse.jgit.transport.UploadPack.serviceV2(UploadPack.java:1030)
        at org.eclipse.jgit.transport.UploadPack.upload(UploadPack.java:726)
        at org.eclipse.jgit.http.server.UploadPackServlet.doPost(UploadPackServlet.java:195)
        at javax.servlet.http.HttpServlet.service(HttpServlet.java:637)

Change-Id: I33df56f1cb1c6c2c25ee95426cb7ad665134ac6b

6 years agoImplement ls-refs in UploadPack 47/119447/10
Jonathan Tan [Thu, 22 Feb 2018 18:24:19 +0000 (10:24 -0800)]
Implement ls-refs in UploadPack

Implement support for Git protocol v2's "ls-refs" command and its
"symrefs" and "peel" parameters.

This adds support for this command to UploadPack but the git://,
ssh://, and git:// transports do not make use of it yet.  That will
have to wait for later patches.

Change-Id: I8abc6bcc6ed4a88c165677ff1245625aca01267b
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoUpgrade error_prone_core to 2.3.1 00/121400/2
David Pursehouse [Thu, 19 Apr 2018 10:16:11 +0000 (12:16 +0200)]
Upgrade error_prone_core to 2.3.1

and plexus-compiler-javac-errorprone to 2.8.4.

Change-Id: I36487ce53e70b54d7a04292e666540224c107b43
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoImplement protocol v2 with no capabilities in UploadPack 06/121506/4
Jonathan Tan [Thu, 22 Feb 2018 18:24:19 +0000 (10:24 -0800)]
Implement protocol v2 with no capabilities in UploadPack

Add initial support for protocol v2 of the fetch-pack/upload-pack
protocol. This protocol is described in the Git project in
"Documentation/technical/protocol-v2.txt".

This patch adds support for protocol v2 (without any capabilities) to
UploadPack. Adaptations of callers to make use of this support will
come in subsequent patches.

[jn: split from a larger patch; tweaked the API to make UploadPack
 handle parsing the extra parameters and config instead of requiring
 each caller to do such parsing]

Change-Id: I79399fa0dce533fdc8c1dbb6756748818cee45b0
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoPacketLineIn, PacketLineOut: Add support for delim-pkt 98/121498/2
Jonathan Tan [Thu, 22 Feb 2018 18:24:19 +0000 (10:24 -0800)]
PacketLineIn, PacketLineOut: Add support for delim-pkt

Most pkt-lines (data-pkts) have the form

pkt-len pkt-payload

where pkt-len is a string of 4 hexadecimal digits representing the
size in bytes of the pkt-line.  Since this size includes the size of
the pkt-len, no data-pkt has a length less than 4.

A pkt-line with a length field less than 4 can thus be used for
other purposes.  In Git protocol v1, the only such pkt-line was

flush-pkt = "0000"

which was used to mark the end of a stream.  Protocol v2 (see
Documentation/technical/protocol-v2.txt in git.git) introduces a
second special pkt-line type:

delim-pkt = "0001"

used to mark the end of a section within a stream, for example to
separate capabilities from the content of a command.

[jn: split out from a larger patch that made use of this support]

Change-Id: I10e7824fa24ed74c4f45624bd490bba978cf5c34
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoAdd RefDatabase#getRefsByPrefix method 73/120873/4
Jonathan Tan [Fri, 6 Apr 2018 22:50:01 +0000 (15:50 -0700)]
Add RefDatabase#getRefsByPrefix method

The existing RefDatabase#getRefs abstract method (to be implemented by
ref database backends) has the following issues:

 - It returns a map with a key (the name of the ref with the prefix
   removed) which is potentially superfluous (it can be derived by the
   caller if need be) and confusing (in that the prefix is removed).
 - The prefix is required to end with a '/', but some backends (e.g.
   reftable) have fast search by prefix regardless of what the last
   character of the prefix is.

Add a new method #getRefsByPrefix that does not have these issues. This
is non-abstract with a default implementation that uses #getRefs (for
backwards compatibility), but ref database backends can reimplement it.

This also prepares for supporting "ref-prefix" in the "ls-refs" command
in the fetch-pack/upload-pack protocol v2, which does not require that
the prefix end with a '/'.

Change-Id: I4c92f852e8c1558095dd460b5fd7b602c1d82df1
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
6 years agoHandle Gerrit Change-Ids for merge commits 71/120871/4
Thomas Wolf [Fri, 6 Apr 2018 21:58:34 +0000 (23:58 +0200)]
Handle Gerrit Change-Ids for merge commits

Otherwise successful, non-conflicting merges will never get a
Gerrit Change-Id.

Bug: 358206
Change-Id: I9b599ad01d9f7332200c1d81a1ba6ce5ef990ab5
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
6 years agoPushProcess: Remove unused import of HashMap 08/121208/1
David Pursehouse [Mon, 16 Apr 2018 14:51:32 +0000 (16:51 +0200)]
PushProcess: Remove unused import of HashMap

Bug: 533621
Change-Id: Ie125818d5fc8fef5800108db2ca21151d8d3bfa3
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
6 years agoConfigure WindowCache settings to use in JGit CLI 95/118595/3
Matthias Sohn [Sun, 4 Mar 2018 09:42:46 +0000 (10:42 +0100)]
Configure WindowCache settings to use in JGit CLI

Set the same defaults as in EGit.
Use mmap to map git packfiles into memory.

Change-Id: I7997b9123448a762192d7eeb55e01432549fba98
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
6 years agoMerge "Push: Ensure ref updates are processed in input order"
David Pursehouse [Fri, 13 Apr 2018 21:43:12 +0000 (17:43 -0400)]
Merge "Push: Ensure ref updates are processed in input order"