]> source.dussan.org Git - jgit.git/log
jgit.git
8 years agoMerge "Fix CheckoutCommand to return updated files even on NONDELETED status"
Christian Halstrick [Mon, 10 Oct 2016 14:19:07 +0000 (10:19 -0400)]
Merge "Fix CheckoutCommand to return updated files even on NONDELETED status"

8 years agoDelay inserter.flush in WalkFetchConnection 03/82303/2
Zhen Chen [Fri, 30 Sep 2016 19:11:46 +0000 (12:11 -0700)]
Delay inserter.flush in WalkFetchConnection

Spawn an ObjectReader from the ObjectInserter, so the flush can be
delayed at the end of the fetch.

Change-Id: I35fe8c8370c06c25262645202aec2b1318057c19
Signed-off-by: Zhen Chen <czhen@google.com>
8 years agoRemove unused code from class Repository 11/81911/2
Rüdiger Herrmann [Mon, 26 Sep 2016 13:47:48 +0000 (15:47 +0200)]
Remove unused code from class Repository

The package-private method Repository::gitInternalSlash() is not
referenced from anywhere within the package.

Last uses were removed with
0f8743d4 "Remove deprecated Tree, TreeEntry, FileTreeEntry and friends"
6e9fdce9 "Kill GitIndex"

Signed-off-by: Rüdiger Herrmann <ruediger.herrmann@gmx.de>
Change-Id: I514bf684ad0da808f6523e9e46db9674a25e1fb5

8 years agoFix CheckoutCommand to return updated files even on NONDELETED status 98/81998/1
Christian Halstrick [Tue, 27 Sep 2016 13:38:53 +0000 (15:38 +0200)]
Fix CheckoutCommand to return updated files even on NONDELETED status

CheckoutCommand was not returning updated and removed files in case of
an overall status of NONDELETED. That's status which occurs especially
on the Windows platform when Checkout wanted to delete files but the
filesystem doesn't allow this. The situation is more seldom on linux/mac
because open filehandles don't stop a deletion attempt and checkout
succeeds more often.

Change-Id: I4828008e58c09bd8f9edaf0f7eda0a79c629fb57

8 years agoMerge branch 'stable-4.5' 44/81844/1
Matthias Sohn [Sat, 24 Sep 2016 08:51:05 +0000 (10:51 +0200)]
Merge branch 'stable-4.5'

* stable-4.5:
  Fix carrying over flags during a RevWalk

Change-Id: Ibf4573c5664271dfa7a6ecc3ede6eaad749f89d8
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge "Fix carrying over flags during a RevWalk" into stable-4.5
Shawn Pearce [Sat, 24 Sep 2016 05:16:47 +0000 (01:16 -0400)]
Merge "Fix carrying over flags during a RevWalk" into stable-4.5

8 years agoFix carrying over flags during a RevWalk 91/81491/2
Christian Halstrick [Mon, 12 Sep 2016 23:29:11 +0000 (01:29 +0200)]
Fix carrying over flags during a RevWalk

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

8 years agoMerge "Use consistent feature and category names"
Matthias Sohn [Fri, 23 Sep 2016 08:57:16 +0000 (04:57 -0400)]
Merge "Use consistent feature and category names"

8 years agoRepositoryCache#unregisterAndCloseRepository: Remove unused db parameter 07/81707/2
David Pursehouse [Thu, 22 Sep 2016 14:54:33 +0000 (16:54 +0200)]
RepositoryCache#unregisterAndCloseRepository: Remove unused db parameter

Change-Id: Ibee48cba85089324dc8db45066d311ad5db4f3ee
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years ago[releng] maven: compile against the Java 1.8 libraries 45/81745/1
Thomas Wolf [Fri, 23 Sep 2016 05:37:21 +0000 (07:37 +0200)]
[releng] maven: compile against the Java 1.8 libraries

Remove the JDK_HOME settings that were introduced to build against the
Java 1.7 libraries. Now that the minimum requirement is Java 1.8, this
is wrong and prevents use of Java 1.8 features.

Change-Id: I91a194c9449d7810ef02b038907dbbc708e600a5
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
8 years agoUse consistent feature and category names 12/81712/2
Matthias Sohn [Thu, 22 Sep 2016 15:39:03 +0000 (17:39 +0200)]
Use consistent feature and category names

These names are presented to users when they install JGit so they should
be consistent.

Bug: 496101
Change-Id: I88a1abd8c3717d9a2f958e3a39edb2dbce3a415b
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agopgm, test: Add missing dependency for buck build 67/81667/2
David Pursehouse [Thu, 22 Sep 2016 09:30:14 +0000 (11:30 +0200)]
pgm, test: Add missing dependency for buck build

Change-Id: I26d69fb6d35c6fb120360ef143d1b1f565d4014c
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoDfsBlockCache: Suppress warnings about unchecked conversion of Ref<T> 44/81544/3
David Pursehouse [Wed, 21 Sep 2016 08:07:34 +0000 (10:07 +0200)]
DfsBlockCache: Suppress warnings about unchecked conversion of Ref<T>

There are already suppressions for the same warnings in other parts
of this class.

Change-Id: Ic3b45525c6c8200cba975d14c7650cedb4409a4d
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoLfsServerTest: Remove unnecessary suppression of 'restriction' warning 42/81542/4
David Pursehouse [Wed, 21 Sep 2016 07:54:22 +0000 (09:54 +0200)]
LfsServerTest: Remove unnecessary suppression of 'restriction' warning

Change-Id: I23386808848746d201ca1ac13f114c06fc8e86c5
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoWrappedRequest: Don't use deprecated ServletRequestWrapper#realPath 41/81541/4
David Pursehouse [Wed, 21 Sep 2016 07:53:14 +0000 (09:53 +0200)]
WrappedRequest: Don't use deprecated ServletRequestWrapper#realPath

Change-Id: I268e66d2299a1a12f7ae44b67a6b947339038245
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoSuppress deprecation warnings related to UploadPackLogger 33/81533/4
David Pursehouse [Wed, 21 Sep 2016 07:07:24 +0000 (09:07 +0200)]
Suppress deprecation warnings related to UploadPackLogger

UploadPackLogger is deprecated but will not be removed until
JGit version 5.0.

Suppress the unavoidable deprecation warnings on usages of the
interface that are kept for backwards compatibility.

Add a TODO so that we don't forget to remove it in 5.0.

Change-Id: Id248002b9bdf23db192427196d54c722a012106c
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoAppServer: Remove unnecessary 'unused' suppressions 32/81532/4
David Pursehouse [Wed, 21 Sep 2016 07:03:29 +0000 (09:03 +0200)]
AppServer: Remove unnecessary 'unused' suppressions

Change-Id: Ic27106f38af14833147f739179c3ef8ec6b6ee31
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoHttpClientConnection: Don't use deprecated HttpClient classes 05/81505/4
David Pursehouse [Tue, 20 Sep 2016 19:19:44 +0000 (21:19 +0200)]
HttpClientConnection: Don't use deprecated HttpClient classes

- raise minimum version for HttpClient packages to 4.3 since some of the
used classes aren't available in older versions
- recompute OSGi uses clauses

Change-Id: I8f0bff1433762561e02f7439db27a6a9e846c290
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoFix wrong @since tags introduced by 1beb3ccd 41/81641/1
Matthias Sohn [Wed, 21 Sep 2016 21:37:18 +0000 (23:37 +0200)]
Fix wrong @since tags introduced by 1beb3ccd

1beb3ccd1ed3be0dde703a604deeff3a4b9f27eb was submitted in 4.6.0-SNAPSHOT

Change-Id: I33e027dbd0995ba36491bd9e2d4b3c84f9f0abd6
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge branch 'stable-4.5' 89/81589/2
Matthias Sohn [Wed, 21 Sep 2016 16:08:00 +0000 (18:08 +0200)]
Merge branch 'stable-4.5'

* stable-4.5:
  Turn off doclint also during Maven site generation
  Prepare 4.5.1-SNAPSHOT builds
  JGit v4.5.0.201609210915-r
  Unconditionally close repository in unregisterAndCloseRepository

Change-Id: Ibfd11669cd74d2e62b014c18fd39b646b200c8c5
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoTurn off doclint also during Maven site generation 06/81606/2
Matthias Sohn [Wed, 21 Sep 2016 15:01:13 +0000 (17:01 +0200)]
Turn off doclint also during Maven site generation

Change-Id: Iefc77114de21e7a101642f5c3a8f0fb317886ba2
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoPrepare 4.5.1-SNAPSHOT builds 03/81603/1
Matthias Sohn [Wed, 21 Sep 2016 15:05:27 +0000 (17:05 +0200)]
Prepare 4.5.1-SNAPSHOT builds

Change-Id: I3305e8a09a3fb06f25a316cff2bdbb551d3ade68
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoJGit v4.5.0.201609210915-r 86/81586/1 v4.5.0.201609210915-r
Matthias Sohn [Wed, 21 Sep 2016 13:06:18 +0000 (15:06 +0200)]
JGit v4.5.0.201609210915-r

Change-Id: Idc02a1a1d74f84605d764c239803f0cfbad94eb7
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoUnconditionally close repository in unregisterAndCloseRepository 63/81563/2
Saša Živkov [Wed, 21 Sep 2016 09:32:36 +0000 (11:32 +0200)]
Unconditionally close repository in unregisterAndCloseRepository

Repository.close() method is used when reference counting and expiration
needs to be honored. The RepositoryCache.unregisterAndCloseRepository
method should close the repository unconditionally. This is also indicated
from its javadoc.

Change-Id: I19392d1eaa17f27ae44b55eea49dcff05a52f298

8 years agoMerge "Add ReceivedPackStatistics for PackParser"
Shawn Pearce [Tue, 20 Sep 2016 19:29:24 +0000 (15:29 -0400)]
Merge "Add ReceivedPackStatistics for PackParser"

8 years agoChange JGit minimum execution environment to JavaSE-1.8 82/79782/4
Matthias Sohn [Thu, 25 Aug 2016 23:36:28 +0000 (01:36 +0200)]
Change JGit minimum execution environment to JavaSE-1.8

Bug: 500059
Change-Id: I47f3f6749a67da52029f84e002d9b155ed56d2b7
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMove AtomicObjectOutputStream to lfs/internal 88/76588/8
Christian Halstrick [Tue, 5 Jul 2016 07:05:59 +0000 (09:05 +0200)]
Move AtomicObjectOutputStream to lfs/internal

The class AtomicObjectOutputStream should be available to all lfs
related classes, not only to the server side. Move the class from
org.eclipse.jgit.lfs.server.fs to org.eclipse.jgit.lfs.internal to
achieve that.

Change-Id: I028e1c9ec7c21f316340b21d558b9a6b77e2060d

8 years agoAdd built-in LFS smudge filter for local case 31/76131/10
Christian Halstrick [Mon, 27 Jun 2016 14:00:44 +0000 (16:00 +0200)]
Add built-in LFS smudge filter for local case

Adds a JGit built-in implementation of the "git lfs smudge" filter. This
filter should do the same as the one described in [1] besides that it
only supports the local case when the lfs objects are already present in
the media directory. Remote cases where download of LFS objects from an
LFS server is needed will be done in a later commit.

[1] https://github.com/github/git-lfs/blob/master/docs/man/git-lfs-smudge.1.ronn

Change-Id: I8ff661d4edd3667ef7f86f3b4fa33e568eb4c8f4

8 years agoAdd built-in LFS clean filter 30/76130/9
Christian Halstrick [Mon, 27 Jun 2016 14:00:09 +0000 (16:00 +0200)]
Add built-in LFS clean filter

Adds a JGit built-in implementation of the "git lfs clean" filter. This
filter should do the same as the one described in [1]. But since this
filter is written in Java and can be called by JGit without forking new
processes it should be much faster

[1]
https://github.com/github/git-lfs/blob/master/docs/man/git-lfs-clean.1.ronn

Change-Id: If60e387e97870245b4bd765eda6717eb84cffb1d

8 years agoAdd configuration parameter to enable built-in hooks/filters 84/71984/12
Christian Halstrick [Tue, 3 May 2016 11:41:39 +0000 (13:41 +0200)]
Add configuration parameter to enable built-in hooks/filters

If the configuration parameter filter.<filterDriverName>.useJGitBuiltin
is set to true then for all corresponding filters JGit will try to
execute the built-in filter instead of the filter-command which is
defined in git configuration. It will fallback to the non-built-in
filters if no built-in filters are registered or if constructing them
leads to exceptions. If set to false JGit will not try to execute
built-in filters for the specified filter driver.

Example: The configuration contains the following lines

  [filter "lfs"]
    clean = git-lfs clean -- %f
    smudge = git-lfs smudge -- %f
    useJGitBuiltin = true

Addtionally the .gitattributes file in the root of the working tree
contains:

    *.bin filter=lfs

In this case when new content is added similar to "git add 1.bin" then
the following will happen:

  - jgit will check whether a built-in command factory was registered
    for the command "jgit://builtin/lfs/clean". If that is true the
    factory is used to create a built-in filter command and that
    command is used to filter the content
  - Otherwise jgit will call the external program "git lfs clean ..."
    to do the filtering

Change-Id: Idadb1db06b1e89e7031d7ed6319904973c367d38

8 years agoAdd support for built-in smudge filters 83/71983/12
Christian Halstrick [Tue, 3 May 2016 12:59:57 +0000 (14:59 +0200)]
Add support for built-in smudge filters

JGit supports smudge filters defined in repository configuration. The
filters are implemented as external programs filtering content by
accepting the original content (as seen in git's object database) on
stdin and which emit the filtered content on stdout. This content is
then written to the file in the working tree. To run such a filter JGit
has to start an external process and pump data into/from this process.

This commit adds support for built-in smudge filters which are
implemented in Java and which are executed by jgit's main thread. When a
filter is defined in the configuration as
"jgit://builtin/<filterDriverName>/smudge" then JGit will lookup in a
static map whether a builtin filter is registered under this name. If
found such a filter is called to do the filtering.

The functionality in this commit requires that a program using JGit
explicitly calls the JGit API to register built-in implementations for
specific smudge filters. In follow-up commits configuration parameters
will be added which trigger such registrations.

Change-Id: Ia743aa0dbed795e71e5792f35ae55660e0eb3c24

8 years agoAdd support for built-in clean filters 82/71982/11
Christian Halstrick [Mon, 2 May 2016 13:24:26 +0000 (15:24 +0200)]
Add support for built-in clean filters

JGit supports clean filters defined in repository configuration. The
filters are implemented as external programs filtering content by
accepting the original content (as seen in the working tree) on stdin
and which emit the filtered content on stdout. To run such a filter JGit
has to start an external process and pump data into/from this process.

This commit adds support for clean filters which are implemented
in Java and which are executed by jgit's main thread. When a filter is
defined in the configuration as
"jgit://builtin/<filterDriverName>/clean" then JGit will lookup in a
static map whether a filter is registered under this name. If found
such a filter is called to do the filtering.

The functionality in this commit requires that a program using JGit
explicitly calls the JGit API to register built-in implementations for
specific clean filters. In follow-up commits configuration parameters
will be added which trigger such registrations. Other commits will add
implementations for lfs filters.

Change-Id: I0344d3c54801c9a46e5a606c5df17e5f2e17b2be

8 years agoAdd ReceivedPackStatistics for PackParser 50/81050/4
Masaya Suzuki [Wed, 14 Sep 2016 00:43:18 +0000 (17:43 -0700)]
Add ReceivedPackStatistics for PackParser

This is like PackStatistics, but for PackParser.

Change-Id: I854215c0956fd0b36843d631780be303e021b8be
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
8 years agoPrepare 4.6.0-SNAPSHOT builds 63/81363/1
Matthias Sohn [Mon, 19 Sep 2016 15:06:42 +0000 (17:06 +0200)]
Prepare 4.6.0-SNAPSHOT builds

Change-Id: Id2eafc331ee32c332c2a9b867b05c260beb0d10f
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoHandle all values of branch.[name].rebase 31/79031/3
Thomas Wolf [Mon, 15 Aug 2016 05:55:44 +0000 (07:55 +0200)]
Handle all values of branch.[name].rebase

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>
8 years agoIgnore trailing spaces in directory rule patterns 58/81058/3
Andrey Loskutov [Wed, 14 Sep 2016 07:29:20 +0000 (09:29 +0200)]
Ignore trailing spaces in directory rule patterns

Bug: 500967
Change-Id: I7fabc2654af97011c62f46d5c30ee992341e45e2
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>
8 years agoAdd support for post-commit hooks 65/80865/3
Martin Goellnitz [Sat, 10 Sep 2016 09:29:30 +0000 (11:29 +0200)]
Add support for post-commit hooks

Change-Id: I6691b454404dd4db3c690ecfc7515de765bc2ef7
Signed-off-by: Martin Goellnitz <m.goellnitz@outlook.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoDon't log error if system git config does not exist 60/80360/3
Matthias Sohn [Sun, 4 Sep 2016 10:07:37 +0000 (12:07 +0200)]
Don't log error if system git config does not exist

- 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>
8 years agoAdd missing dependency to slf4j-log4j bridge 52/80352/2
Matthias Sohn [Sat, 3 Sep 2016 09:10:38 +0000 (11:10 +0200)]
Add missing dependency to slf4j-log4j bridge

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>
8 years agoRename FSJava7Test to FSTest 51/80351/1
Matthias Sohn [Fri, 2 Sep 2016 22:03:00 +0000 (00:03 +0200)]
Rename FSJava7Test to FSTest

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>
8 years agoImprove JavaDoc for LfsProtocolServlet.getLargeFileRepository 62/80062/2
Matthias Sohn [Tue, 30 Aug 2016 19:24:19 +0000 (21:24 +0200)]
Improve JavaDoc for LfsProtocolServlet.getLargeFileRepository

Guide implementors which exception to throw in case of errors.

Change-Id: I74fb76cdf6b7cdef513f3fe8c144572e869cc533
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoCLI: implement option -d for deleting tags 27/79327/4
Ned Twigg [Fri, 18 Mar 2016 10:08:44 +0000 (03:08 -0700)]
CLI: implement option -d for deleting tags

Change-Id: I438456b76aefd361384729686271288186d3be3b
Signed-off-by: Ned Twigg <ned.twigg@diffplug.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoAdded a clean command. 24/79324/5
Ned Twigg [Fri, 18 Mar 2016 07:54:52 +0000 (00:54 -0700)]
Added a clean command.

Change-Id: I05d5392789b5b64e6ee44f678556cf25dc30d7ba
Signed-off-by: Ned Twigg <ned.twigg@diffplug.com>
8 years agoAdd specific exception for LFS unavailable 08/79908/2
David Pursehouse [Mon, 29 Aug 2016 03:18:38 +0000 (12:18 +0900)]
Add specific exception for LFS unavailable

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>
8 years agoAdd Javadoc for LFS exception messages 07/79907/2
David Pursehouse [Mon, 29 Aug 2016 03:04:42 +0000 (12:04 +0900)]
Add Javadoc for LFS exception messages

Change-Id: I4967b5f7a9a7e9488d20f60aaa949efe97442429
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoReceivePack: integrate push option parsing into recvCommands 57/79957/2
Shawn Pearce [Mon, 29 Aug 2016 19:11:11 +0000 (12:11 -0700)]
ReceivePack: integrate push option parsing into recvCommands

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

8 years agoReceivePack: simplify getPushOptions 56/79956/2
Shawn Pearce [Mon, 29 Aug 2016 19:12:23 +0000 (12:12 -0700)]
ReceivePack: simplify getPushOptions

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

8 years agoReceivePack: allow push options to be set 55/79955/3
Shawn Pearce [Mon, 29 Aug 2016 18:54:15 +0000 (11:54 -0700)]
ReceivePack: allow push options to be set

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

8 years agoMake GC.RepoStatistics static 89/79889/1
Jonathan Nieder [Fri, 26 Aug 2016 23:28:43 +0000 (16:28 -0700)]
Make GC.RepoStatistics static

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

Change-Id: Ie271e9e789a5bc764a287f7bd2a8c8af297d7d41

8 years agoReceivePack: refactor push option parsing 87/79887/4
Shawn Pearce [Fri, 26 Aug 2016 21:44:44 +0000 (14:44 -0700)]
ReceivePack: refactor push option parsing

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

8 years agoFix push option initalization on HTTP 86/79886/2
Stefan Beller [Fri, 26 Aug 2016 21:10:06 +0000 (14:10 -0700)]
Fix push option initalization on HTTP

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>
8 years agoMerge "DfsReader: check object type during open"
Shawn Pearce [Fri, 26 Aug 2016 02:45:31 +0000 (22:45 -0400)]
Merge "DfsReader: check object type during open"

8 years agoMerge "Fix HttpClientConnection leaking temporary buffer files"
Shawn Pearce [Fri, 26 Aug 2016 02:31:08 +0000 (22:31 -0400)]
Merge "Fix HttpClientConnection leaking temporary buffer files"

8 years agoDfsReader: check object type during open 86/79786/2
Shawn Pearce [Fri, 26 Aug 2016 01:59:15 +0000 (18:59 -0700)]
DfsReader: check object type during open

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

8 years agoClarify the semantics of DfsRefDatabase#compareAndPut 33/77433/5
Masaya Suzuki [Fri, 26 Aug 2016 01:25:48 +0000 (18:25 -0700)]
Clarify the semantics of DfsRefDatabase#compareAndPut

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>
8 years agoMerge "Do not fake a SymbolicRef as an ObjectIdRef"
Shawn Pearce [Fri, 26 Aug 2016 00:57:39 +0000 (20:57 -0400)]
Merge "Do not fake a SymbolicRef as an ObjectIdRef"

8 years agoNoteMapTest: Add missing @Test annotations 03/53103/2
Dave Borowitz [Mon, 3 Aug 2015 20:58:39 +0000 (13:58 -0700)]
NoteMapTest: Add missing @Test annotations

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>
8 years agoAdd a RepeatRule to help repeating flaky tests 31/79731/2
Matthias Sohn [Tue, 24 May 2016 23:07:18 +0000 (01:07 +0200)]
Add a RepeatRule to help repeating flaky tests

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>
8 years agoDo not fake a SymbolicRef as an ObjectIdRef 16/77416/9
Masaya Suzuki [Wed, 13 Jul 2016 23:57:03 +0000 (16:57 -0700)]
Do not fake a SymbolicRef as an ObjectIdRef

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]: https://eclipse.googlesource.com/jgit/jgit/+/01b5392cdbc12ce2e21fd1d1afbd61fdf97e1c38
[2]: https://eclipse.googlesource.com/jgit/jgit/+/3a86868c0883d2a564db88bf9ae4a5fe235bb63f

Change-Id: I118f85f0414dbfad02250944e28d74dddd59469b
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
8 years agoLfsProtocolServlet: Add support for insufficient storage error 15/79315/4
David Pursehouse [Fri, 19 Aug 2016 02:29:00 +0000 (11:29 +0900)]
LfsProtocolServlet: Add support for insufficient storage error

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>
8 years agoPacket logging for JGit 72/79472/9
Dan Wang [Tue, 23 Aug 2016 02:16:51 +0000 (19:16 -0700)]
Packet logging for JGit

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< 2632c897f74b1cc9b5533f467da459b9ec725538 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< e0c1ceafc5bece92d35773a75fff59497e1d9bd5 refs/heads/maint

Change-Id: I5028c064f3ac090510386057cb4e6d30d4eae232
Signed-off-by: Dan Wang <dwwang@google.com>
8 years agopush: Do not use push options unless requested 53/79653/3
Stefan Beller [Wed, 24 Aug 2016 19:47:10 +0000 (12:47 -0700)]
push: Do not use push options unless requested

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>
8 years agoFix HttpClientConnection leaking temporary buffer files 22/79622/1
Matthias Sohn [Wed, 24 Aug 2016 12:08:00 +0000 (14:08 +0200)]
Fix HttpClientConnection leaking temporary buffer files

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>
8 years agoUse FS#lastModified instead of File#lastModified 07/79407/5
Masaya Suzuki [Sun, 21 Aug 2016 22:27:30 +0000 (15:27 -0700)]
Use FS#lastModified instead of File#lastModified

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>
8 years agoDon't check lastModified, length on folders for submodules 86/78586/4
Christian Halstrick [Sun, 7 Aug 2016 17:59:49 +0000 (18:59 +0100)]
Don't check lastModified, length on folders for submodules

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

Bug: 498759
Change-Id: If5db69ef3868e475ac477d3e8a7750b268799b0c

8 years agoAdded Java 7 launch config with LANG env. variable set 82/79582/1
Andrey Loskutov [Wed, 24 Aug 2016 07:05:22 +0000 (09:05 +0200)]
Added Java 7 launch config with LANG env. variable set

This avoids symlink test errors on Linux

Change-Id: Id8193524c40394a90b8315ab0b8256670d618cb5
Signed-off-by: Andrey Loskutov <loskutov@gmx.de>
8 years agoAdd HTTP status code to ServiceMayNotContinueException 37/79237/6
Masaya Suzuki [Wed, 24 Aug 2016 00:57:59 +0000 (17:57 -0700)]
Add HTTP status code to ServiceMayNotContinueException

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>
8 years agoLfsProtocolServlet: Add support for rate limit and bandwidth limit errors 14/79314/2
David Pursehouse [Fri, 19 Aug 2016 01:58:06 +0000 (10:58 +0900)]
LfsProtocolServlet: Add support for rate limit and bandwidth limit errors

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>
8 years agoLfsProtocolServlet: Always include message in error response 14/79214/4
David Pursehouse [Wed, 17 Aug 2016 16:18:15 +0000 (01:18 +0900)]
LfsProtocolServlet: Always include message in error response

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>
8 years agoLfsRepository{NotFound,ReadOnly}: Externalize message strings 46/79246/2
David Pursehouse [Thu, 18 Aug 2016 06:58:43 +0000 (15:58 +0900)]
LfsRepository{NotFound,ReadOnly}: Externalize message strings

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>
8 years agoMerge "DownloadTest: Use enums from org.apache.http.HttpStatus"
Matthias Sohn [Fri, 19 Aug 2016 23:19:13 +0000 (19:19 -0400)]
Merge "DownloadTest: Use enums from org.apache.http.HttpStatus"

8 years agoMerge "Ignore IOException thrown from close"
Shawn Pearce [Fri, 19 Aug 2016 23:10:58 +0000 (19:10 -0400)]
Merge "Ignore IOException thrown from close"

8 years agoDownloadTest: Use enums from org.apache.http.HttpStatus 12/78912/7
David Pursehouse [Fri, 12 Aug 2016 03:47:34 +0000 (12:47 +0900)]
DownloadTest: Use enums from org.apache.http.HttpStatus

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>
8 years agoIgnore IOException thrown from close 86/79386/2
Masaya Suzuki [Fri, 19 Aug 2016 20:51:05 +0000 (13:51 -0700)]
Ignore IOException thrown from close

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>
8 years agoDfsObjDatabase: clear PackList dirty bit if no new packs 74/79374/2
Shawn Pearce [Fri, 19 Aug 2016 18:51:40 +0000 (11:51 -0700)]
DfsObjDatabase: clear PackList dirty bit if no new packs

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

8 years agoEnhance ResetCommand to allow disabling reflog update 62/77362/3
Matthias Sohn [Thu, 14 Jul 2016 22:13:01 +0000 (00:13 +0200)]
Enhance ResetCommand to allow disabling reflog update

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>
8 years agoLfsProtocolServlet: Always set the Content-Type header on response 09/79009/2
David Pursehouse [Sat, 13 Aug 2016 07:56:20 +0000 (16:56 +0900)]
LfsProtocolServlet: Always set the Content-Type header on response

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>
8 years agoLfsProtocolServlet: Don't set pretty printing on Gson 53/78953/2
David Pursehouse [Fri, 12 Aug 2016 13:26:26 +0000 (22:26 +0900)]
LfsProtocolServlet: Don't set pretty printing on Gson

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>
8 years agoMerge "BaseReceivePack: null and IllegalStateException cases for getPushOptions"
Jonathan Nieder [Mon, 15 Aug 2016 19:40:49 +0000 (15:40 -0400)]
Merge "BaseReceivePack: null and IllegalStateException cases for getPushOptions"

8 years agoBaseReceivePack: null and IllegalStateException cases for getPushOptions 67/78367/5
Dan Wang [Wed, 3 Aug 2016 01:34:07 +0000 (18:34 -0700)]
BaseReceivePack: null and IllegalStateException cases for getPushOptions

Change-Id: I5c6790719991931d615d821c900bfd90a20e540b
Signed-off-by: Dan Wang <dwwang@google.com>
8 years agoUpdate .mailmap 59/78659/3
David Pursehouse [Tue, 9 Aug 2016 01:46:18 +0000 (10:46 +0900)]
Update .mailmap

Change-Id: I167f9e4746b8aa814b361e884c3c9a25e95b9d54
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoLfsProtocolServlet: Reuse existing Writer when sending error response 88/78588/4
David Pursehouse [Mon, 8 Aug 2016 00:40:34 +0000 (09:40 +0900)]
LfsProtocolServlet: Reuse existing Writer when sending error response

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>
8 years agoRepoCommand: Avoid group lists shadowing groups strings 60/78660/1
Jonathan Nieder [Tue, 9 Aug 2016 01:50:01 +0000 (18:50 -0700)]
RepoCommand: Avoid group lists shadowing groups strings

Reported-by: David Pursehouse <david.pursehouse@gmail.com>
Change-Id: I9e9b021d335bda4d58b6bcc30f59b81ac5b37724
Signed-off-by: Jonathan Nieder <jrn@google.com>
8 years agoPackWriter: Fix Javadoc tag for thrown exception in preparePack 57/78657/1
David Pursehouse [Tue, 9 Aug 2016 01:26:34 +0000 (10:26 +0900)]
PackWriter: Fix Javadoc tag for thrown exception in preparePack

Use @throws instead of @param

Change-Id: Ic9419d254c617e60a9b10e49205b11069442eb27
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
8 years agoMerge "Document new PackWriter#preparePack variant's parameters and exceptions"
Jonathan Nieder [Mon, 8 Aug 2016 23:55:57 +0000 (19:55 -0400)]
Merge "Document new PackWriter#preparePack variant's parameters and exceptions"

8 years agoDocument new PackWriter#preparePack variant's parameters and exceptions 55/78655/1
Jonathan Nieder [Mon, 8 Aug 2016 23:38:21 +0000 (16:38 -0700)]
Document new PackWriter#preparePack variant's parameters and exceptions

Change-Id: Id4fa272c611a855bf4ef1bf5399f3e4305664103

8 years agoSilence API errors in LfsProtocolServlet 22/78622/2
Matthias Sohn [Mon, 8 Aug 2016 15:10:53 +0000 (17:10 +0200)]
Silence API errors in LfsProtocolServlet

bb9988c2 changed the signature of getLargeFileRepository() which is only
breaking implementors which is ok according to OSGi semantic versioning
rules.

Change-Id: I68bda7900b72e217571f74aee53705167f8100a2
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoAdd missing @since tags for new API 21/78621/2
Matthias Sohn [Mon, 8 Aug 2016 15:07:29 +0000 (17:07 +0200)]
Add missing @since tags for new API

Change-Id: I8db29a0313fbc476152cef47f2eaa76954f1e280
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge "Require-Bundle com.jcraft.jsch replaced by Import-Package statement"
Matthias Sohn [Mon, 8 Aug 2016 22:43:54 +0000 (18:43 -0400)]
Merge "Require-Bundle com.jcraft.jsch replaced by Import-Package statement"

8 years agoRequire-Bundle com.jcraft.jsch replaced by Import-Package statement 78/78578/6
Jens Offenbach [Sat, 6 Aug 2016 15:30:42 +0000 (11:30 -0400)]
Require-Bundle com.jcraft.jsch replaced by Import-Package statement

Bug: 359288
Change-Id: Ifbbf953f5389c6bd3ba960b598c0e92656b522e3
Signed-off-by: Jens Offenbach <wolle5050@gmx.de>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge changes Ib0d8c294,Idfb83482 44/78644/2
Jonathan Nieder [Mon, 8 Aug 2016 19:46:56 +0000 (12:46 -0700)]
Merge changes Ib0d8c294,Idfb83482

* changes:
  Shallow fetch: Pass along "shallow"s in unparsed-wants case, too
  Shallow fetch: Pass a DepthWalk to PackWriter

Change-Id: I7d1c3b4d0b7ebc254b53404d1618522b0174ac23

8 years agoShallow fetch: Pass along "shallow"s in unparsed-wants case, too 43/78643/3
Jonathan Nieder [Mon, 8 Aug 2016 19:35:36 +0000 (12:35 -0700)]
Shallow fetch: Pass along "shallow"s in unparsed-wants case, too

Since 84d2738ff21c (Don't skip want validation when the client sends no
haves, 2013-06-21), this branch is not taken.  Process the
"shallow"s anyway as a defensive measure in case the code path gets
revived.

Change-Id: Idfb834825d77f51e17191c1635c9d78c78738cfd
Signed-off-by: Jonathan Nieder <jrn@google.com>
8 years agoShallow fetch: Pass a DepthWalk to PackWriter 42/78642/2
Jonathan Nieder [Mon, 8 Aug 2016 19:31:39 +0000 (12:31 -0700)]
Shallow fetch: Pass a DepthWalk to PackWriter

d385a7a5e5ca (Shallow fetch: Respect "shallow" lines, 2016-08-03) forgot
that UploadPack wasn't passing a DepthWalk to PackWriter in the first
place.  As a result, shallow clones fail:

  java.lang.IllegalArgumentException: Shallow packs require a DepthWalk
        at org.eclipse.jgit.internal.storage.pack.PackWriter.preparePack(PackWriter.java:756)
        at org.eclipse.jgit.transport.UploadPack.sendPack(UploadPack.java:1497)
        at org.eclipse.jgit.transport.UploadPack.sendPack(UploadPack.java:1381)
        at org.eclipse.jgit.transport.UploadPack.service(UploadPack.java:774)
        at org.eclipse.jgit.transport.UploadPack.upload(UploadPack.java:667)
        at org.eclipse.jgit.http.server.UploadPackServlet.doPost(UploadPackServlet.java:191)

Change-Id: Ib0d8c2946eebfea910a2b767fb92e23da15d4749

8 years agoFix non-parameterized generic type warning 56/76556/3
Matthias Sohn [Mon, 4 Jul 2016 21:58:56 +0000 (23:58 +0200)]
Fix non-parameterized generic type warning

Change-Id: Ib857166f64420aebf7c31d72825cac44bd770dbd
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge "Add path src/ to source path in build.properties"
Matthias Sohn [Mon, 8 Aug 2016 06:42:20 +0000 (02:42 -0400)]
Merge "Add path src/ to source path in build.properties"

8 years agoMerge "Skip cleaning inner repositories by default in CleanCommand"
Christian Halstrick [Sun, 7 Aug 2016 13:37:47 +0000 (09:37 -0400)]
Merge "Skip cleaning inner repositories by default in CleanCommand"

8 years agoMerge "Add testCleanDirsWithSubmodule test to CleanCommandTest"
Christian Halstrick [Sun, 7 Aug 2016 13:37:20 +0000 (09:37 -0400)]
Merge "Add testCleanDirsWithSubmodule test to CleanCommandTest"

8 years agoAdd path src/ to source path in build.properties 76/77376/2
Matthias Sohn [Fri, 15 Jul 2016 06:41:23 +0000 (08:41 +0200)]
Add path src/ to source path in build.properties

This fixes the warning "src/ is missing from source.."

Change-Id: I166e3a6a3d5230e4110d3283ec4dbc7d1dfe6732
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
8 years agoMerge changes I27961679,I91be6165,If0dbd562
Jonathan Nieder [Sat, 6 Aug 2016 00:36:08 +0000 (20:36 -0400)]
Merge changes I27961679,I91be6165,If0dbd562

* changes:
  LfsProtocolServlet: Allow access to objects in request
  LfsProtocolServlet: Allow getLargeFileRepository to raise exceptions
  Remove references to org.eclipse.jgit.java7

8 years agoShallow fetch/clone: Make --depth mean the total history depth 81/78481/5
Terry Parker [Thu, 4 Aug 2016 18:14:33 +0000 (11:14 -0700)]
Shallow fetch/clone: Make --depth mean the total history depth

cgit changed the --depth parameter to mean the total depth of history
rather than the depth of ancestors to be returned [1]. JGit still uses
the latter meaning, so update it to match cgit.

depth=0 still means a non-shallow clone. depth=1 now means only the
wants rather than the wants and their direct parents.

This is accomplished by changing the semantic meaning of "depth" in
UploadPack and PackWriter to mean the total depth of history desired,
while keeping "depth" in DepthWalk.{RevWalk,ObjectWalk} to mean
the depth of traversal. Thus UploadPack and PackWriter always
initialize their DepthWalks with "depth-1".

[1] upload-pack: fix off-by-one depth calculation in shallow clone
https://code.googlesource.com/git/+/682c7d2f1a2d1a5443777237450505738af2ff1a

Change-Id: I87ed3c0f56c37e3491e367a41f5e555c4207ff44
Signed-off-by: Terry Parker <tparker@google.com>