Include the net.i2p.crypto.eddsa bundle via a hard dependency.
Add tests for dealing with ed25519 host keys and user key files.
Manual tests: fetching from git.eclipse.org with an ed25519 user key,
and pushing this change itself using the same ed25519 key.
Note that sshd 2.0.0 does not yet support encrypted ed25519 private
keys.
Bug: 541272
Change-Id: I7072f4014d9eca755b4a2412e19c086235e5eae9
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Set GIT_DIR and GIT_WORK_TREE when calling hooks.
Bug: 541622
Change-Id: I6153d8a6a934ec37a3a5e7319c2d0e516f539ab7
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
The new dependency to mockito for JDKHttpConnectionTest wasn't added for
bazel test run.
Change-Id: Ie967b7152340a0b3e05cced7eadfd40af78d9995
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
After cloning a repo with a submodule, non-recursively, JGit would
encounter in its TreeWalk in IndexDiff:
* first, a missing gitlink (in index & HEAD, not in working tree)
* second, the untracked folder (not in index and head, in working tree)
As a result, it would report the submodule as missing. Canonical git
reports a clean workspace.
The root cause of this is that the path of a gitlink "x" did
not compare equal to the path of a tree "x" in JGit.
Correct Paths.compare() to account for that. If two paths are otherwise
equal, then let gitlinks match both trees and files. Matching trees
solves the bug. Matching files is necessary to handle the case where
the gitlink directory was replaced by a file; see the new test case
IndexDiffSubmoduleTest.testSubmoduleReplacedByFile(). Comparisons of
unequal paths are left untouched, so the sort order is unchanged.
After the fix, another bug(?) in WorkingTreeIterator became apparent:
with core.dirNoGitLinks = true, it was no longer possible to overwrite
a gitlink in the index. This is now fixed in WorkingTreeIterator.
Add new test cases for the bug itself and for some related cases
(submodule directory deleted or replaced by a file) in
IndexDiffSubmoduleTest. Add a test for missing files in IndexDiffTest,
and adapt the PathsTest to test matching gitlinks.
Bug: 467631
Change-Id: I0549d10d46b1858e5eec3cc15095aa9f1d5f5280
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
when multiple match options are given in git describe the result must
not depend on the order of the match options. JGit wrongly picked the
first match using the match options in the order they were defined. Fix
this by concatenating the streams of matching tags for all match options
and then choosing the first match on the concatenated stream sorted in
tie break order.
See https://git-scm.com/docs/git-describe#git-describe---matchltpatterngt
Change-Id: Id01433d35fa16fb4c30526605bee041ac1d954b2
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
According to RFC 2616 [1] header field names are case insensitive.
Header fields defined as a comma separated list can have multiple header
fields with the same field name. Add a method to HttpConnection which
retrieves all values with a given header field name with the field name
compared case insensitive.
[1] https://tools.ietf.org/html/rfc2616#section-4.2"
Change-Id: I7f601b21cda99e84f43f866c7c7cb4cb0e3cf5c3
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
This partially reverts commit a551b646: revert the changes in
RawParseUtils.lineMap(). Forcing all blobs containing a NUL byte
as a single line causes blame to produce useless results as soon
as it hits any version containing a NUL byte.
Doing binary detection at this level also has the problem that the
user cannot control it. Not by setting the text attribute nor in any
other way.
This came up in bug 541036, where a Java source inadvertently
contained NUL bytes in strings. Even fixing this by using escapes
"\000" will not fix JGit's blame for this file because the past
versions will still contain the NUL byte.
Native git can blame that file from bug 541036 fine.
Added new tests verifying that blaming a text file containing a NUL
byte produces sensible results.
Bug: 541036
Change-Id: I8991bec88e9827cc096868c6026ea1890b6d0d32
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Add tests for password and keyboard-interactive authentication.
Implement password authentication; the default provided by sshd
is non-interactive, which is not useful for JGit.
Make sure the CredentialsProvider gets reset on successive password
retrieval attempts. Otherwise it might always return the same non-
accepted password from a secure storage. (That one was discovered
by actually trying this via EGit; the JGit tests don't catch this.)
Change the default order of authentication mechanisms to prefer
password over keyboard-interactive. This is a mitigation for upstream
bug SSHD-866.[1]
Also include a fix for upstream bug SSHD-867.[2]
[1] https://issues.apache.org/jira/projects/SSHD/issues/SSHD-866
[2] https://issues.apache.org/jira/projects/SSHD/issues/SSHD-867
Bug: 520927
Change-Id: I423e548f06d3b51531016cf08938c8bd7acaa2a9
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Create the bundle and move the SshTestGitServer there. Verified that
the Eclipse build still works and ran JSchSshTest and ApacheSshTest as
junit tests inside Eclipse.
Update maven build and features to account for that. Verified by
running full maven build including packaging.
Update bazel build files to account for that. Verified by a
clean-slate bazel build :all, followed by running the JSchSshTest
and the ApacheSshTest via bazel.
Change-Id: Ia084942f4425b454529de148e00417e7da786a90
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
sshd only asks exactly once for the password. C.f. upstream issue
SSHD-850.[1] So we have to work around this limitation for now.
Once we move to sshd > 2.1.0, this can be simplified somewhat.
[1] https://issues.apache.org/jira/browse/SSHD-850
Bug: 520927
Change-Id: Id65650228486c5ed30affa9c62eac982e01ae207
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
By default sshd will use its default built-in list, which matches
the one of openssh (as far as the algorithms exist in sshd at all).
But it doesn't handle HostKeyAlgorithms from the ssh config at all.
Implement this as in openssh, including the '+' and '-' modifiers
and reordering the default if there are known host keys for a
server already.
Add tests for the reordering.
Also use a more robust reader for the known hosts file. The default
aborts on the first error.
Bug: 520927
Change-Id: Ib1684440bfe2e96140536aa1a93c4bd4a0d35916
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Tests that it works with unknown authentications in the list, and
fails if there are no common authentications between server and
client. The latter also tests that the ssh config setting is taken
into account at all.
And promptly the JGit sshd client didn't. Add a fix for this. It's
a tiny bit hacky: Apache MINA looks up a custom property set on a
hierarchy of "PropertyResolver"s starting with the session. On the
session itself this property can never be set since it's read
already in the session constructor before anyone had any chance
to set it. The next element in the resolver hierarchy is the sshd
SshClient, and so we set that property there. Since we use one
SshClient and one ClientSession per JGit SshdSession, this is OK.
Bug: 520927
Change-Id: I62446fc1fffde125a8965c030240f0918ae234b7
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Add a new ssh client implementation based on Apach MINA sshd 2.0.0.
This implementation uses JGit's own config file parser and host entry
resolver. Code inspection of the Apache MINA implementation revealed
a few bugs or idiosyncrasies that immediately would re-introduce bugs
already fixed in the past in JGit.
Apache MINA sshd is not without quirks either, and I had to configure
and override more than I had expected. But at least it was all doable
in clean ways.
Apache MINA boasts support for Bouncy Castle, so in theory this should
open the way to using more ssh key algorithms, such as ed25519.
The implementation is in a separate bundle and is still not used in
the core org.eclipse.jgit bundle. The tests re-use the ssh tests from
the core test bundle.
Bug: 520927
Change-Id: Ib35e73c35799140fe050d1ff4fb18d0d3596580e
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Add support for git-receive-pack to the ssh git server and add two
new tests for pushing.
This actually uncovered an undocumented requirement in TransportSftp:
the FTP rename operation assumes POSIX semantics, i.e., that the
target is removed. This works as written only for servers that
support and advertise the "posix-rename@openssh.com" FTP extension.
Our little Apache MINA server does not advertise this extension.
Fix the FtpChannel implementation for Jsch to handle this case in a
meaningful way so that it can pass the new "push over sftp" test.
Add more tests to test the behavior of server host key checking.
Also refactor the tests generally to separate better the test
framework from the actual tests.
Bug: 520927
Change-Id: Ia4bb85e17ddacde7b36ee8c2d5d454bbfa66dfc3
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Correct behaviour as git 1.7.1.1 is to resolve tie-breakers to choose
the most recent tag.
https://github.com/git/git/blob/master/Documentation/RelNotes/1.7.1.1.txt:
* "git describe" did not tie-break tags that point at the same commit
correctly; newer ones are preferred by paying attention to the
tagger date now.
Bug: 538610
Change-Id: Ib0b2a301997bb7f75935baf7005473f4de952a64
Signed-off-by: Håvard Wall <haavardw@gmail.com>
Suppose that a repository has the following commit graph:
B C
\ /
A
and it was cloned with --shallow-exclude=A. DepthGenerator does not mark
C as shallow, causing an invalid repository to be produced on the
client, because A is not sent. (A similar issue occurs when
--shallow-since is used to exclude A but neither B nor C.)
This happens whenever an excluded commit has more than one child that is
to be sent to the client. Fix DepthGenerator to handle this case
correctly.
While we're editing DepthWalk.Commit, fix the documentation of
DepthWalk.Commit#isBoundary.
Change-Id: I7068abf0fe0c864d1b0e56e1616dad1aa8719411
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Move the bulk of the basic parsing and host entry handling into a
new class OpenSshConfigFile that has no dependencies on any concrete
ssh implementation. Make the existing OpenSshConfig use the new
parser.
Introduce a new class SshConstants collecting all the various ssh-
related string literals. Also use TreeMaps with a case-insensitive
key comparator instead of converting keys to uppercase. Add a test
to verify that keys are matched case-insensitively.
Most of the parsing code was simply moved, except that the new
parser supports looking up entries given host name, port, and user
name, and can thus handle more %-substitutions correctly. This
feature is not yet used and cannot be used with JSch since JSch
only has a ConfigRepository.getConfig(String) interface.
The split is still worth the trouble as it opens the way to using
another ssh client altogether. Apache MINA sshd, for instance,
resolves host entries giving host name, port, and user name.
(Apache MINA has a built-in ssh config handling, but that has
problems, too: its pattern matching is case-insensitive, and its
merging of host entries if several match is not the same as in
OpenSsh. But with this refactoring, it will be possible to plug in
OpenSshConfigFile into an Apache MINA sshd client without dragging
along JSch.)
One test case that doesn't make sense anymore has been removed. It
tested that repeatedly querying for a host entry returned the same
object. That is no longer true since the caching has been moved to
a deeper level.
Bug: 520927
Change-Id: I6381d52b29099595e6eaf8b05c786aeeaefbf9cc
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Add a simple ssh git server based on Apache MINA sshd, and use it
in new tests that verify ssh operations and in particular a number
of bugs that had cropped up over time in JSch.
The git server supports fetching only, and sftp access.
The tests are all in an abstract base class; the concrete JschSshTest
class only provides ssh-specific test setup. So the same tests could
be run easily also with some other ssh client.
Bug: 520927
Change-Id: Ide6687b717fb497a29fc83f22b07390a26dfce1d
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
When a source folder is marked as a test folder, JDT requires that it
has an output folder different from the one used for regular sources.
Therefore give the test folders in org.eclipse.jgit.test a separate
output folder "bin-tst".
Moreover JDT reports errors if non-test classes have dependencies on
test classes. Therefore remove the "test" annotation from
org.eclipse.jgit.junit.
Change-Id: Ib527439ff5b7d7b570b8a60819ecaa70f59c63a3
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
The copyfile entry in the manifest file copies the contents of the file
but doesn't keep the executable flag. This is inconsistent with repo
tool behaviour, plus is natural to expect that the copy of a executable
file is executable.
Transfer the executable bit when copying the file, aligning the
RepoCommand with repo tool and user expectations.
Change-Id: I01b24f482d5939e01d496f032388b3a5c02a912a
Signed-off-by: Ivan Frade <ifrade@google.com>
The RepoCommand.RemoteReader interface doesn't offer access to the mode
of a file. Caller can only default to mark the copied objects as regular
files, losing e.g. the executable bit (if set).
Add a new method readFileWithMode that returns the contents and mode of
the remote file. It supersedes the readFile method, that is marked as
deprecated.
Now callers can set correctly the file mode of the copied file.
Change-Id: I8fce01e4bc5707434c0cbc4aebbae1b6b64756f0
Signed-off-by: Ivan Frade <ifrade@google.com>
Many tests verify the contents of files in a try-with-resources
incantation that clutters the code.
Extract that verification to an "assertContents" method, that is easier
to read.
Change-Id: If430eac6f5b9ae352e42b2d43867ceb6cd618fbb
Signed-off-by: Ivan Frade <ifrade@google.com>
Patterns should treat \r in file names as normal characters
Change-Id: Ica3e0fa4a58acf5326db46bb28571fe5f20f6cd2
Signed-off-by: Marc Strapetz <marc.strapetz@syntevo.com>
ProtocolV2Parser has unit tests but protocol v0/v1 is not covered.
Change-Id: I96022e8f8eb60d4da748d1042474fd1efd67e882
Signed-off-by: Ivan Frade <ifrade@google.com>
Parsed requests represent object ids (SHA1) in ObjectId instances but tests
use strings for those ids because they are easier to define.
Create a custom matcher that hides the conversion from string to
ObjectId. Note that this reverses the existing code conversion (it was
transforming ObjectIds into string).
This produces more readable code, consistent with the other hamcrest
assertions.
Change-Id: I47ba1d25557d791fe74fb93c740ff7de9923cc00
Signed-off-by: Ivan Frade <ifrade@google.com>
This allows clients to use the --shallow-exclude parameter (producing a
"deepen-not <ref>" line when communicating with the server) in their fetch
commands when fetching against a JGit server using protocol v2.
Note that the implementation in this commit is somewhat inefficient, as
described in the TODO comment in DepthGenerator.
Change-Id: I9fad3ed9276b624d8f668356ffd99a067dc67ef7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
In C Git, when a client fetches with "git fetch --shallow-since=<date>
origin <ref>", and all commits reachable from <ref> are older than
<date>, the server dies with a message "no commits selected for shallow
requests". That is, (1) the --shallow-since filter applies to the commit
pointed to by the ref itself, and (2) there is a check that at least one
commit is not filtered out. (The pack-protocol.txt documentation does
not describe this, but the C implementation does this.)
The implementation in commit 1bb430dc21 ("UploadPack: support
deepen-since in protocol v2", 2018-09-27) does neither (1) nor (2), so
do both of these.
Change-Id: I9946327a71627626ecce34ca2d017d2add8867fc
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
The request receives a list of capabilities and takes out the "agent" to
offer it on its own setter (getAgent).
Do this at parse time: when reading the line if the capability is
"agent" set it directly in the builder.
This makes the treatment of "agent" consistent in v0/v1 and v2.
Change-Id: Ie4f9f2cad8639adeeaef4921df49a30a8ce5b42f
Signed-off-by: Ivan Frade <ifrade@google.com>
All of the input lines passed to pre-push hook scripts must be properly
terminated by '\n', so that normal shell scripts like the git-supplied
pre-push.sample work properly, even when pushing just a single branch.
With the old code, hook scripts that use the following pattern didn't
process the last line, because 'read' has a non-zero exit status when
EOF is encountered:
while read local_ref local_sha remote_ref remote_sha; do ... done
Change-Id: Id899662ed3fedef6c314fc4b2ddf91a6dcb98cbb
Signed-off-by: Markus Keller <markus.kell.r@gmail.com>
Before this commit, a force checkout would fail if there
were any conflicting files. After this commit, a force
checkout will overwrite the conflicting files, as expected.
Making this work required fixing a bug in DirCacheCheckout.
Before this commit, when DirCacheCheckout had
failOnConflict=false, it would delete all conflicting files
from the working copy and just leave them missing. After
this commit, DirCacheCheckout overwrites conflicting files
with the merge tree.
This change in DirCacheCheckout causes "reset --hard" and
"revert --abort" to behave as expected (previously they
would simply delete conflicting files, now they will be
overwritten from the merge tree).
Change-Id: If7e328ee792ef6511ab7d9c26d8d77c39210ec9f
Signed-off-by: Ned Twigg <ned.twigg@diffplug.com>
UploadPack.getPeerUserAgent() doesn't produce the expected results for
protocol v2 requests. In v2, the agent reported in the request (in an
"agent=" line) is not in the clientCapabilities but in a field on its
own. This makes getPeerUserAgent default to the transport user agent.
Making "agent" a shared property between protocol v0/v1 and v2 fixes the
problem, simplifies the function and harmonizes the implementation
between protocol versions.
In a follow up commit the "agent" will be identified on parsing time,
instead of taking it from the client capabilities.
Change-Id: Idf9825ec4e0b81a1458c8e3701f3e28aafd8a32a
Signed-off-by: Ivan Frade <ifrade@google.com>
In protocol v2, a command request can be followed by server options
(lines like "agent=<>" and "server-option=<>"), but current code
doesn't accept those lines.
Advertise the "server-option" capability, parse the lines and add
them to the request objects.
Other code in JGit can see this options and act accordingly via the
protocol v2 hooks.
This should not require any change in the client side.
Change-Id: If3946390f9cc02d29644b6ca52534b6f757bda9f
Signed-off-by: Ivan Frade <ifrade@google.com>
Matthias Sohn
Thomas Wolf
Gunnar Wagenknecht
David Pursehouse
Håvard Wall
Jonathan Tan
Ivan Frade
Marc Strapetz
Markus Keller
Ned Twigg