Kevin Sawicki [Mon, 9 Jan 2012 16:46:13 +0000 (08:46 -0800)]
Provide helper for unlocking a file
This will allow recovery from a LockFailedException where
the file associated with an exception is passed to FileUtils.unlock
to attempt an unlock on the file so the operation can be retried
Matthias Sohn [Sun, 8 Jan 2012 21:08:08 +0000 (13:08 -0800)]
Narrow exceptions declared by LsRemoteCommand
API commands either throw GitAPIException or JGitInternalException.
Also add missing javadoc and reduce nesting of catch blocks.
Change-Id: I9a3b302e1b3f373ee11a977a0e3d6213bfbd3cdf Signed-off-by: Matthias Sohn <matthias.sohn@sap.com> Signed-off-by: Kevin Sawicki <kevin@github.com>
Kevin Sawicki [Tue, 3 Jan 2012 23:07:02 +0000 (15:07 -0800)]
Retain executable mode of existing files on Windows
Currently files in a repository marked as executable will have
that mode unset when modified and committed on systems that
do not support detection of this mode since the working tree
iterator will never report this mode for any entries.
This change updates WorkingTreeIterator to be able
to determine the target file mode to be used for the index
through consideration of the configured WorkingTreeOptions.
The places where ResolveMerger was doing content merges have been
refactored. The goal was to have one single method where content merge
was done and to factor out other topics (updating the index, updating
the working tree) into own methods. This was done to allow adding
pluggable content mergers in change
I7817e2123d254f3eeb315b47a61d2c55bd202c12
Change-Id: I8529697b197372a284bcd5ab2c9ba1adb925a520 Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Kevin Sawicki [Tue, 13 Dec 2011 01:49:09 +0000 (17:49 -0800)]
Add exception class for when locking a file fails
This will allows calling classes to handle lock failures
without checking against the message and will also provide
access to the file that could not be locked.
Kevin Sawicki [Sat, 17 Dec 2011 01:05:42 +0000 (17:05 -0800)]
Support resolving integer-based reflog revisions
Revision strings such as 'master@{0}' can now be resolved
by Repository.resolve by reading the reflog for the ref and
returning the commit for the entry number specified.
This still throws an exception for cases not supported
such as 'master@{yesterday}'.
Matthias Sohn [Fri, 16 Dec 2011 07:49:59 +0000 (08:49 +0100)]
Merge branch 'stable-1.2'
* stable-1.2:
Add API checking using clirr
Fix MergeCommandTest to pass if File.executable is not supported
Fix ResolveMerger not to add paths with FileMode 0
Change-Id: I86e7194a40acd6dfa3d433f1d17c01bdf5bb0d9c Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Robin Rosenberg [Fri, 16 Dec 2011 00:01:36 +0000 (01:01 +0100)]
Enforce the use of Java5 API:s only (with a few exceptions)
This only works with Eclipse 3.6 and newer and requires installation
of new package. Documentation is not very good, but there is a blog
about it here:
http://eclipseandjazz.blogspot.com/2011/10/of-invalid-references-to-system.html
API checking is especially useful on OS X where Java5 is not readily
available.
Change-Id: I3c0ad460874a21c073f5ac047146cbf5d31992b4 Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Fix ResolveMerger not to add paths with FileMode 0
When ResolveMerger finds a path where it has to do a content merge it
will try the content merge and if that succeeds it'll add the newly
produced content to the index. For the FileMode of this new index entry
it blindly copies the FileMode it finds for that path in the common base
tree. If by chance the common base tree does not contain this path it'll
try to add FileMode 0 (MISSING) to the index.
One could argue that this can't happen: how can the ResolveMerger
successfully (with no conflicts) merge two contents if there is no
common base? This was due to another bug in ResolveMerger. It failed to
find out that for two files which differ only in the FileMode (e.g. 644
vs. 755) it should not try a content merge.
Change-Id: I7a00fe1a6c610679be475cab8a3f8aa4c08811a1 Signed-off-by: Christian Halstrick <christian.halstrick@sap.com> Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com>
Matthias Sohn [Sat, 10 Dec 2011 00:06:47 +0000 (01:06 +0100)]
Merge branch 'stable-1.2'
* stable-1.2:
Fix version.sh
Throw API exception when MergeCommand hits checkout conflicts
Add methods for configuring platform emulation
Fix history rendering not to occupy too many lanes
Fix History rendering
Change-Id: If71cc760423ae2b76c7435ca4830edc1745556de Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Matthias Sohn [Thu, 8 Dec 2011 22:33:53 +0000 (23:33 +0100)]
Throw API exception when MergeCommand hits checkout conflicts
When MergeCommand hit checkout conflicts it did throw an internal JGit
exception org.eclipse.jgit.errors.CheckoutConflictException instead of
org.eclipse.jgit.api.errors.CheckoutConflictException which it
declares to throw. Hence translate the internal exception to the
exception declared in the API.
Bug: 327573
Change-Id: I1efcd93a43ecbf4a40583e0fc9d8d53cffc98cae Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Kevin Sawicki [Tue, 6 Dec 2011 18:00:28 +0000 (10:00 -0800)]
Throw invalid ref exception instead of invalid remote exception
ReflogCommand command was throwing an incorrect exception type
when an IOException was wrapped and rethrown from the underlying
ReflogReader. The IOException cause is now provided to the thrown
exception as well.
Robin Rosenberg [Tue, 6 Dec 2011 23:42:35 +0000 (00:42 +0100)]
Force jgit.sh to use UTF-8 encoding on Mac
For some obscure reason the platform encodng in Java on Macs
is MacRoman. OS X used UTF-8 as the default encoding so set
the encoding property in jgit.sh.
Fix history rendering not to occupy too many lanes
There was a bug in history rendering which caused jgit to use too
many lanes in case lanes get repositioned. Looking at commit 90c11cbaeb83ee9b02238cbd2c0e5bcf68068772 in JGit was one example. Vadim
Dmitriev found the problem and the solution.
There was the possibility that during history rendering we draw a lane
"trough" a passed commit. Vadim Dmitriev found that out in bug 335818.
I added the needed check to that block of code where it was missing.
Bug: 335818
Change-Id: Ic944193b2aca55ff3eb0235d46afa60b7896aa0f Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Robin Rosenberg [Tue, 29 Nov 2011 23:16:40 +0000 (00:16 +0100)]
Add percent-encoding of reserved characters in URIish
We do this for the the names that have an explicit scheme and
do it both ways. The URIish is parsed before decoding. Only
a few special characters are encoded for the path part of the
URI, i.e. space, non-ASCII and control characters. The percent
encoding is assumed to be a stream encoding so we interpret it
as UTF-8.
There was the possibility that during history rendering we draw a lane
"trough" a passed commit. Vadim Dmitriev found that out in bug 335818.
I added the needed check to that block of code where it was missing.
Bug: 335818
Change-Id: Ic944193b2aca55ff3eb0235d46afa60b7896aa0f Signed-off-by: Christian Halstrick <christian.halstrick@sap.com>
Shawn O. Pearce [Thu, 1 Dec 2011 01:36:32 +0000 (17:36 -0800)]
Discard request HTTP bodies for status code <400
The HTTP RFCs require a server to fully consume the request body before
it can return a non-error status code, which is any code below 400.
JGit returns most Git level errors inside of an HTTP 200 OK response,
and sometimes this happens before the entire request was consumed from
the servlet container. In such cases the body must be skipped or read
until EOF is reached, ensuring the HTTP keep-alive semantics will work
for the next request on the same TCP connection.
HTTP status codes >= 400 may be returned without consuming the body,
and a servlet container must set "Connection: close" in the response
headers when this happens, since the state of the request body is not
well defined with an early abort.
With the introduction of sendError() in GitSmartHttpTools there are
only a handful of locations that need to worry about the request body
being consumed, so sprinkle the call in as necessary.
Shawn O. Pearce [Thu, 1 Dec 2011 01:28:54 +0000 (17:28 -0800)]
Ensure all smart HTTP errors are sent to clients
Error messages are typically short, below the 32 KiB in-memory buffer
size of the SmartOutputStream. When an error is queued up for sending
to a client and an exception is thrown up into the servlet handler we
discarded the message and sent nothing to the client, as the messages
were stuck inside of the SmartOutputStream buffer.
Hoist the creation of the output stream above the invocation of try
block of the service, and use close() in the few catch blocks that
assume there are buffered messages ready for transmission. This will
ensure errors from unpacking a stream in ReceivePack are sent off to
a client correctly, as previously these were causing no status report
to arrive at the client side as the data was stuck in the buffer.
Shawn O. Pearce [Thu, 1 Dec 2011 01:17:59 +0000 (17:17 -0800)]
Always close the GZIPOutputStream to release Deflater
The stream must be closed to ensure the native resources associated
with its internal Deflater instance are cleaned up early, instead of
waiting for GC to identify the dead object and finialize it.
Shawn O. Pearce [Thu, 1 Dec 2011 01:40:10 +0000 (17:40 -0800)]
Fix HTTP unit tests
I modified the way errors are returned, and this particular test is
now getting a different access denied response. The new text happens
to be what I intended to have here, so update the test.
Matthias Sohn [Sun, 27 Nov 2011 00:16:49 +0000 (01:16 +0100)]
Reset SSH connection and credentials on "Auth fail"
When SSH user/password authentication failed this may have been caused
by changed credentials on the server side. When the SSH credentials of a
user change the SSH connection needs to be re-established and
credentials which may have been stored by the credentials provider
need to be reset in order to enable prompting for the new credentials.
Bug: 356233
Change-Id: I7d64c5f39b68a9687c858bb68a961616eabbc751 Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
projects / jgit.git / log