This introduces ReftableBatchRefUpdate and ReftableDatabase, as
generic classes, with some code moved to DfsReftableBatchRefUpdate and
DfsReftableDatabase.
Clarify thread-safety requirements by asserting locked status in
accessors, and acquiring locks in callers. This does not fix threading
problems, because ReftableBatchRefUpdate already wraps the whole
transaction in a lock.
This also fixes a number of bugs in ReftableBatchRefUpdate:
* non-atomic updates should not bail on first failure
* isNameConflicting should also check for conflicts between names that
are added and removed in the BatchRefUpdate.
Change-Id: I5ec91173ea9a0aa19da444c8c0b2e0f4e8f88798
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
On changing a ref, the old SHA1 is not updated in the object => ref
mapping. This means search by object ID may still turn up a ref from
deeper within the stack. To fix this, check all refs produced by the
merged iterator against the merged reftables.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Change-Id: I41e9cd395b0608eedeeaead0a9fd997238d747c9
The flag enabling sideband-all is used in two places: in UploadPack
for advertisement and in the protocol parser to read it from the
request.
This leds to problems in distributed deployments where the two requests of
a fetch can go to different servers with different configurations.
Use the existing allowsidebandall to accept the sideband-all request
(and respond to it) and introduce a new "advertisesidebandall" to toggle
the advertising of the feature.
Change-Id: I892d541bc3f321606c89bad1d333b079dce6b5fa
Signed-off-by: Ivan Frade <ifrade@google.com>
If the SubModuleWalk isn't closed its TreeWalk's ObjectReader won't
be closed. Re-loading the DirCache during an IndexDiff is not only
inefficient but could also give strange results if an external
process had modified the index in the meantime: file diffs would
be based on a "before" state, but submodule diffs on an "after"
state.
Change-Id: Iab948c08ac342138b37263c9028d80b84101f6d6
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Add to statistics the amount and size of packfiles offloaded to HTTP
download.
Change-Id: I895a7219ecac2794368bfc4fdfae74c1238deed9
Signed-off-by: Ivan Frade <ifrade@google.com>
The object identifying packfiles to send them via packfile-uri contains
only the uri and the hash. This is the information that goes through the
wire. It would be useful to know also the size of those packfile, for
example to track how many bytes have been offloaded to HTTP.
Add size field the CachedPackUriProvider.PackInfo object.
Change-Id: If6b921b48a4764d936141c777879b148cc80bbd3
Signed-off-by: Ivan Frade <ifrade@google.com>
Exception handling can be isolated from UploadPack. This makes it
possible to make the exception handler pluggable.
Change-Id: Ieebbd6711963c7f2e47a98783b4ad815793721c7
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
This consolidates the sideband stream creation code and the error
handling code for the sideband-allowed part in the Git protocol to one
place.
Change-Id: I0e3e94564f50d1be32006f9d8bcd1ef1ce6bf07e
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
ErrorWriter writes an error message to the user. The implementation is
swapped once it detects that the client supports sideband. By default it
uses the protocol level ERR packet, which was introduced recently.
In total the error output is done in two different places;
UploadPack#upload and UploadPack#sendPack. These will be consolidated in
the next change.
Change-Id: Ia8d72e31170bbeafc8ffa8ddb92702196af8a587
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
Also correctly parse the "always" value (allowed in canonical git
since git 2.12.0[1]). Adapt the ReflogWriter.
[1] https://github.com/git/git/commit/341fb2862
Bug: 551664
Change-Id: I051c76ca355a2ac8d6092de65f44b18bf9aeb125
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
MergedReftable is not used as an AutoCloseable, because closing tables
is currently handled by DfsReftableStack#close.
Encode that a MergedReftable is a list of ReftableReaders. The previous
code suggested that we could form nested trees of MergedReftables,
which is not how we use reftables.
Change-Id: Icbe2fee8a5a12373f45fc5f97d8b1a2b14231c96
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
In the Config#StringReader we relied on ArrayIndexOutOfBoundsException
to detect the end of the input. Creation of exception with (deep) stack
trace can significantly degrade performance in case when we read
thousands of config files, like in the case when Gerrit reads all
external ids from the NoteDb.
Use the buf.length to detect the end of the input.
Change-Id: I12266f25751373a870ce3fa623cf2a95d882d521
Otherwise the paths modified by a cherry-pick with conflicts won't be
reported as modified via WorkingTreeModifiedEvents.
Change-Id: I875b67c0d2f68efdf90a9c32b80a2e074ed3570d
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
This makes the intended use of the classes more clear. It also
simplifies generic functions that write reftables: they only need a
ReftableWriter as argument, as the stream is carried within the
ReftableWriter.
Change-Id: Idbb06f89ae33100f0c0b562cc38e5b3b026d5181
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
The functionality in ReftableStack is specific to DFS.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Change-Id: If6003d104b1ecb0f3ca7e9c3815b233fa0abf077
This makes maxUpdateIndex() available in MergedReftable, so we can
know generically at which index to create the next reftable in a
stack.
Change-Id: Ia2314bc57c8b5dd7e69d5e61096fdce1d35abd11
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Older JGit stored only milliseconds timestamps in the index. Newer
JGit may get finer timestamps from the file system. This leads to
slow index diffs when a new JGit runs against an index produced
by older JGit because many timestamps will differ and JGit will
then do many content checks. See [1].
Handle this migration case by only comparing milliseconds if the
index entry has only millisecond precision.
The inverse may also occur; also compare only milliseconds if the
file timestamp has only millisecond precision.
Do the same also for microsecond resolution. On Windows, NTFS may
provide 100ns resolution and may be used by external programs writing
the index, but Java's WindowsFileAttributes may provide only
microseconds.
File timestamp precision in Java depends not only on the Java APIs
used by different JGit versions but may also change when running the
same Java code on different VMs. And of course the resolution may
vary among operating and file systems. Moreover, timestamp precision
in the index depends on the program that wrote the index. Canonical
git may use a different resolution, maybe even different between git
versions.
[1] https://www.eclipse.org/forums/index.php/t/1100344/
Change-Id: Idfd08606c883cb98787b2138f9baf0cc89a57b56
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
The removed code was trying to avoid mistakenly reporting differences
when core.autocrlf was set to "input" but a file had already been
committed with CR-LF. It did that by running the blob from the cache
through a CRLF-to-LF filter because older JGit would also run the file
from the working tree through such a filter.
The real fix for this case was done in commit 60cf85a. Since then files
are not normalized if they have already been committed with CR-LF and
this old fix attempt from bug 372834 is no longer needed.
Change-Id: Ib4facc153d81325cb48b4ee956a596b423f36241
Signed-off-by: Thomas Wolf <thomas.wolf@paranor.ch>
This removes one use of DFS specific code in this class.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
Change-Id: I3ef6a4b98357cc6dc480892244ddc51d2fd751a2
This lets us write reftables generically with functions that take
just ReftableWriter argument
Change-Id: I7285951f62f9bd4c78e8f0de194c077d51fa4e51
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
allRefs determined the end of the ref block without accounting for
index or log blocks. This could cause other blocks to be interpreted
as ref blocks, leading to "invalid block" error messages.
Change-Id: I7b9323e7d5e0e7d64535b3ec1efd576aed1e9870
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
If CheckStat is MINIMAL or timestamps have no nanosecond part
WorkingTreeIterator.compareMetaData only checks the second part of
timestamps and ignores nanoseconds which may have ended up in the index
by using native git.
If
fileLastModified.getEpochSecond() == cacheLastModified.getEpochSecond()
we currently proceed comparing fileLastModified and cacheLastModified
with full precision which is wrong since we determined that we detected
reduced timestamp resolution.
Fix this and also handle smudged index entries for CheckStat.MINIMAL.
Change-Id: I6149885903ac63d79b42d234cc02aa4e19578f3c
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
We don't need to update time atomically since it's only used to order
cache entries in LRU order.
Change-Id: I756fa6d90b180c519bf52925f134763744f2c1f1
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Since commit c3b0dec509fe136c5417422f31898b5a4e2d5e02, Git has
disallowed writing a non-commit to refs/heads/* refs. JGit still
allows that, which can put users in a bad situation. For example,
git push origin v1.0:master
pushes the tag object v1.0 to refs/heads/master, instead of the
intended commit object v1.0^{commit}.
Prevent that by validating that the target of the ref points to a
commit object when pushing to refs/heads/*.
Git performs the same check at a lower level (in the RefDatabase). We
could do the same here, but for now let's start conservatively by
handling it in pushes first.
[jn: fleshed out commit message]
Change-Id: I8f98ae6d8acbcd5ef7553ec732bc096cb6eb7c4e
Signed-off-by: Yunjie Li <yunjieli@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
Han-Wen Nienhuys
Ivan Frade
Thomas Wolf
David Pursehouse
Matthias Sohn
Masaya Suzuki
Minh Thai
Saša Živkov
Yunjie Li