| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Change-Id: I87b951065a8331e635f042ece7266c9e859516be
|
|
|
|
| |
Change-Id: I177ccfd2a3db387822085602dfc66267703c0a48
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As part of the update to patch_set_approavals, the category names
are changed. Doing so causes "ON UPDATE CURRENT_TIMESTAMP" in at
least MySQL to replace all of the existing timestamps with the
time of the migration. Since the purpose of CURRENT_TIMESTAMP
here is to track when the approval vote happened, this is a loss
of data.
Amend the data migration to include granted=granted in the update
list so that anyone running the migration will not lose their
data.
Change-Id: If80891018eae0fab75770a5aea5a382200e9112f
|
|
|
|
|
|
|
|
|
| |
This patch adds gerrit.war into api_{install,deploy} buck build step.
This allows third party maven based builds depend on gerrit.war
artifact.
Change-Id: I22c8d3339a1647913967e423e45462c2c5c255de
Signed-off-by: Dariusz Luksza <dariusz@luksza.org>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Since sshd 0.10.0 the handling of global requests
is refactored out of [1] ServerConnectionService.globalRequest
method to AbstractConnectionService.globalRequest.
The implementation has been refactored into a list
of RequestHandlers, one for each type of request.
These handlers are never set in SshDaemon constructor
as they should be.
See SshServer.getUpDefaultServer line 503
Change-Id: I7a38145c021538838d11427950db66eb57a50582
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* changes:
Use Provider for IdentifiedUser in CreateBranch constructor
Split PGPEncryptedDataGenerator creation out into a utility method
ChangeScreen2: Only reset the commit message text on cancel
Fix log spew caused by DeleteBranch constructor
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This is the same fix that was already done for DeleteBranch's
constructor in If7fd7f4c38bcbc685d4f9451479da7a9ed432564.
Change-Id: Ied38df303eaf4ea1893527e947633445974e9d19
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Split the creation of PGPEncryptedDataGenerator out into a utility
method. Inline the call to the open() method on the returned generator
rather than storing the generator in a temporary variable.
Suppress deprecation warnings caused by using methods that have been
deprecated in the new version of Bouncycastle that was introduced
recently. We cannot yet replace the calls with the newer versions
because some Gerrit sites still use the older version of Bouncycastle.
Change-Id: I98f4b7c7bf8ab1ee74eb8bbc86b673d057d98e46
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Allow the user to begin editing the commit message, dismiss the
box by clicking outside of it (e.g. to copy part of a file name
from the Files table), and then re-open the current draft text
without resetting the box.
Only reset the box when the user explicitly clicks Cancel. Track
this by using the empty string in the text editor to mean there
is no valid message and the current original should be used.
Change-Id: Icd99b92c50ff92e0b1dc31a78fc57612f57a143b
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Every time branches are accessed the server logs a nasty Guice
trace preventing DeleteBranch from being created. The view must
create without enforcing the IdentifiedUser to exist. The solution
throughout the rest of Gerrit REST API code is to use Provider<X>
instead of X to allow Guice to defer this resolution.
Change-Id: If7fd7f4c38bcbc685d4f9451479da7a9ed432564
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Do not allow users to replace a more recent patch set with an older
patch set when there is a race between the web UI and the command
line git client:
1. Upload a new change (Patch Set 1)
2. Open the code review page in a browser
3. Upload a second version of the change (Patch Set 2)
4. Without refreshing the code review page, so that it's still
showing Patch Set 1, click the Edit Message button, change the
commit message, and hit save.
5. Note that this has created a Patch Set 3 with file contents taken
from Patch Set 1 but the edited message.
Change-Id: I73af89989b72d9e702061b3746390df68cef4b62
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
937d527ac0e4ee685fccb60ae604406a6638daec upgraded bouncy castle to 1.49.
The old JAR artifact was renamed and split to number of files. For Gerrit
two libraries must be downloaded and be available under `$gerrit_site/lib`:
* bcprov-jdk15on-149.jar
* bcpkix-jdk15on-149.jar
Update download logic to support a basic concept of required dependencies,
ensuring BC provider is downloaded if the SSL library is installed by init.
Change-Id: I60092ebe136f78a5649fe8512737275f50195a34
|
|/
|
|
|
|
|
|
|
|
| |
Many bugs were fixed since version 1.44 that we are using[1].
JARs artifacts were renamed.
[1] http://www.bouncycastle.org/releasenotes.html
Change-Id: Ie2b22328c77b0100dfff61139b737894b88e7664
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Intraline edit information can include a negative delta on the B
side of an edit, for example:
{
a: [...],
b: [...],
edit_a: [[4, 11], [14, 12]]
edit_b: [[4, 66], [37, 41], [-2, 1]]
}
In the edit_a or edit_b block the list pairs are the number of
characters to advance since the last position. -2 encodes the edit
iterator has to go back 2 positions to return the correct value.
This situation arose from an intraline edit list of:
REPLACE( 4-15, 4- 70),
REPLACE(29-41,107-148),
INSERT (62-62,146-147)
The EditList is sorted by the beginA values, where 29 < 62. This
forced the final INSERT to sort in the edit_b list after the REPLACE
at 107. Since 146 < 148 the final edit_b entry was encoded with a
negative delta (148 - 146 = -2).
Update the UI code to accept a negative delta in the iterator. There
are already many /diff responses with negative delta that will be
cached by browsers for 7 days.
There may be other issues with the intraline diff algorithm that
allows it to create this final INSERT span within the REPLACE
span, but the code is essentially unchanged for many years in
Gerrit Code Review. Handle the corner case for now in the UI while
the server code is investigated further.
Change-Id: Ibec0fc845356af7c02f25acac23c4e709f7c3a34
|
|\
| |
| |
| |
| |
| | |
* changes:
Update sshd to 0.11.0
Bump SSHD version to 0.10.1 and enable nio2 backend
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Fixes a few issues including a null pointer exception [1] that causes
ssh commands to hang.
It also adds server driven rekeying after 1G bytes of data transferred,
which is a useful security feature.
[1] https://issues.apache.org/jira/browse/SSHD-307
Change-Id: I441995eaf76e4fe53f7991f5aef00e062900791b
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
46f8488a78742799a4f5f781279c8955460e7964 removed nio2 backend because of
a bug [1]. Given that the bug is fixed and released, enable it again.
Another problem was also fixed and released [2].
Upgrade to the official version 0.10.1 makes Gerrit's own custom patch of
SSHD and hosting it on Google bucket unnecessary.
[1] https://issues.apache.org/jira/browse/SSHD-252
[2] https://issues.apache.org/jira/browse/SSHD-255
Bug: Issue 2406
Change-Id: I6668f6194427f153e1982e14b3936d2f3132fa7d
|
| |
| |
| |
| |
| |
| |
| | |
Making it more consistent with the tooltips on the other
buttons.
Change-Id: I3f26b08866452eb666da2a23be7201528b5f81a7
|
|/
|
|
|
|
|
|
|
|
|
|
|
| |
The reason is the ReplaceRequest.validate() should compare
to 'newCommit' of ReplaceRequest instance, else when
ReceiveCommits.autoCloseChanges() working on the second one
of a series of commits, the 'commit already exists(in the project)'
will be thrown wrongly.
Fixed.
Bug: issue 2607
Change-Id: Ib4db351d39f85d57ccf532da4d6422bdcbb88105
|
|
|
|
|
|
|
|
|
|
|
|
| |
When IOException is thrown in insertPatchSet() or
insertChange(), e.g. ClosedByInterruptException thrown
from AutoCommitWriter.manualFlush(), it will be
mapped to OrmException with error message
"Error updating database" by ORM_EXCEPTION, this becomes
misleading when the administrator analyzes the error log.
Fixed. Replaced the ORM_EXCEPTION with INSERT_EXCEPTION.
Change-Id: I49a6dfea7e49df72db775989a3a0501f60db76d2
|
|
|
|
|
| |
Change-Id: I8047fa702c12150232ebf62bcd16a27eb638c5e8
Signed-off-by: Edwin Kempin <edwin.kempin@sap.com>
|
|
|
|
|
| |
Change-Id: I109bced2f9309dd20a7962d5fdbcd999fa6f87ff
(cherry picked from commit a2bf824fa154b7cb51e8f89df8321d02c25f3425)
|
|
|
|
|
| |
Bug: issue 2569
Change-Id: Ic192daa21cfcc1f26fa9f00feb14f41013a29fb9
|
|
|
|
|
|
|
|
| |
Group list screen now uses pagination similar to project list screen.
The number of groups displayed is determined by the 'Maximum Page
Size' in user preferences.
Change-Id: Id4649fda4b06ef62bf9d630460aeefc1c5f65989
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
List groups endpoint now has an option to limit the number of results
returned and a start to define nonzero offset in order to support
groups pagination.
Add and document 'n' and 'S' function to allow query group list from
a specified limit and start.
Both options will then be used to render group list with pagination
in web UI.
Change-Id: Ibf244b1cfe83c56dc1c23aabf70a55697b52c835
|
|
|
|
|
|
|
|
|
|
| |
Every time key up event is raised in the group list filter box, the
group list was getting refreshed even if the filter did not change
(e.g. moving the cursor was refreshing the list).
Only call the refresh if the filter is changed or enter is pressed.
Change-Id: I64d24eeae08f35723b401fffd9a3e8ed84c8d00c
|
|
|
|
|
|
|
| |
Currently the button is enabled for all open changes, but if
the patch set is a draft, pressing it results in an error.
Change-Id: I382fc9330b73d357c00ba0209f18d2c1584e9256
|
|
|
|
|
|
|
|
|
|
|
| |
When a draft change is abandoned it is currently published to all
users by setting the status to ABANDONED. Restoring the change will
effectively publish the change as /restore sets the status to NEW.
To not complicate the workflow, prevent draft changes from being
abandoned.
Change-Id: If747264fa66a2139e0d8f13987125d853ec3bed9
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The internal waiting generation counter in
ControlledRealTimeReopenThread is only increased when
waitForGeneration is called. This controls whether the min or max
refresh time is used for waiting. Because NrtFuture.get is not called
directly and listeners may be used instead, this was not guaranteed
to happen, leading to long delays. Fix this by doing a dummy wait for
0ms at future construction time.
Change the order of refresh listener initialization so any
removeIfDone calls are made only after the internal listener within
ControlledRealTimeReopenThread updates its searching generation.
Change-Id: I0edf4a3d38a160496afaf0587cf14b6fee2f2aa5
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Because WorkQueue.Module and LuceneIndexModule are in the same
Injector, the ordering of their LifecycleListeners' stop() methods is
not defined. Tasks on the queue, e.g. background mergeability checks,
might try to perform index writes after the SubIndex is shut down.
Shutting down the SubIndex stops the ControlledRealTimeReopenThread,
which immediately sets the generation to Long.MAX_VALUE and notifies
any callers currently waiting on a generation. But the call sites
using NrtFutures in LuceneChangeIndex are not calling NrtFuture.get,
which waits on a generation; they are using Futures.allAsList, which
depends on addListener.
So, we need to check the generation number in the call to addListener
as well as in in get.
This may also improve performance, since futures in some cases will
not need to wait for a reopen thread refresh cycle before notifying
their listeners.
Change-Id: I0ec3e315638a90dd5da3cf7f0d2cc50599a5bb36
|
|
|
|
| |
Change-Id: If89b43f642e2cd06c29e1549273832ca70d07cb4
|
|\
| |
| |
| | |
cherry-picking" into stable-2.8
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When using the web ui (both old and new ChangeScreen) to cherry-pick a
change to a branch that already has this change (e.g.: cherry-picking
on the same branch to get rid of dependencies), the corresponding
patchset-created event had its patchSet.uploader set to the Change's
owner instead of the current user. We now set it to the current user,
so stream-events consumers can properly detect who uploaded the
rebased patch set.
Change-Id: I0613ecae06b7843b8ca0294f83378c3917c26810
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The JavaDoc generation reported the following warning:
.../EmailReviewCommentsExecutor.java:29: warning - Tag @link:
reference not found: EmailReviewComments
Change-Id: If30a0e08ccbcea0ee55d208cc2dc6840c4dab00e
Signed-off-by: Edwin Kempin <edwin.kempin@sap.com>
(cherry picked from commit a5d2ad3933bd021d6a5a947642bf14bcfd1b998b)
|
| |
| |
| |
| |
| |
| |
| |
| | |
In the top right corner of a file the navigation cluster has a
tooltip on the up arrow but did not show the tooltip on the left
or right arrows. Show the key in the tooltip.
Change-Id: I192ef0ab8155c15b4bb35a18da33f74f15bb5537
|
|\ \
| |/
|/|
| | |
into stable-2.8
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If a group contains more than 'addreviewer.maxWithoutConfirmation'
members adding it as reviewer to a change requires a confirmation. A
user should only be asked for the confirmation when reviewers are
added from the WebUI but not when the set-reviewers SSH command is
used. This is also how 'addreviewer.maxWithoutConfirmation' is
documented in 'config-gerrit.txt'.
Bug: issue 2599
Change-Id: Ieab601e573115a0b034e1dcebe1c602adda6770b
Signed-off-by: Edwin Kempin <edwin.kempin@sap.com>
|
|/
|
|
|
| |
Change-Id: I5d136856f4d9794b74914579b39f2a1da3eb4ce8
Signed-off-by: Edwin Kempin <edwin.kempin@sap.com>
|
|\
| |
| |
| | |
stable-2.8
|
| |
| |
| |
| |
| | |
Change-Id: I804ab1f8034ac9fb1be9d178acce509f6d997d09
Signed-off-by: Edwin Kempin <edwin.kempin@sap.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The project list screen was taking a long time to render over a large
amount of projects (1,000+) and with even larger number of projects
(3,000+), it could make the browser unresponsive.
Project list screen now uses pagination to resolve this issue. The
number of projects displayed is determined by the 'Maximum Page Size'
user preference.
Bug: issue 2215
Change-Id: Icd0a7d54fd5c5b3c2301c31026e7c6717a648a24
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
List projects endpoint has an option to limit the number of results
returned but it was missing an option to start at a nonzero offset in
order to support results paging.
Add and document 'S' option which allow to query project list starting
at a nonzero offset. Also add missing documentation for the 'n' option
which limits the number of results.
Those 2 options will be then used to render project list with pagination
in web UI.
Change-Id: I19c4cd1cfa6ed78f22c84f249af54a5c30387561
|
|
|
|
|
|
|
| |
a7e34313 broken the acceptance tests.
Fixed.
Change-Id: I577cc875bcce3013cabd2f1b6813e7be820d4b3c
|
|
|
|
|
|
|
|
|
|
| |
Every time key up event is raised in the project list filter text box,
the project list was getting refreshed even if the filter did not
change (e.g. moving the cursor was refreshing the list).
Only call the refresh if the filter changed or if enter key is pressed.
Change-Id: I2e608d89e16dbef0cfe1e503ac2feb4639c27f5d
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In the new change screen, if the user clicks on a binary file in
the file list, the unified view is used. Then when navigating to
a previous or next file that is not binary, the diff view stays in
the old unified setting. It is only possible to get back to the
new side-by-side view by going back to the change screen and then
opening the non-binary file from there.
To keep consistent behavior of diff view, always use unified diff
for binary files. Use the user's preference setting for non-binary
files.
Change-Id: Ie4ece4cb740df8a69cbf3d2d42e4b0fb05461520
|
|
|
|
|
|
| |
- Fix plugin name in REST API documentation synopsis lines
Change-Id: I3bcf6da0e4ec3fb76aab3a93e08da427df39c09b
|
|
|
|
|
|
|
| |
The resource parameter specifies the resource to be read,
not the resource to be modified.
Change-Id: I4010674a348cefad02c234118ff25cf332dd76b4
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The SubIndex.NrtFuture objects are added as listeners of
searchManager who was found to hold on to them forever.
Fixed. There are also some code refactoring in NrtFuture.
Change-Id: If87cb62890a1cfa6c6336f6c7953a1cb56d42937
|