Running the script in the root of the Gerrit folder:
./tools/version NEW_VERSION
will replace the value in the first occurrence of the <version> tag
in the plugin archetype pom files with the value specified by
NEW_VERSION.
Change-Id: I9f529852af22a1eef7f30e2fcaea8acea28057d9
The GWT compiler swamps a system when it builds, using multiple
threads to build each permutation in parallel. It also requires
a lot of RAM. Buck is not aware of the cost of the genrule and
currently schedules both to build at the same time.
Make ui_opt depend on ui_dbg so Buck is forced to serialize these
build steps, reducing the load on average systems while possibly
extending build time on very powerful (e.g. 16 core) systems.
Change-Id: Icc1856c9c4e7919c28dc533e792fa25aadcdf801
This commit changes the mysql setup documentation because
there is no need to use latin1 encoding if you are using
another engine than the MyISAM engine.
Bug: Issue 1273
Change-Id: I432957381385c3e0390bb0db245139954300ecb8
(cherry picked from commit 13d07ecc7cd98cfd928a2b06d8755aea1fddde61)
When pushing changes it is easy to make a typo in the refspec and
in this case new branches should not be created. If administrators
want to create branches by push they should explicitly assign
themselves the needed access rights.
This was broken by [1].
[1] https://gerrit-review.googlesource.com/#/c/54048/
Change-Id: I768bfd6d7cb193b0ec9f51c3fa768087ceb91acd
Bug: Issue 2557
Disable the 'commitWithin' from within Reindex by overriding
the configuration with '-1'. Treat negative values as the
original behavior, auto-flushing but not auto-committing, which
is the least safe but the most efficient for reindexing the
entire site.
Change-Id: Ifdba797bee871d2a3d8928810a6304bacb850c8c
After editing the commit message and pressing "save", the button
remains active until the message is saved and the dialog is closed.
Usually the edit completes quickly, but on some systems where the
network is slow it can take a few seconds. If the user presses the
"save" button again (perhaps more than once) it will result in more
than one edit request being sent. This can in some cases result in
inconsistency between the database and the git repository, which
prevents the change from being viewed any more until a new patch set
is manually uploaded.
Disable the "save" button after it has been pressed, to prevent this.
Change-Id: I820095a1d91b7fd7c498c7c24fa5e7e81e40a7eb
As the owner still can't create a branch by pushing a reference over
ssh I think the previous statement isn't as clear as it could be.
Original commit 7ec4ac71ffa58f32e1443b0ca0110417244983c0 seems to
only deal with the create-branch command, not changing the behavior
over creating a reference by pushing it (over ssh).
Change-Id: Iac3f52b10ee6a92862622e8699c5daa1d4749247
Signed-off-by: Fredrik Luthander <fredrik@gandaraj.com>
Escape dollar sign in plugin manifest entries, to prevent that
it is interpreted by the shell and thus get removed.
Change-Id: Iefd8abdcd207cb985f8156c7e41c094c1aa12b27
Move the paragraphs about running the reindex program and the index
names up to the top of the index section. It makes more sense to
have them there, than in the middle of the setting descriptions.
Change-Id: I875a66173f752554803ca95bc166fba0dfbcd315
If the commitWithin setting is low, it can cause the reindex to
run slowly. Add a note in the documentation about this.
Change-Id: If4c394443bde89be4bfc1505d33322089e671493
Second and additional changes in the same submit attempt must be
built on the previously constructed merge, not the branch tip.
Include tests for these strategies and the same bug in Cherry Pick
that was fixed by cd60499cd65812d926806e49d15a6852c7798871.
Change-Id: I62524b39ee4104ab53b36c75b64c02b20b4d5e94
Second and additional changes in the same submit attempt must be built
on top of the previously cherry picked change, not the branch tip.
Change-Id: I9183f68b5780af1f28cb357217663e4bdb2d4e76
The bullet list for API artifact repository types was not being
rendered correctly due to a missing blank line.
Also, add backticks around the version numbers in the list.
Change-Id: I58d8845acde464d223e7838160be8d222e059f2b
Adding the ~/bin folder to the path in the current shell session
is not enough. It also needs to be added globally.
Update the documentation to mention this.
Change-Id: Ic5d4cf1a02b3e52eccb6e34ba3513a69785df1f9
Under certain circumstances, when pushing to 'refs/changes/n', the same
commit can be pushed onto multiple changes even if they are on the same
branch.
As a result of this, if the commit sha1 is used as change identifier
in the 'review' ssh command, it will fail with the error:
'fatal: <SHA1> matches multiple patch sets.'
This commit performs the same check as when pushing to 'refs/for/branch'
to prevent this, and will show "commit already exists (in the project)"
error message to reject uploading a commit to an existing change via
`refs/changes/n` if the commit was already successfully pushed to a change
in project scope.
Bug: issue 2374
Change-Id: I1c000b2cc3e155617cdc9c295a46ce2107ec47ca
The message reads:
"This patchset was cherry picked to change: <Change-Id>"
This is not very informative, since the Change-Id is identical
to the current change.
A better message would be:
"This patchset was cherry picked to branch <branch name> as commit <SHA1>"
Bug: issue 2513
Change-Id: I8bfb61b1a314cf2a985df7a571bb28e3f6cb49d4
When the 'commentlink.<subsection-name>.match' is configured wrongly,
e.g. invalid regular expression, the change screen cannot be loaded,
and there is no error message or log.
Show an error dialog.
Bug: Issue 2514
Change-Id: Ie36fcdb0be13690cacaa64e235a08d7913dea937
When the value of 'diff_intraline.timeout' is not long enough
to let IntraLineLoader compute the intraline, Gerrit will set its
status as TIMEOUT and cache it. When the user opens the diff view
a dialog will appear with Error title to say "Intraline difference
not available due to timeout".
Change the 'Error' title to 'Warning' to reduce number of complaints
from users.
Change-Id: I440d03593eb82d9c6c213cdc77d3fd3d50a760eb
New change screen changed the logic for submit status evaluation
by moving it partially to the client side. The change can be submitted
only when all needed labels have max scores and there is no labels exist
with min score. The computation of rejection is based on `rejected`
attribute in LabelInfo class. This attribute was set in two places:
1. As outcome of submit rule evaluation (with blocking functions)
2. In setLabelScores() to render the result for non blocking labels
After overriding of rejected attribute for non blocking labels it was
not possible any more to differentiate on the client side if one
particular label was rejected for information purpose only or because
it was blocked.
5a32a6dc15b73d412645598648ac089eacf29b65 was trying to solve it by
not overriding the rejected attribute in setLabelScores(). The
consequence of doing it was however, that custom labels with min
score were rendered not as rejected but as disliked.
This change introduces the `blocking` attribute in LabelType class to
carry out the missing information to the client to perform the correct
submit status computation. Now the rejected attribute can be overridden
for custom non blocking labels and rendered correctly.
Bug: Issue 2453
Bug: Issue 2497
Change-Id: Ic5092342b0e2153eb43f5e06fda4fcac4699844e
When updating an existing label value with command like
ssh -p 29418 user@review gerrit review <change>,<patchset> --label code-review=+1
where the lower case label name will make the update operation
become insert operation. This causes an exception:
"Unique index or primary key violation"
Use standard label name instead of lower case name.
Bug: Issue 2515
Change-Id: I735a5f84d0569554c050b60b8984ee83aa941d39
Direct push local commits to create a new branch is rejected,
even when user group can
- 'Read',
- 'Create Reference',
- 'Push'
on 'refs/heads/*'. User have to firstly create new branch with
a commit which exists on server side, then push other local new
commits.
The reason is that Gerrit take it for granted that the commits
used to create new branch exist on server side already, so Gerrit
only check the commits are reachable or not on the server side.
According to the same reason if user group can 'Read' on 'refs/*'
or have Global Capabilities 'Administrate Server', Gerrit will
think the commits sure will be reachable for user and stop do
further checking.
So Gerrit does not work in the cases where creating new branch with
local new commits, e.g. the commits come from outer partners and need
not review, and the user group is not of Admin group and should not
be able to see every thing for some business rules reason.
Taking local new commits into account, also including the related
permission checking, now it can be finished in one step to create
new branch and push local new commits.
Bug: Issue 2441
Change-Id: I4e580e2fe0b40b8d0fbb06c5c8d1efd61127a842
