| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
Release-Notes: skip
Change-Id: I6b222674624e9d7955c1cf2622779806c44ebf2f
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* stable-3.4:
Set version to 3.4.5-SNAPSHOT
Set version to 3.4.4
Set version to 3.3.11-SNAPSHOT
Set version to 3.3.10
Ignore '--no-limit' query changes option for anonymous users
Show HTTP(S) message in Missing Change ID Error message
Update reload4j to 1.2.19 and slf4j to 1.7.36
Move SLF4J to nongoogle.bzl
Separate reviewer-recipients from project-watch-recipients
Restore Java 11 binary bytecode for releases to Maven
Update git submodules
Release-Notes: skip
Change-Id: I057b11ba549129dfbb1cc8a72cd9940d4fba9b9a
|
| |
| |
| |
| |
| | |
Release-Notes: skip
Change-Id: I235e37ce7f38c3bd3bd944e6a3315795e08c795f
|
| |
| |
| |
| |
| | |
Release-Notes: skip
Change-Id: I761ffa6a464a58e616cb282beba0ca950a257030
|
| |\
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.3:
Set version to 3.3.11-SNAPSHOT
Set version to 3.3.10
Ignore '--no-limit' query changes option for anonymous users
Release-Notes: skip
Change-Id: I323b6708075bbe54624d72f39a5a3e39fb0856f4
|
| | |
| | |
| | |
| | |
| | | |
Release-Notes: skip
Change-Id: Iffa172335f1347e466349d54ef1ede871aaa7dfb
|
| | |
| | |
| | |
| | |
| | | |
Release-Notes: skip
Change-Id: Ifad11be501dcec40077e4b5e201c9fe7d10307e4
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Adding 'no-limit' option to query changes REST API can result in
substantial resources usage. This change ensures that it cannot be used
(or abused) by anonymous users.
Notes:
* one can still configure them to request unlimited results by setting
'Query Limit' Global Capability to Integer.MAX_VALUE for 'Anonymous
Users' group
* 'no-limit' option is only a part of query changes API hence accounts,
groups and projects are not affected by this change
Release-Notes: Ignore '--no-limit' for anonymous users change queries
Change-Id: Ic789690ffd2f94f02989c2906fcd75e442df86f8
|
| |\|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.3:
Show HTTP(S) message in Missing Change ID Error message
Update reload4j to 1.2.19 and slf4j to 1.7.36
Move SLF4J to nongoogle.bzl
Separate reviewer-recipients from project-watch-recipients
Restore Java 11 binary bytecode for releases to Maven
Release-Notes: skip
Change-Id: I194e44a9939a89fb06a37eacf780e0461100631c
|
| | |\ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
reload4j 1.2.19 fixes
- CVE-2022-23305
- CVE-2020-9488
slf4j 1.7.36 updates its reload4j integration to reload4j 1.2.19
Release-Notes: Update reload4j to 1.2.19 and slf4j to 1.7.36
Change-Id: I6b00f94b644cacc777d7b92f85e26bec44b9e40d
|
| | |\ \
| | | |/
| | |/| |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The assumption that the existence of the SSH host key means that
all users use SSH protocol is incorrect, since some might use HTTP
protocol.
Since a CommitValidationListener doesn't know which protocol was
used, provide both SSH and HTTP variant of the command to install
the commit message hook.
Release-Notes: skip
Change-Id: I10ae721315de8afb964e720033e2d6498bce3e55
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Gerrit core and all plugins used by Google migrated to Flogger so that
Gerrit at Google no longer depends on SLF4J. Hence the SLF4J version
doesn't need to match the SLF4J version that is used at Google.
Signed-off-by: Edwin Kempin <ekempin@google.com>
Change-Id: I40ff875a33aa628eef5f387e22768665188b434f
Release-Notes: skip
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
To ensure that project-watch emails are sent regardless of other
notfication configurations. Without this fix you will not get any
email notifications for watched projects when you have set
"Only when I am in the Attention-Set" and you are outside of
the attention-set.
Release-Notes: Fixed lost watch notifications when attention-set-only is configured
Bug: Issue 14912
Change-Id: I0707c588f23f13c617cdfb565171adff12833ea9
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The switch to Java 8 as the default java toolchain in
Change-Id: Ieb0d627ce5f4 has missed that the Maven
release script generation was relying on the default
toolchain for publishing.
Introduce the Java 11 toolchain as the target for publishing
to Maven, so that the binary bytecode can be restored
as it was expected for all current v3.3.x versions.
Release-Notes: skip
Bug: Issue 15759
Change-Id: Ia4c29d1db800220cdbc4f1c015fdec9bd552188b
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Update plugins/replication from branch 'stable-3.4'
to 44171b65a0c1c390764cdd045aff1cd138c4b802
- Ensure states are updated for canceled replication tasks
When using 'replication start --wait' we need
ReplicationState.waitForReplication() to return when the task we're
waiting on has been canceled, either through an admin action or because
the replication distributor determined another node already completed
it.
Add a couple tests for PushAll that confirm this behavior was previously
broken and is fixed now.
Change-Id: I36320ae079af5d7673e05d20ddc94b42a9b04347
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Do not rely to throwing AuthExceptions for checking
change visibility but ForChange.test instead, which
returns a boolean.
Relying to exceptions to be thrown and catching them
for a simple change filtering would cause the JVM to
trigger multiple context switching and slowing down
the overall ACL evaluation.
A slow ACL evaluation for changes filtering may
result in the increase of threads utilisation
while the change query is executed.
Release-Notes: skip
Change-Id: Ie488a41b28b5c19710a11c2840793f04b1d3861b
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Compiling regular expressions is known to be expensive. If the access
section contains a regular expression, we pre compile and memoize it.
Release-Notes: Memoize reg-ex compilation in projects cache
Change-Id: I84f50323e9f08e69ab6414203927c7d2eacb23e4
(cherry picked from commit e235c42b099138f713ceef034923bfd7731f75ad)
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This code moves ordering from ProjectState to the cached value creation. The
ordering is univerally useful and it's better for performance to cache it.
Release-Notes: Order project's access section of ACLs
Change-Id: Icda97cd401212d7139c2b91b789f33fe99234f50
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Certain permissions cannot be used in All-Projects and are filtered
out before we evaluate them. This used to be done in ProjectState
and was cached when that entity was cached. Now, we cache immutable
data objects (mostly AutoValues) for better thread safety.
When the ProjectCache was refactored, this logic was left in
ProjectState. This means it has to be recomputed quite often when
permissions are evaluated. A recent profile showed that we can
save CPU time (hence request time) as well as heap if we cache
the outcome of this filtering.
This is what this commit does. Existing tests check that permissions
that are not allowed on All-Projects (e.g. OWNERS on refs/*) are
indeed not evaluated.
Since this filtering depends on if a project is considered to
be All-Projects or not, we have to invalidate the cache when
the name of All-Projects changes. We document this fact.
Release-Notes: Cache permission filtering for All-Projects
Change-Id: I01e08e70bf36aaf4be522564286ab4a86e7681a3
|
|\| |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.4:
config-gerrit: Fix 'database' references
Remove leftover receive.changeUpdateThreads references
config-gerrit: Remove extra list continuations
Fix the ${hash} substitution for diff file web links
Add ${hash} as a substitution variable for gitweb file links
Change-Id: I5377a1cb9007e7de545d6b6874592a80d06aaa4a
Release-Notes: skip
|
| |\|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.3:
config-gerrit: Fix 'database' references
Remove leftover receive.changeUpdateThreads references
config-gerrit: Remove extra list continuations
Change-Id: Ia5ee3ed9473f4f48bba55f5e8f0f413b322cb083
Release-Notes: skip
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
These were missed in the NoteDb conversion.
Change-Id: I1d223e19b2e0771a9ddf3f9aaf8ca4fa04b6c9c3
Release-Notes: config-gerrit doc fixup
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Change I8c435856c5 removed the dead code that used this config, but
missed removing the config from the docs and tests.
Change-Id: I2dbd2e93b334fdc7ea4f3abdb88a27c904ab48c0
Release-Notes: Remove non-existent receive.changeUpdateThreads setting from docs and tests
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Also add a blank line where one was missing.
Change-Id: I141347972e835d9797d6801a259b6587abb180c5
Release-Notes: config-gerrit doc fixup
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
It seems that when we were refactoring this change to address review
comments we accidentally swapped out the ${hash} replacement for a
revision replacement. Fix this by swapping it back to a hash replacement
which was the intended behavior.
Note we skip release notes as this addresses a bug in a recent change
that should have accurate release notes already.
Release-Notes: skip
Change-Id: I0b1c56d9aa80b0014e26f2a2385df90b1f5ae413
|
| |\ \
| | | |
| | | |
| | | | |
stable-3.4
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Gitweb file links interpolate ${commit} but this may be the ref rather
than the commit hash when interpolating file links. This is problematic
because some gitweb like systems (in this case gitea) rely on the hash
rather than the ref path.
We address this by adding a new ${hash} interpolation value that will
always be the sha commitId value in the file links. We don't change the
value of ${commit} as it appears gitiles may rely on the current
implemenation and there may be other implementations out there relying
on the current values for that variable.
Bug: Issue 15589
Co-Authored-By: David Ostrovsky <david@ostrovsky.org>
Release-Notes: Add SHA-1 hash variable to gitweb file links
Change-Id: Ifabc2c27c62cd55a7b8d43cfa43f6dc1539423eb
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Update plugins/gitiles from branch 'stable-3.5'
to 9011b868d1fab8ae35d41cafed922b0b19899dde
- Merge branch 'stable-3.4' into stable-3.5
* stable-3.4:
Update getFileWebLink signature to match Gerrit
Change-Id: I7b8b02b8ccc230abc07d70722c600a811d838896
Depends-on: https://gerrit-review.googlesource.com/c/gerrit/+/331440
- Update getFileWebLink signature to match Gerrit
Gerrit is getting a new getFileWebLink method signature in order to
support additional gitweb alternatives like gitea. Gitiles needs to
update its overridden method to match the new signature.
Depends-On: https://gerrit-review.googlesource.com/c/gerrit/+/329279
Change-Id: I9b0e73504245e22cf5ad1b0075ca8a972d0a703e
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Pick up a couple fixes before plugin/gitiles:master moved to gitiles
1.0.0 and use this as the point to create a stable-3.5 branch for the
repo.
$ git log --oneline fa993c0..a0709a4
a0709a4 Simplify FilteredRepository.FilteredRefDatabase#getAdditionalRefs()
97ce60f Ignore JdkObsolete issue with StringBuffer
Change-Id: I385c332710fd833329b53e5dd09346414679f823
Release-Notes: skip
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The DefaultPermissionBackend envisaged the need of
of a thread-specific cache for avoid creating
and parsing the permissions of a project multiple times
during the same API execution.
However, the change missed the inclusion of the ProjectState
creation inside the lambda, causing the parsing of the ProjectConfig
multiple times for the same project even when it was unneeded because
the ForProject object was already cached.
Release-Notes: Improve performance in ACL permissions evaluation
Change-Id: Ia28419dd6912ef712630554184abb47ed63b68a9
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This fixes an issue where the background for the description
box was grey. It should be white if you can edit and darkened
if you cannot.
Release-Notes: skip
Change-Id: I38ea8a2543b53c8319264fec686a709a69c43234
(cherry picked from commit 278bbfb32cf976b82da47354f2b141f6b506de62)
|
|\ \ \ \ |
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This is being done because change [0] brought includedIn to
change-actions. It was being hidden for logged out users even though
previously it was shown for logged out users.
It correctly hides the actions that the current user cannot do
even if logged out. Thus we can show it regardless.
[0] I6cfde70242d6a2bf45d1b3c3dd25ed1ca05af1be
Release-Notes: Show gr-change-actions regardless if logged in or not
Change-Id: I33470ca664f87b88347b16dcb2407d6f389d25de
(cherry picked from commit 7e2f9f36904f1cf3a3a0932aab509977551606d5)
|
|/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This fixes an issue where the background for the description
box was grey. It should be white if you can edit and darkened
if you cannot.
Release-Notes: skip
Change-Id: I185b49a8c770b27e1c6fea81be16602527b9a7c1
(cherry picked from commit 4d87db8d5e425c637c4fac1e347dbe8a31f7f58e)
|
|\ \ \
| | | |
| | | |
| | | | |
tree" into stable-3.5
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Versions of Gerrit older than v2.13 may have cache-automerge refs
which point to a tree. These older refs are not considered valid by
AutoMerger, but are not removed or invalidated.
BaseCommitUtil's methods' callers expect to be returned objects that
are persisted in the repository and creating auto-merges in-memory
is incompatible with that expectation. Solve this by no longer doing
in-memory merges and by validating if the cache-automerge ref is
actually usable(i.e exists and points to a commit). If the ref is found
to be invalid, we create the auto-merge commit in git and force update
the ref. Other callers of AutoMerger should be unchanged as the behavior
of methods in there is not modified. Instead a new helper is created for
BaseCommitUtil.
Release-Notes: skip
Change-Id: I15f9c951cdaa3733963e6f8b38066c11be8712f7
|
|\ \ \ \
| |/ / /
|/| / /
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.4:
Fix gerrit ls-members --recursive
Fix gerrit ls-groups flag in documentation
Test that endless loop on Prolog type error is fixed
PrologRuleEvaluator: Fix endless loop on type error
Bazel: Use Java 8 per default to guarantee source level compatibility
Release-Notes: skip
Change-Id: Icedae403f14deac962137eb7645b1229a81d798b
|
| |\|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* stable-3.3:
Fix gerrit ls-members --recursive
Fix gerrit ls-groups flag in documentation
Bazel: Use Java 8 per default to guarantee source level compatibility
Release-Notes: skip
Change-Id: I6501e5fbeb13c35c4e40efdf2d1420c9d4425a62
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Id3ec26677 introduced a regression to the ls-members SSH command that
caused it to ignore the --recursive flag. This was caused by replacing
apply() with getDirectMembers() which only looks at direct membership
and not transitive membership.
To fix this we update the parent restapi class to include a getMembers()
method that inspects the recursive flag and applies the correct
memerbship lookup method.
Release-Notes: Honor --recursive option on ls-members command
Change-Id: I7379be1a99b58e32da0e1fc518c774ea65ae8e88
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
In Gerrit 2.14 the -q flag to ls-groups was deprecated in favor of the
-g flag. With change I1a1bdbbcd -q has been removed and -g must be
used. The docs were never updated fix this.
It is worth noting that there may be other flags that should be added to
the docs as it appears the rest api defines what we can pass in here. I
have not confirmed this, but a docs audit for incorrect or missing flags
may be worthwhile.
Change-Id: If9ea99724f182641bf349ce7faaa24f43d988bd6
Release-Notes: Fix gerrit ls-groups flag in documentation
|
| | |\
| | | |
| | | |
| | | | |
compatibility" into stable-3.3
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
At the time the change 264535 was submitted, that switched Java default
to Java 11, not all Gerrit users were ready to abandon Java 8
compatibility. To provide Java 8 compatibility it was decided to perform
CI verification with Java 8, by overriding the java toolchain option
on GerritForge CI (core build) and Zuul (plugins build).
Today the situation is different. Starting from stable-3.5 branch Java 8
compatibility has been dropped. Still we guarantee the Java 8
compatibility on stable-3.3 and stable-3.4 branches. To achieve this,
the java toolchain option is overridden on stable-3.3 and stable-3.4
branches on the CI.
To simplify the build toolchain and the CI configurations, we would like
to rely only on the default java toolchain (provided in .bazelrc) to be
the minimum guaranteed compatible Java version (on each branch).
That way the default bazel invocation:
$ bazel build :release
would always provide the desired result per default.
This change should be reverted when merged up to stable-3.5 branch.
To summarize:
before this change is submitted, the situation is:
stable-3.3: default toolchain: Java 11, guaranteed compatibility: Java 8
stable-3.4: default toolchain: Java 11, guaranteed compatibility: Java 8
from stable-3.5: default toolchain/guaranteed compatibility: Java 11
after this change is submitted, merged to stable-3.4 branch and reverted
started from stable-3.5 branch, the situation is:
stable-3.3: default toolchain/guaranteed compatibility: Java 8
stable-3.4: default toolchain/guaranteed compatibility: Java 8
from stable-3.5: default toolchain/guaranteed compatibility: Java 11
Change-Id: Ieb0d627ce5f4d2753fe637fb67123e6431796940
|
| |\ \ \ |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The new test reproduced the endless loop that was fixed by change
Ie922c8db1. Without the fix the change creation via push fails with an
internal error due to the stack overflow.
Signed-off-by: Edwin Kempin <ekempin@google.com>
Change-Id: I4112cfc455940d0f1a019f825b0dd85e8c6a221f
Release-Notes: skip
|
| |\| | | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
If a Prolog rule fails to load and PrologOptions#logErrors() is true
(which is e.g. the case if the request is being traced) there was an
endless loop in PrologRuleEvaluator:
1. typeError(String, Exception) is invoked which logs the exception and
then calls typeError(String) with DEFAULT_MSG as parameter
2. typeError(String) delegates to typeError(String, Exception) with the
exception parameter being null which brings us back to 1.
The issue was introduced by change I018ce9e13.
Signed-off-by: Edwin Kempin <ekempin@google.com>
Change-Id: Ie922c8db1ffa51275865f7dfe3432462995b5c84
Release-Notes: Fix endless loop when handling Prolog type errors
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
In Ice695c47c8 the id of grp-app element was changed from "app" to
"pg-app" but one place was missed to be renamed.
To avoid proliferation of pg-app id across the code base, replace lookup
done by getElementById with lookup done by querySelector with gr-app
element.
Bug: Issue 15686
Change-Id: I4fbf7ecc2644567919672bd91a1de506797408c9
Release-Notes: skip
|