| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
Release-Notes: skip
Change-Id: I15d05b44cd22f9935fc8027e58f8bac052020cba
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
Replace with firing a toast instead.
Release-Notes: skip
Change-Id: I6c05f49ed78424ab144f65432cb8c00eae7bc255
(cherry picked from commit 2b4fc746b19198e04bc45d0fa02d5eba0f2ff24e)
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently gr-change-view can get to state when it add twice listener
to _handleCommitMessageSave. This results to 2 /message requests with
the same content, that currently back-end doesn't process well and
change can get corrupted by 2 same patchsets.
This is quick fix to disable 2 /message request to be sent. The fix
for multiple listeners will be done in following change.
Google-Bug-Id: b/231981018
Release-Notes: skip
Change-Id: Ib14db7ae3257fc716926c0060d1d749348c56141
(cherry picked from commit cfcf0af437c15d12c9ed2bc9961193afa3af5935)
|
|\
| |
| |
| | |
into stable-3.6
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When checking ref visibility, we need to parse change metadata.
To preserve existing behavior, we pretend that such change refs are
not visible.
Release-Notes: skip
Change-Id: I65331af8e5e8d66669d6a87f24fc945cafad01f7
(cherry picked from commit 766062542fe670ac514a96fbae129a79d185210f)
|
|\ \ |
|
| |\ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This commit was reverted during the merge:
Make PatchSetApproval UUID parsing forwards compatible
* stable-3.5:
Add a build icon for plugin use
gr-icons: Update links to material icons
Make PatchSetApproval UUID parsing forwards compatible
Release-Notes: skip
Change-Id: I89c184841aee850d7cf100308d37cffde5232518
|
| | |\ \ |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The change 324937 is backward incompatible: the data,
written by release stable-3.6+ can not be parsed by prior versions.
This fix skips UUID from the footer, allowing old releases to parse new
data format.
Issue: 15909
Change-Id: I9aeabd2fc34d1d2c616d4a06de368cff72754036
Release-Notes: this change has to be reverted, when merging to stable-3.6 branch.
|
| | |\ \ \ |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Screenshot: https://imgur.com/3ANfGwR
Release-Notes: skip
Change-Id: I0102ef00c6e6ad6014507e5e5634846e5e637db4
|
| | |/ / /
| | | | |
| | | | |
| | | | |
| | | | | |
Release-Notes: skip
Change-Id: I5fe855733c5951c1ff32ae7ef610d4977c897504
|
|\ \ \ \ \ |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Change Ief321b76 introduced a functionality whereby soy template are
refreshed asynchronously through the usage of an in-memory cache, which
stores compiled soy templates.
In case the soy templates were not customized in the
`<gerrit_site>/etc/mail`, the method performing the refreshing however
failed with:
WARNING: Exception thrown during refresh
java.util.concurrent.CompletionException:
com.google.common.util.concurrent.UncheckedExecutionException:
java.lang.IllegalArgumentException: resource
com/google/gerrit/server/mail/Abandoned.soy not found.
This is because the reloading happens asynchronously in the context of a
thread which has a different class loader from `MailSoySauceLoader`,
where the reload is triggered (this implementation is provided by guava
[1]).
This causes the soy templates to be searched in a path relative to the
reloading thread rather then main class loader.
Load the soy template by explicitly using the class loader of
`MailSoySauceLoader` instead.
[1] https://github.com/google/guava/blob/v30.1/guava/src/com/google/common/io/Resources.java#L192
Release-Notes: skip
Bug: Issue 15911
Change-Id: I41407ca342e209a7861a606cff7557280dab1e61
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
* changes:
Fix issue with closing comment
Finish rendering a message before scrolling
Migrate gr-message to Lit
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
While we typically want to use new objects, so that lit knows that
something has changed. This breaks for messages, as we need the original
message in the model to store whether it is expanded or not, otherwise
when a comment is saved and gr-comment-messages is retriggered to
render, the messages have all expanded set to false again.
Google-Bug-Id: b/232286621
Release-Notes: skip
Change-Id: Id116e2b4d2c86e0890bb2635571ec4507048a4e3
(cherry picked from commit ed4af620b17899fde19b8827bf0ce58067914162)
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Release-Notes: skip
Change-Id: Ibdc1c20dfa43904ceb4b760d3ebd9649f6e11d5b
(cherry picked from commit 541f5bb80fefa0246738b56345005719f9d7668a)
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Release-Notes: skip
Change-Id: I3de8ec7086e9d4d7807edda62dbb4e11697586c6
(cherry picked from commit f0e3e9ed53c44c35db48abe9e9538ede6e7ab5c1)
|
|\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | |
| | | | | | | | |
stable-3.6
|
| | |_|_|_|_|/
| |/| | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
When user change patch range from patchset 1 -> patchset 2. File list
correctly shows file change between 1-2. When user clicks on change
number, patchset range is reset to base vs patchset 2. However
file list is not reloaded to show changes base vs 2.
This is caused because patchNum is undefined, which is first check
in file-list reload function. It works ok on first render, because
patch gr-patch-range-select set latest to patchNum.
Google-Bug-Id: b/230097909
Release-Notes: skip
Change-Id: Icb96b5d7d861adfb4c8a4d0e4b1209b4975ada13
(cherry picked from commit 025bf8fa9cafba9ec6ae2540e84beafaf75f289d)
|
|/ / / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
After clicking edit button, editable-content tried to put focus into
textarea. However iron-autogrow-textarea was not yet rendered so
it throw js error:
Error: selector 'iron-autogrow-textarea' did not match anything'
Google-Bug-Id: b/228800894
Release-Notes: skip
Change-Id: I51a48f84f750798d6a93a00ca765b308341b7517
(cherry picked from commit 43d22deb14e3d25b67166e44b3f7a033d6c1ff27)
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
* changes:
Render pre-formatted text as both <pre> and <code>
Do not linkify text inside of code-blocks.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This is following the convention on https://daringfireball.net/projects/markdown/syntax#precode
Release-Notes: skip
Change-Id: Iaaaea16ec794bc9bb080696ee0011f69712eee9a
(cherry picked from commit 81198eeeb2e9f38b2ac86cb7f9b125c85d5ab589)
|
| | |_|/ / /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Google-Bug-Id: b/230529142
Release-Notes: skip
Change-Id: I522ee8ef87a4640e449be492aa2351a736bdff13
(cherry picked from commit d5b98ee409cb43765686b4518fd66646c432b193)
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Broke in 75b390cb1b2c5db48e4f3da19f2237f61cba12f8 and
0fefcf1ff7bd934edf325dfab957915e4a1335da
Release-Notes: skip
Change-Id: I932841f091b523dad8e875be921fa2dec6be03ad
|
| |_|/ / /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Before: https://imgur.com/a/IO0sbAE
After: https://imgur.com/a/DsFFqSt
Google-Bug-Id: b/232135614
Release-Notes: skip
Change-Id: Ifcc10f48c3815b447afaae91596270bfc4e8ef1a
(cherry picked from commit 68a4112eb1362fabbf70afe973f82ec934677e1a)
|
|\ \ \ \ \
| |/ / / /
|/| | | | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Release-Notes: skip
Change-Id: I3c570fa09c3f4b0812d7723d6dd71cbb85fd6c27
(cherry picked from commit 3f50213c67dbb0dc2172fe0deb195f040874e754)
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
We previously cached a DefaultRefFilter instance in ForProject. This
was done because we cached state we wanted to reuse. It's no longer
true, so we remove the member variable.
We also move the per-instance cached visbileChanges map to just be
an argument. This makes the machinery harder to use the wrong way.
Release-Notes: skip
Change-Id: Ia2748a8f976eb9c5f27de74ce1e48b27d85a3863
(cherry picked from commit 696758eddd4451d87b226310ce8f95536f5bc83b)
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Git Protocol v0 was a 'server speaks first' setting. That meant, that
the Gerrit server had to advertise all refs available for fetching
before knowing what the client wants.
For repositories with many changes, this meant that we had to
advertise thousands, sometimes millions of refs. To cap expensive
filtering and excessive wire transfer (sometimes 10s of MiB), we
used the change index to only advertise the N most recent changes.
At Google, that N was approximately 10k. So we advertise
- refs/heads/*
- refs/tags/*
- refs/changes/ -> 10k most recent (so /meta + patchset refs)
This heuristic cap means that:
1) Users can't fetch older changes
2) We always filter N changes (expensive) for visibility
Git protocol v2 added refs-in-wants so that the client tells
Gerrit what refs it actually wants. We never fully implemented
that in Gerrit's ref filtering. This commit does that by:
- Detecting that the number of change refs that we should filter
visibility for is small (<5)
Here it is important that we check on the number of changes (not refs)
because refs-in-wants allows to fetch more than one ref at once.
Git's tag autofollowing also always requests refs/tags/*, even if
you want to fetch just a single ref, say refs/changes/<id>.
- If the number is small, evaluate visibility using PermissionBackend
- If the number is larger, use the existing ChangeIndex backed filtering
to be efficient enough
This means this commit is a no-op for Git clients with v0 and makes a
better experience for v2 users with refs-in-wants.
Release-Notes: Protocol V2 allows fetching old changes
Google-Bug: b/122888978
Google-Bug: b/142876048
Change-Id: I61b3b6763a5cfecb1e662d62b5782d37d04b0919
(cherry picked from commit e2781b4d8fc5335b7f95e2c24fd565ed2babb121)
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
AbstractDaemonTest used to contain logic to disable secondary indices.
This means more code in an overly full class and also made it hard
to reuse that code in tests that did not inherit from AbstractDaemonTest.
This commit factors out this logic in the spirit of *Operations test APIs.
You can inject a IndexOperations.{Change,Account,...} and use it to
disable reads/writes.
Release-Notes: n/a
Change-Id: I26f59ac8a667b29d4138e8aadad647488d36f082
|
| |/ / /
|/| | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Permission on refs/changes are not respected by the permission
backend. Change refs are visible if the change is visible to the
user. In these tests, the change was already made visible by
pushing them to master, which is visible to registered users.
This commit reworks the permission grants to work on All-Projects
instead of the test repository. It wipes existing permissions
before the test. This makes it more obvious what is happening
because the repo under test inherits permissions from
All-Projects and setting permissions on that repo means the reader
must also understand what permissions are set in All-Projects
which makes it harder to know what is going on.
Release-Notes: n/a
Change-Id: Ib5d3bbfeb4238ef0545024738555d96b3f06acfa
(cherry picked from commit 30d43cd7d2b816712578203e64708bcf60b22c19)
|
|\ \ \ \ |
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This is based on the typescript style guide.
Release-Notes: skip
Change-Id: Ib648c7ef178339a68a3c3ae251a36df1009f5fd4
(cherry picked from commit 5ea7d6526b1f24097fc7b0b016dd9181471b9310)
|
|\ \ \ \
| |/ / /
|/| | |
| | | |
| | | |
| | | | |
* changes:
Simplify gr-change-actions willUpdate
Migrate gr-change-actions to lit
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Test needed updating because the download action was added to the overflow list.
Google-Bug-Id: b/231946357
Release-Notes: skip
Change-Id: I10fc8a84382b7db8687017231540ae4f7c18b096
(cherry picked from commit 9369a9fb65c836f7a1f5a1636a4d8a35f3ade4ed)
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Release-Notes: skip
Change-Id: I87f3ded5863ba98df2cd6083407f279220e42d25
(cherry picked from commit 32b861af38fc81de08a443df1216e28c3360028a)
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Change I893acbe61 erroneously removed the impl-log4j dependency, which
is actually needed to avoid using the NOOP logging implementation in the
tests.
This fix was already applied for the gerrit war (in I86af69dd), but
never for the testing targets.
Release-Notes: skip
Change-Id: I5492e96cdcc6ff69050a44d1cb07a89d3a101185
|
|\ \ \ \ |
|
| |\ \ \ \
| | | |/ /
| | |/| |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
* stable-3.5:
Fix the gr-autocomplete.
Consider non-empty performance log-records as non-empty
LoggingContext: Consider empty aclLogRecords as emtpy
Make tracing.performanceLogging default disabled
Fix bug in gr-avatar which wasn't hiding avatars when "hidden" is set
Release-Notes: skip
Change-Id: I8a696ca92f1698bb61fbf30ddf32897fe78646ff
|
| | |\ \ \ |
|
| | | |\ \ \
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
* stable-3.4:
Consider non-empty performance log-records as non-empty
LoggingContext: Consider empty aclLogRecords as emtpy
Fix bug in gr-avatar which wasn't hiding avatars when "hidden" is set
Release-Notes: skip
Change-Id: I15770e210e1ee3ac631720ccc8157bbe35584b55
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
The improper check for whether performanceLogRecords was null or not
was removed in [1] but wasn't replaced with a check whether the
performanceLogRecords was actually empty or not.
[1] https://gerrit-review.googlesource.com/c/gerrit/+/295125
Release-Notes: skip
Change-Id: I70ff8046226914ed53196d3f2b8b9b6468a6d25e
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Bug: Issue 15889
Release-Notes: Fix "Logging context is not empty" log spam.
Change-Id: Id34aa123a771466196049635c5ee13c9def19063
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This happened for sites that did not set an avatar. Seems a recent
regression as it works on gerrit 3.3.
Release-Notes: skip
Change-Id: I148716b36c3d57e3eb3f1486b9539f1741d7288b
(cherry picked from commit e0358c2e4e7db76a7efe01de9cc14ef16af03a2e)
|
| | |/ / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Don't listen to keys on the gr-autocomplete-dropdown, instead do the
right thing for item-selected in gr-autocomplete.ts
Bug: Issue 15887
Google-Bug-Id: b/207422954
Release-Notes: Fix the gr-autocomplete.
Change-Id: I5bb51bc54c817830a92c66f982863c13b5fa8497
(cherry picked from commit be1d4b5369b7ec5ee33de88ada85baa5c8906c34)
|
| | |\ \ \ \ |
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Having this enabled caused a significant heap-usage increase even if
there were no performance logging plugins installed.
Bug: Issue 15531
Release-Notes: tracing.performanceLogging is default disabled
Change-Id: I4d2674eeeff64fe1ea1d79ef1f3472944936e054
|
|/ / / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
.name=${} doesn't work properly. It was found by fixing bug in
change 336724. Fixing rest of names in this change
Release-Notes: skip
Change-Id: I9b9ad3c0cc5f90c417e4020429e7cf3503fa6dd9
(cherry picked from commit 11168908a773ed5ae31bb9c4469a4b9fb76e095d)
|