Two test-only changes that cut the `-race` backend unit job's critical
path, with no behavior change.
- **`modules/auth/password/hash`** — `TestHashing`/`TestVectors`
exercised the CPU-bound KDFs (scrypt `N=65536`, pbkdf2, bcrypt, argon2)
serially on one core. Marking the subtests `t.Parallel()` fans them
across cores. The hasher registry they read is only mutated by the
non-parallel `Test_registerHasher`, so this is race-free.
- **`services/release`** — `TestRelease_Update`/`TestRelease_createTag`
slept `6x time.Sleep(2s)` only to cross the 1-second `CreatedUnix`
boundary. Replaced with an advancing mocked clock (`timeutil.MockSet`),
making the timestamp assertions deterministic and removing the real
waits.
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Updates `github.com/urfave/cli/v3` to
[v3.9.0](https://github.com/urfave/cli/releases/tag/v3.9.0) and removes
the renovate pin now that
[urfave/cli#2319](https://github.com/urfave/cli/pull/2319) (the `-c`
help flag parsing fix) is merged.
v3.9.0 prepends the default command name to the root command's args,
which broke the old `Root().Args()` check in `isValidDefaultSubCommand`.
It now uses the command's own `Args()`.
Behavior change: `./gitea web <extra-positional-arg>` now errors with
`unknown command` instead of starting the web server and ignoring the
trailing arg. `web` takes no positional args, so this is stricter (and
arguably more correct) input handling. The intended `./gitea bad-cmd`
rejection is unchanged.
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
## Summary
This fixes an OIDC sign-in edge case where a stale `external_login_user`
record can still point to an organization or a deleted user.
In that situation, Gitea may keep resolving the external login to the
wrong account during sign-in. For affected instances, this matches the
behavior reported in #36439 and #37812, where a user signing in with
OIDC/Entra ID could appear as an organization, or hit a 404 after that
organization was removed.
## What changed
- validate the user resolved from `external_login_user` during
OAuth2/OIDC login
- ignore stale links when the linked user no longer exists
- ignore stale links when the linked user is not an individual user
- remove the stale external login row so the sign-in flow can relink the
external account to the correct user
## Related
- Fixes#37812
- Related to #36439
---------
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Removes redundant/obsolete WebKit prefixes:
- `-webkit-mask-*` — duplicate the unprefixed `mask-*` siblings already
present; every supported browser handles unprefixed CSS Masking
longhands.
- `-webkit-overflow-scrolling: touch` — a no-op outside iOS Safari <13.
Browser floor (all support unprefixed `mask`): Chrome 120+, Safari
15.4+, Firefox 53+, and PaleMoon/Goanna (verified: unprefixed `mask`
longhands implemented unconditionally in UXP).
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
## Summary
This PR improves reusable workflow support for Gitea Actions. The
parsing of the called workflow now happens on Gitea side, not on the
runner. When the caller becomes ready, Gitea fetches the called workflow
source, parses it, and inserts each child job into the database as a
`ActionRunJob` linked to the caller via `ParentCallJobID`. As a result,
every callee job is dispatched as its own task and its logs surface as
an independent job entry in the UI, rather than being inlined into the
caller's "Set up job" step.
This PR supports two kinds of `uses` :
- same-repo call: `uses: ./.gitea/workflows/foo.yaml`
- cross-repo call: `uses: OWNER/REPO/.gitea/workflows/foo.yaml@REF`
## **⚠️ BREAKING ⚠️**
External reusable workflows (`uses:
https://other-gitea-instance/OWNER/REPO/.gitea/workflows/test.yaml@REF`)
are no longer supported. To keep using them, clone the repositories to
the local instance.
## Main changes
### Execution model
- Each caller job carries `IsReusableCaller=true` and won't be fetched
by runners.
- `ParentCallJobID` can link a called job to its caller.
- Caller status is derived from its direct children.
### Workflow syntax
- `jobparser` now supports parsing `on: workflow_call` trigger with
`inputs:`, `outputs:`, and `secrets:` declarations.
- **Max nesting depth**: capped at `MaxReusableCallLevels = 9`, which
means a top-level caller may have at most 9 nested callers below it.
- **Cycle prevention**: at expansion time, `checkCallerChain` walks the
caller's ancestor chain via `ParentCallJobID` and rejects if the same
`uses:` string appears anywhere upstream (`reusable workflow call cycle
detected`). This catches both direct (`A -> A`) and indirect (`A -> B ->
A`) cycles.
### Cross-repo access
- To share reusable workflows from private repos, use `Collaborative
Owners` introduced by #32562
### Rerun semantics
- `expandRerunJobIDs` partitions the latest attempt's jobs into:
- a **rerun set**: jobs being rerun + downstream siblings within the
same scope.
- an **ancestor set**: reusable callers whose only *some* descendants
are being rerun (the caller itself is not).
- Cloning behavior for callers in `execRerunPlan`:
- **Caller is fully rerun** (caller's `AttemptJobID` in `rerunSet`):
none of its descendants are cloned. The caller is cloned with
`IsCallerExpanded=false`, and re-expansion (which reinserts the children
fresh) happens later when the resolver brings the caller to `Waiting`
again.
- **Caller is in ancestor set** (only some descendants rerun): the
caller is pass-through (`Status` will be updated by its fresh children).
Its non-rerun descendants are also pass-through clones (point
`SourceTaskID` at the original task). Their `ParentCallJobID` is
remapped to the new attempt's caller row.
### UI
- Job list in `RepoActionView.vue` is now tree-shaped: callers indent
their children. Callers default to collapsed.
- New caller detail page using `WorkflowGraph` to show direct children
only; the run summary's `WorkflowGraph` shows top-level callers and
their immediate descendants.
### Known trade-offs
- **Caller expansion runs inside the enclosing write transaction.**
`expandReusableWorkflowCaller` performs a git read of the called
workflow while holding the row locks that update the caller and insert
its children. This is intentional: the caller-row update and child-row
inserts must commit atomically. None of the call sites is hot (each
caller is expanded once per attempt), so the trade-off is acceptable.
- **A malformed `if:` expression on a job leaves it `Blocked`
silently.** `evaluateJobIf` now runs server-side as part of resolver
passes; deterministic expression errors (typos, undefined context
fields) are logged but do not surface in the UI. This is the same
behavior the resolver already had for concurrency-expression errors.
Distinguishing transient DB errors from user-authored expression errors
and writing the latter back as `StatusFailure` is a follow-up.
#### Screenshots
<img width="1600" alt="image"
src="https://github.com/user-attachments/assets/bfaa9b7a-07e9-4127-8de9-a81f86e82828"
/>
<img width="1600" alt="image"
src="https://github.com/user-attachments/assets/8af109b3-ef28-4b53-aaad-d4632b923224"
/>
## References
-
https://docs.github.com/en/actions/how-tos/reuse-automations/reuse-workflows
-
https://docs.github.com/en/actions/reference/workflows-and-actions/reusing-workflow-configurations
---
Replace #36388
---------
Signed-off-by: Zettat123 <zettat123@gmail.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
At a 512m heap the CI Elasticsearch GC-thrashes under the jobs' memory
pressure and goes unresponsive, flaking `test-unit` (ES indexer tests
time out) and `test-mysql` (the ES-backed issue indexer blocks the
per-test queue flush). Raise the heap to 1g and disable ML + the startup
GeoIP download.
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
The Workflow Dependencies graph in the Actions run details view had
hard-coded English strings.
Also in projects view and contributors view I found some hard-coded
strings.
The other items in the issue #37787 (Summary / All jobs / Run Details /
Workflow file / Triggered via / Total duration) were already wired
through ctx.Locale.Tr; their translations just need to land in the
non-English locale_*.json files via the translation pipeline.
Fixes#37787
---------
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Adds bulk actions on the site-admin runner list
(`/-/admin/actions/runners`). Site admins can now select multiple
runners and **Delete**, **Disable**, or **Enable** them in one go
instead of clicking through each runner's edit page.
Scope is intentionally limited to the admin page. The user, org, and
repo runner pages keep their existing per-row UX — the shared list
template gates the bulk UI behind an `AllowBulkActions` flag set only by
the admin handler.
## Screenshots
<img width="1582" height="353"
src="https://github.com/user-attachments/assets/2125661f-aac0-4168-990a-97995a26abd2"
/>
---------
Signed-off-by: Nicolas <bircni@icloud.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Bump the pinned `giteabot` action to the
[`v1.0.3`](https://github.com/go-gitea/giteabot/releases/tag/v1.0.3)
release in both `giteabot.yml` and `giteabot-backport.yml`. v1.0.3 moves
label/state queries off the search API on top of the existing retry
logic.
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Use job.LoadAttributes (loads run + attributes) instead of an open-coded
load plus a dead nil-check, matching the concurrency path. Pre-size the
children slice to the known combination count.
Co-Authored-By: Claude (Opus 4.8) <noreply@anthropic.com>
A matrix built from a dependent job's runtime output (fromJson(needs.*.outputs.*))
has no inherent size bound, so a large or buggy upstream output could insert an
arbitrary number of jobs in one transaction. The per-attempt AttemptJobIDs also
back the job-index URLs, which must stay below MaxJobNumPerRun. Skip the
expansion with a clear reason when the resulting jobs would exceed that limit.
Also rename a test variable that tripped the misspell linter and clarify the
deferred-placeholder commit-status comment.
Co-Authored-By: Claude (Opus 4.8) <noreply@anthropic.com>
Expand a needs-dependent matrix through the new jobparser.ExpandMatrixWithNeeds,
which evaluates the raw matrix against the needs context via NewInterpeter the
same way EvaluateConcurrency does. This replaces synthesizing a workflow with
stub jobs and re-parsing it once the dependencies complete.
Parse now emits a single placeholder per needs-dependent matrix and the whole
matrix expands exactly once when the needs finish, instead of being split per
static value at planning time and re-expanded per placeholder. This removes
ExtractRawStrategies, HasMatrixWithNeeds and constructWorkflowWithNeeds.
Side effects of deferring the whole matrix: a static+dynamic matrix no longer
duplicates N times, needs-output detection no longer relies on string matching
(so bracket notation is handled), and a deferred placeholder gets no commit
status until it expands, avoiding an orphaned status from the name change.
Co-Authored-By: Claude (Opus 4.8) <noreply@anthropic.com>
- expanded matrix jobs inherit RunAttemptID, Attempt, AttemptJobID and
TokenPermissions, so they stay in the attempt and downstream needs/status
aggregation account for them
- reuse the placeholder as the first combination instead of leaving a skipped
phantom job that poisoned downstream `needs`
- jobparser only clones + pre-evaluates a matrix when it contains a `${{ }}`
expression; static matrices keep the original path
- handle the expanded-job marshal error, drop the unused mergeNeedsIntoVars,
gate and attempt-scope the post-resolve job reload, trim resolver logging
- migration: add IgnoreConstrains for mssql; revert the unrelated go.mod
golang.org/x/net downgrade
Co-Authored-By: Claude (Opus 4.8) <noreply@anthropic.com>
- When a commit subject is a bare URL, `linkProcessor` wrapped it in its
own `<a>` to that URL. Because HTML cannot nest anchors, the wrapping
default link (the action run / commit link) was lost and the action
title became unclickable — clicking it sent the user to the URL from the
commit message instead of the action log.
- Drop `linkProcessor` from `PostProcessCommitMessageSubject` so the
whole subject stays wrapped in the default link. URLs in subjects now
render as text inside that link; URLs in commit bodies are unaffected.
Fixes#37865
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Giteabot <teabot@gitea.io>
### Description
Replaces all remaining direct `gopkg.in/yaml.v3` imports with
`go.yaml.in/yaml/v4` across models, modules, routers, services, and
integration tests. `gopkg.in/yaml.v3` moves from a direct to an indirect
dependency in `go.mod`.
#### API compatibility
The yaml.Node type, node.Kind/node.Content traversal style
(modules/markup/markdown/convertyaml.go), and the
UnmarshalYAML(*yaml.Node) interface signature
(modules/optional/serialization.go) are all preserved in v4 — no
call-site changes were required beyond the import path.
**Related:**
- https://github.com/go-gitea/gitea/pull/36564#issuecomment-4526536805
---------
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
## Summary
Fixes#37528
This PR makes the workflow dispatch API reject workflows that do not
declare `workflow_dispatch`. Previously, `POST
/repos/{owner}/{repo}/actions/workflows/{workflow_id}/dispatches` could
create an `ActionRun` for a workflow that only declared another event
such as `push`.
The service now validates that the target workflow has a
`workflow_dispatch` trigger before inserting the run. The API maps that
validation failure to `422 Unprocessable Entity`, matching existing
validation failures in this handler.
The regression test creates a push-only workflow, dispatches it through
the public API, asserts the `workflow_dispatch` validation message, and
verifies that no run was inserted.
## Disclosure
Developed with assistance from OpenAI Codex.
---------
Co-authored-by: Nicolas <bircni@icloud.com>
## Background
`MAX_CREATION_LIMIT` applies to whoever owns a new repository, with no
distinction between individual users and organizations. Admins who want
different limits for the two - most commonly "block personal repos but
let orgs create freely" - currently have to set per-user / per-org
overrides on every entity.
## Changes
Adds two new `[repository]` settings:
- `USER_MAX_CREATION_LIMIT`: global limit for individual users
- `ORG_MAX_CREATION_LIMIT`: global limit for organizations
`MAX_CREATION_LIMIT` is kept as a shortcut: when set, it becomes the
default value for both new keys. When the new keys are explicitly
configured, they take precedence. Deployments that only set
`MAX_CREATION_LIMIT` see behavior identical to now.
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
In dev mode `/api/swagger` returned HTTP 500 (`Failed to locate local
path for managed asset URI: css/swagger.css`): the backend synthesised
asset keys from the Vite entry name instead of reading the manifest,
which only worked by coincidence and broke once a source file name
diverged from its entry name.
This keys the manifest by its source path (e.g. `web_src/js/index.ts`)
and resolves entries directly — hashed `file` in prod, dev-server source
in dev. A new `AssetCSSLinks` helper renders a JS entry's stylesheet
`<link>` tags from the manifest (the entry's CSS plus the CSS of its
statically-imported chunks).
Fixes: https://github.com/go-gitea/gitea/issues/37830
Fixes: https://github.com/go-gitea/gitea/pull/37832
Fixes: https://github.com/go-gitea/gitea/pull/37876
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: prakhar0x01 <prakharporwal2004@gmail.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Giteabot <teabot@gitea.io>
The Actions log parser only recognized `[command]`, so runner command
output emitted as `##[command] ...` was not shown in expanded step logs.
Add `##[command]` support to `LogLinePrefixCommandMap` in
`web_src/js/components/ActionRunView.ts` and cover it with a regression
test in `web_src/js/components/ActionRunView.test.ts`.
Changes
- Fixes Actions UI log rendering for runner command output
- Adds support for ##[command] in the Actions log parser
- Ensures runner echo ... lines are rendered when expanding step logs
- Includes a regression test covering ##[command] foo parsing
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
After posting a comment the page reloads via fetch-action. Clicking
"Close Issue" before the form re-initializes triggers a native form
submit, which navigates to the raw JSON redirect response
(`{"redirect":...}`) instead of the issue, so "Reopen Issue" never
appears and the test times out (observed on Firefox in CI).
Wait for the comment button to become disabled — which only happens once
the form re-initializes — before clicking "Close Issue".
---
This PR was written with the help of Claude Opus 4.7
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
- Update `github.com/alecthomas/chroma/v2` to `v2.25.0`.
- Migrate `github.com/dlclark/regexp2` to `/v2` (incorporates
https://github.com/go-gitea/gitea/pull/37664); drop the renovate pin.
- Replace the unmaintained `github.com/dimiro1/reply` (the last consumer
of `regexp2` v1 in our own code) with a small built-in reply parser for
incoming mail.
Signed-off-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Giteabot <teabot@gitea.io>
Co-authored-by: Nicolas <bircni@icloud.com>
- Expand NewInterpeter ownership comment to explicitly confirm that
run.Workflow is fresh per call and job is only read (not written),
so concurrent calls for different matrix combinations are race-free
- Downgrade the two remaining log.Info lines (matrix re-evaluation
complete / succeeded) to log.Debug per reviewer request
- Add unit tests for constructWorkflowWithNeeds (happy path, invalid
payload, missing jobs section), mergeNeedsIntoVars, and a
multi-dimension dynamic matrix case in TestHasMatrixWithNeeds
Co-Authored-By: Claude <claude-sonnet-4-5@anthropic.com>
- Fix double-expansion race: wrap claimMatrixExpansion (conditional UPDATE on
is_matrix_evaluated=false AND status=Blocked) + InsertActionRunJobs in a
single db.WithTx transaction; second concurrent caller gets n==0 and the
transaction rolls back without inserting duplicate jobs
- Remove matrix_cache.go, matrix_metrics.go, matrix_metrics_prometheus.go, and
matrix_metrics_test.go
- Replace HasMatrixWithNeeds string matching with a YAML tree walker that only
reports true when a ${{ ... }} expression block inside the matrix key
contains needs.<id>.outputs.<key>; avoids false positives on job names like
needs.review-runner that appear as plain YAML values
- Adapt the logging
Co-Authored-By: Claude <claude-sonnet-4-5@anthropic.com>
## Summary
- Add a Branch filter dropdown to the repo Actions run list web UI
- Wire `?branch=` query param through the web handler, matching the
existing REST API filter behavior
- Source the Branch dropdown from the indexed `branch` table (filtering
out deleted branches) instead of scanning `action_run.ref`, addressing
review feedback about unindexed columns
The Event filter was dropped after review: a static list of supported
events was noisy as UX, and querying distinct values from
`action_run.trigger_event` is slow because the column is not indexed.
`FindRunOptions.TriggerEvent` is kept for the REST API.
Closes#25042
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
The modernc SQLite driver (default since
https://github.com/go-gitea/gitea/pull/37562) returns `SQLITE_BUSY` once
the busy timeout is reached, unlike mattn which waited indefinitely.
`TestResourceIndex` fires many concurrent `NewIssue` writers, but SQLite
serializes all writers, so they queue on a single `BEGIN IMMEDIATE`
write lock. Under `-race` (modernc is much slower) the goroutines at the
back of the queue exceeded the hardcoded 5s test timeout, producing
`database is locked (5) (SQLITE_BUSY)`.
Changes:
- Reduce the concurrent inserts from 25 to 10. Since SQLite serializes
writers, the extra goroutines only deepen the busy-lock queue without
adding coverage. 10 still exercises concurrent index allocation while
cutting the test's `-race` runtime ~3x (2.76s to 0.86s locally).
- Share the busy-timeout constant: export `DefaultSQLiteBusyTimeout`
(20s, the production default) and reference it from the test engine
instead of the hardcoded `5000`.
Observed flake:
https://github.com/go-gitea/gitea/actions/runs/26394082930/job/77690496092
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
GiteabotandGitHub
