# Tekton v1.12.0

2026-05-04 · [Original release notes](https://github.com/tektoncd/pipeline/releases/tag/v1.12.0) · [Full summary](https://ratatosk.io/en/releases/tekton/v1.12.0)

Security 6 (high 1) · Breaking changes 5 · Deprecations 1 · Dependency updates 9 · Other changes 14

## Security (6)

### [high] Git resolver argument injection fixed

- Affects: setups using the deprecated or current git resolver
- [GHSA-94jr-7pqp-xhcq](https://github.com/tektoncd/pipeline/security/advisories/GHSA-94jr-7pqp-xhcq)

In versions 1.10.0 through 1.11.0, the git resolver passed `revision` directly to `git fetch` without checking for a leading `-`, allowing argument injection.

> * :lock: fix: prevent git argument injection via revision parameter (GHSA-94jr-7pqp-xhcq) (0967cc8e6e9b)

### VolumeMount restriction check now uses normalized paths

- Affects: setups with VolumeMount paths checked against `/tekton/`

> * :lock: fix: normalize VolumeMount paths before /tekton/ restriction check (c8cc73a2ceca)

### System API tokens rejected with a user-controlled `serverURL`

- Affects: setups using a system API token with a user-controlled `serverURL`

> * :lock: Security: reject system API token with user-controlled serverURL (e964bce1a130)

### HTTP resolver response size limited

- Affects: setups using the HTTP resolver

> * :lock: fix: limit HTTP resolver response body size to prevent OOM DoS (d4ba8cae9edf)

### Resolver prefix stripping and non-capturing pattern anchoring

- Affects: resolver setups using prefixes and pattern anchoring (Inferred)

> * :lock: fix: strip resolver prefixes and use non-capturing group for pattern anchoring (b8905600322a)

### Source URI whitespace trimming before pattern matching

- Affects: setups using source-URI pattern matching (Inferred)

> * :lock: fix: trim whitespace from source URI before pattern matching (37f9657b4bd1)

## Breaking changes (5)

### CustomRun CloudEvents sent by default

- Affects: CustomRuns with a CloudEvents sink configured in `config-events`
- Removed in: not stated

Previously, CustomRun CloudEvents were not sent unless `send-cloudevents-for-runs` was set to `true`. They are now sent when a sink is configured in `config-events`, unless `send-cloudevents-for-runs` is set to `false`.

Evidence: “The feature flag send-cloudevents-for-runs is deprecated and will be removed in a future release. Its default has changed from false to true: CloudEvents for CustomRuns are now sent by default whenever a sink is configured in config-events.” https://github.com/tektoncd/pipeline/pull/9774

> * :sparkles: [TEP-0137] Deprecate send-cloudevents-for-runs feature flag (#9774)
> The feature flag default changed from `false` to `true`. CloudEvents for CustomRuns are now sent by default.
> * :rotating_light: [TEP-0137] The `send-cloudevents-for-runs` feature flag is deprecated and will be removed in a future release. Its default has changed from `false` to `true` (#9774).

### Specific failure reasons for pod-level TaskRun failures

- Affects: TaskRun `Succeeded` conditions

Previously, all pod-level TaskRun failures used the generic `Failed` reason. Now, the reason identifies pod eviction (`PodEvicted`), init-container OOM or failure (`InitContainerOOM`, `InitContainerFailed`), step OOM or failure (`StepOOM`, `StepFailed`), or sidecar OOM or failure (`SidecarOOM`, `SidecarFailed`). Unknown failures still use `Failed`; when multiple failures occur, the reason follows this priority: pod eviction, init-container failures, sidecar OOM, step OOM, sidecar failure, then step failure.

Evidence: “When multiple containers fail, the reason is chosen by priority: 1. Pod eviction (pod-level, authoritative) 2. Init container failures (run first, block everything) 3. Sidecar OOM → Step OOM (OOM is likely root cause) 4. Sidecar failed → Step failed (non-OOM failures)” https://github.com/tektoncd/pipeline/pull/9368

> * :sparkles: fix: surface specific TaskRun failure reasons when pod fails (#9368)
> TaskRun failure reasons now distinguish between different pod-level failure types: PodEvicted, InitContainerOOM, InitContainerFailed, StepOOM, StepFailed, SidecarOOM, and SidecarFailed, replacing the generic "Failed" reason.

### Duplicate CloudEvents suppressed for run objects

- Affects: CloudEvents sinks configured for run objects
- Action required: Controller concurrency set to 1 so reconcile cycles for the same resource do not run concurrently.

Before this release, `PipelineRun` and `TaskRun` events bypassed the event cache. Repeated condition snapshots are now skipped, and a previously sent event type stays suppressed even after condition changes; `.unknown.` events can still be sent on condition changes, including message-only changes.

Evidence: “// A cache miss requires a new object or change in the condition (including message-only changes)” https://github.com/tektoncd/pipeline/pull/6914

> * :bug: Rework the events controller cache (#6914)

### CloudEvents delivery moved to a dedicated controller

- Affects: setups using CloudEvents for PipelineRuns or TaskRuns
- Action required: the `tekton-events-controller` Deployment must be running

Previously, the PipelineRun and TaskRun controllers sent these CloudEvents; now the dedicated controller sends them, and two queued event types are available. `kubectl describe pipelinerun/taskrun` shows delivery status through `CloudEventSent` and `CloudEventFailed` Kubernetes Events.

> ### :rotating_light: TEP-0137: CloudEvents now sent by dedicated events controller (ACTION REQUIRED)
> CloudEvents for **PipelineRuns** and **TaskRuns** are now sent by the dedicated `tekton-events-controller` and **no longer by the PipelineRun/TaskRun controllers**. Operators must ensure the `tekton-events-controller` Deployment is running.
> New events added:
> - `dev.tekton.event.pipelinerun.queued.v1` — sent when a PipelineRun is created but not yet processed
> - `dev.tekton.event.taskrun.queued.v1` — sent when a TaskRun is created but not yet processed
> CloudEvent delivery visibility is available via `kubectl describe pipelinerun/taskrun` (CloudEventSent/CloudEventFailed k8s Events).

### PipelineRun missing-result messages identify the task and result

- Affects: PipelineRuns with downstream tasks that consume task results

When a completed successful task omits a result referenced by a downstream task, the old message reported only a count; it now names the task, missing result, and downstream task, with details capped at 1024 characters. If the referenced task failed rather than succeeded, its original error message remains unchanged.

Evidence: “If a referenced result is missing from a completed (successful) task, the error message now states which task completed without emitting which result and which downstream task references it. The details are included in the PipelineRun status condition message (capped at 1024 characters). When the referenced task failed (rather than succeeded), the original error message is preserved unchanged.” https://github.com/tektoncd/pipeline/pull/9529

> * :sparkles: fix: surface clear errors when completed tasks miss referenced results (#9529)

## Deprecations (1)

### `send-cloudevents-for-runs` flag deprecated

- Affects: CustomRuns with a sink configured in `config-events`
- Removed in: not stated

Previously, CloudEvents for CustomRuns were not sent by default because the flag defaulted to `false`. With no setting change, they are now sent whenever a sink is configured in `config-events`.

> ### :rotating_light: TEP-0137: `send-cloudevents-for-runs` feature flag deprecated
> The `send-cloudevents-for-runs` feature flag is deprecated and will be removed in a future release. Its default has changed from `false` to `true`: CloudEvents for CustomRuns are now sent by default whenever a sink is configured in `config-events`.
> Users relying on the flag set to `false` to skip events for CustomRuns should be aware of this change.

## Dependency updates (9)

9 more, none mentioning security

- `* :hammer: Remove aws-sdk-go-v2 ECR replace directives (#9773)`
- `* :hammer: build(deps): bump k8s.io dependencies from 0.35.2 to 0.35.4 (#9848)`
- `* :hammer: build(deps): bump github.com/google/cel-go from 0.27.0 to 0.28.0 (#9824)`
- `* :hammer: build(deps): bump google.golang.org/grpc from 1.79.3 to 1.80.0 (#9715)`
- `* :hammer: build(deps): bump go.uber.org/zap from 1.27.1 to 1.28.0 (#9926)`
- `* :hammer: build(deps): bump go.opentelemetry.io/otel/sdk from 1.42.0 to 1.43.0 (#9757)`
- `* :hammer: build(deps): bump github.com/go-jose/go-jose/v3 from 3.0.4 to 3.0.5 (#9722)`
- `* :hammer: build(deps): bump github.com/go-jose/go-jose/v4 from 4.1.3 to 4.1.4 (#9721)`
- `* :hammer: build(deps): bump github.com/sigstore/sigstore from 1.10.4 to 1.10.5 (#9711)`

## Other changes (14)

- Kubernetes Warning events for task result-validation failures
- Completed PipelineRun reconciliation skips repeated work
- TaskRun cleanup no longer applies defaults
- PipelineRun and TaskRun map comparisons switched to `maps.Equal`
- Events controller cache switches to bigcache and hashed keys
- Release pipeline creates categorized GitHub release drafts
- CI workflow replaces `tj-actions/changed-files` with `git diff`
- Automated Dependabot configuration for LTS release branches
- Table-driven structure removed from `TestEmit` test
- End-to-end test for the `TaskRun` pending lifecycle added
- Generated fake clientsets updated for `k8s.io/code-generator` 0.35.3
- Bundle resolver option defaults and requiredness documented
- Metrics documentation updated for the OpenCensus-to-OpenTelemetry migration
- Examples README updated with v1 apply paths

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
