Skip to content

Adding Span Link support for compound header tag extractions with invalid traces #4895

Adding Span Link support for compound header tag extractions with invalid traces

Adding Span Link support for compound header tag extractions with invalid traces #4895

Workflow file for this run

name: Smoke Tests
on:
workflow_call: # allows to reuse this workflow
inputs:
ref:
description: The branch or tag to run the workflow on
required: true
type: string
go-libddwaf-ref:
description: A git ref to update github.com/DataDog/go-libddwaf/v3 to. No-op if empty.
required: false
type: string
push:
branches:
- main
- release-v*
tags:
- '**'
schedule: # nightly
- cron: "0 0 * * *"
workflow_dispatch: {} # manually
pull_request:
branches:
- '**'
env:
TEST_RESULTS: /tmp/test-results # path to where test results will be saved
permissions:
contents: read
jobs:
go-get-u:
# Run go get -u to upgrade dd-trace-go dependencies to their
# latest minor version and see if dd-trace-go still compiles.
# Related to issue https://github.com/DataDog/dd-trace-go/issues/1607
name: 'go get -u smoke test'
runs-on: ubuntu-latest
if: github.repository_owner == 'DataDog' # only run on DataDog's repository, not in forks
env:
PACKAGES: ./internal/... ./ddtrace/... ./profiler/... ./appsec/...
steps:
- uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
# Manually specify the repository, which is necessary in the workflow_call situation, as the default is
# otherwise the repository where the caller workflow started from. In this case, we need to check out the
# repository where the called workflow is (i.e, this repository); but I don't know of a more elegant way to
# obtain its name than hard-coding it.
repository: DataDog/dd-trace-go
- uses: actions/setup-go@v3
with:
go-version: "stable"
cache: true
- name: go get -u
run: |-
go get -u -t $PACKAGES
go mod tidy
- name: Install requested go-libddwaf version
if: github.event_name == 'workflow_call' && inputs.go-libddwaf-ref != ''
run: |-
go get -u -t github.com/DataDog/go-libddwaf/v3@${{ inputs.go-libddwaf-ref }}
go mod tidy
- name: Compile dd-trace-go
run: go build $PACKAGES
- name: Test dd-trace-go
env:
DD_APPSEC_WAF_TIMEOUT: 1h
run: go test $PACKAGES
go-mod-tidy:
# Run go mod tidy to ensure that all go.mod and go.sum files are up-to-date.
name: 'go mod tidy smoke test'
runs-on: ubuntu-latest
env:
# Users may build our library with GOTOOLCHAIN=local. If they do, and our
# go.mod file specifies a newer Go version than their local toolchain, their
# build will break. Run our tests with GOTOOLCHAIN=local to ensure that
# our library builds with all of the Go versions we claim to support,
# without having to download a newer one.
GOTOOLCHAIN: local
steps:
- uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
- uses: actions/setup-go@v3
with:
go-version: "1.22"
cache: true
- name: go mod tidy
run: |-
find -name go.mod -execdir go mod tidy \;
git diff --exit-code
# TODO: macos setup requirements (xcode tools installation, etc.)
setup-requirements-linux:
# Build and deployment setup smoke test of linux containers built from the
# golang docker image to test that dd-trace-go doesn't need more than the
# "out-of-the-box" images. It is expected require a few more tools when CGO
# is enabled, but nothing more than gcc and the C library, but nothing more.
# Anything more than this "standard Go build and deployment requirements"
# must be considered breaking changes.
name: 'Build and deployment requirements smoke tests'
runs-on: ubuntu-latest
if: github.repository_owner == 'DataDog' # only run on DataDog's repository, not in forks
strategy:
fail-fast: false
matrix:
# TODO: cross-compilation from/to different hardware architectures once
# github provides native ARM runners.
go: [ "1.22", "1.23" ]
build-env: [ alpine, bookworm, bullseye ]
build-with-cgo: [ 0, 1 ]
deployment-env: [ alpine, debian11, debian12, al2, al2023, busybox, scratch ]
include:
# GitHub limits the number of matrix jobs to 256, so we need to reduce
# it a bit, and we can reduce redundant tests.
# 1. Building with `go mod vendoring` is not worth it on all the
# possible build and deployment envs.
- build-env: alpine
build-with-vendoring: y
build-with-cgo: 1 # cgo's build tag can impact the vendored files
deployment-env: alpine
- build-env: alpine
build-with-vendoring: y
build-with-cgo: 0 # cgo's build tag can impact the vendored files
deployment-env: alpine
# 2. Given the low blast radius of the busybox deployment environment
# this is the only one where we accept the libdl.so.2 requirement.
# For this reason, we add the datadog.no_waf build tag in this only
# case to avoid the libdl.so.2 dependency.
- deployment-env: busybox
build-with-cgo: 1
build-tags: "datadog.no_waf"
exclude:
# Exclude "out of the box" cases requiring extra setup:
# 1. Building with CGO enabled on alpine but deploying to a non-alpine
# environment: the C library isn't located at the same place.
- build-env: alpine
build-with-cgo: 1
deployment-env: debian11
- build-env: alpine
build-with-cgo: 1
deployment-env: debian12
- build-env: alpine
build-with-cgo: 1
deployment-env: al2
- build-env: alpine
build-with-cgo: 1
deployment-env: al2023
- build-env: alpine
build-with-cgo: 1
deployment-env: busybox
- build-env: alpine
build-with-cgo: 1
deployment-env: scratch
# 2. Too old glibc on the deployment environment than on the build env
- build-env: bookworm
deployment-env: al2
- build-env: bookworm
deployment-env: debian11
# 3. Build with CGO enabled and deploying to a scratch/busybox docker
# image requires copying the dynamic lib dependencies (full example
# provided at https://github.com/DataDog/appsec-go-test-app/blob/main/examples/docker/scratch/Dockerfile)
- build-with-cgo: 1
deployment-env: scratch
steps:
- uses: actions/checkout@v4
with:
ref: ${{ inputs.ref || github.ref }}
# Manually specify the repository, which is necessary in the workflow_call situation, as the default is
# otherwise the repository where the caller workflow started from. In this case, we need to check out the
# repository where the called workflow is (i.e, this repository); but I don't know of a more elegant way to
# obtain its name than hard-coding it.
repository: DataDog/dd-trace-go
- uses: docker/setup-buildx-action@v3
- name: Build
uses: docker/build-push-action@v5
with:
context: .
file: ./internal/setup-smoke-test/Dockerfile
push: false
load: true
tags: smoke-test
build-args: |
go=${{ matrix.go }}
build_env=${{ matrix.build-env }}
build_tags=${{ matrix.build-tags }}
build_with_vendoring=${{ matrix.build-with-vendoring }}
build_with_cgo=${{ matrix.build-with-cgo }}
deployment_env=${{ matrix.deployment-env }}
go_libddwaf_ref=${{ inputs.go-libddwaf-ref }}
- name: Test
run: docker run -p7777:7777 --rm smoke-test