ci: First run at interchaintests. (backport #3205)

.github/workflows/docker-push.yml

@@ -16,7 +16,7 @@ jobs:
   build-and-push-image:
     runs-on: ubuntu-latest
     permissions:
-      contents: read
+      contents: write
       packages: write
 
     steps:
@@ -54,3 +54,14 @@ jobs:
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}
         if: ${{ env.NEW_COMMIT_COUNT > 0 }}
+
+      - name: Dispatch Interchain test
+        uses: peter-evans/repository-dispatch@v3
+        with:
+          event-type: image-pushed
+          client-payload: |
+            {
+              "tag_name": "${{ fromJson(steps.meta.outputs.json)['labels']['org.opencontainers.image.version'] }}",
+              "ref_name": "${{ fromJson(steps.meta.outputs.json)['labels']['org.opencontainers.image.revision'] }}"
+            }
+        if: ${{ env.NEW_COMMIT_COUNT > 0 }}

.github/workflows/interchain-test.yml

@@ -0,0 +1,64 @@
+name: Interchain Test
+on:
+  repository_dispatch:
+    types: [image-pushed]
+  workflow_dispatch:
+
+jobs:
+    prepare-matrix:
+        runs-on: ubuntu-latest
+        steps:
+            # We need to figure out a) a ref to clone based on a docker image, and b) which tag to test
+            # If the event is a registry_package, this comes from the pushed image; for a workflow_dispatch, it's the branch/tag that the user supplied
+            - name: Get metadata
+              id: get-metadata
+              run: |
+                if [[ "${{ github.event_name }}" == 'repository_dispatch' ]]; then
+                  echo "ref_name=${{ github.event.client_payload.ref_name }}" | tee -a $GITHUB_OUTPUT
+                  echo "tag_name=${{ github.event.client_payload.tag_name }}" | tee -a $GITHUB_OUTPUT
+                else
+                  echo "ref_name=${{ github.ref_name }}" | tee -a $GITHUB_OUTPUT
+                  echo "tag_name=${{ github.ref_name }}" | tee -a $GITHUB_OUTPUT
+                fi
+            - name: Check out repository code
+              uses: actions/checkout@v3
+              with:
+                ref: ${{ steps.get-metadata.outputs.ref_name }}
+            - name: Setup go
+              uses: actions/setup-go@v5
+            - name: Prepare matrix
+              id: generate-matrix
+              run: |
+                cd ./tests/interchain
+                echo "matrix=$(go run ./matrix_tool/main.go ${{ steps.get-metadata.outputs.tag_name }})" | tee -a $GITHUB_OUTPUT
+        outputs:
+            matrix: ${{ steps.generate-matrix.outputs.matrix }}
+            ref_name: ${{ steps.get-metadata.outputs.ref_name }}
+    test:
+        needs: prepare-matrix
+        runs-on: ubuntu-latest
+        name: "${{ matrix.previous_version }} -> ${{ matrix.test_version }} test ${{ matrix.test_name }}"
+        strategy:
+            matrix:
+                ${{fromJson(needs.prepare-matrix.outputs.matrix)}}
+            fail-fast: false
+            max-parallel: 10
+        steps:
+            - name: Check out repository code
+              uses: actions/checkout@v3
+              with:
+                ref: ${{ needs.prepare-matrix.outputs.ref_name }}
+            - name: Setup go
+              uses: actions/setup-go@v5
+            - name: Run test
+              env:
+                TEST_DOCKER_REGISTRY: "ghcr.io/${{ github.repository_owner }}"
+                TEST_OLD_GAIA_IMAGE_VERSION: "${{ matrix.previous_version }}"
+                TEST_NEW_GAIA_IMAGE_VERSION: "${{ matrix.test_version }}"
+                TEST_UPGRADE_NAME: "${{ matrix.upgrade_name }}"
+              run: |
+                # This docker pull/tag is a quick hack only necessary for v19, since there were no official v18 images built.
+                # Once we're testing 19 -> 20 this can be removed
+                docker pull "ghcr.io/hyphacoop/gaia:v18.1.0" && docker tag "ghcr.io/hyphacoop/gaia:v18.1.0" "ghcr.io/${{ github.repository_owner }}/gaia:v18.1.0"
+                cd ./tests/interchain
+                go test -v ./... -failfast -p 1 -timeout 5h -run="^${{ matrix.test_name }}"

.github/workflows/test.yml

@@ -46,7 +46,7 @@ jobs:
       - name: test & coverage report creation
         if: env.GIT_DIFF
         run: |
-          go test -v -coverprofile=profile.out -covermode=atomic -coverpkg=./... $(go list ./... | grep -v -e '/tests/e2e')
+          go test -v -coverprofile=profile.out -covermode=atomic -coverpkg=./... $(go list ./... | grep -v -e '/tests/e2e' | grep -v -e '/tests/interchain')
       - uses: actions/upload-artifact@v4
         if: env.GIT_DIFF
         with:

Dockerfile

@@ -24,7 +24,7 @@ RUN echo "Ensuring binary is statically linked ..."  \
     && file /src/app/build/gaiad | grep "statically linked"
 
 FROM alpine:$IMG_TAG
-RUN apk add --no-cache build-base
+RUN apk add --no-cache build-base jq
 RUN addgroup -g 1025 nonroot
 RUN adduser -D nonroot -u 1025 -G nonroot
 ARG IMG_TAG

tests/interchain/README.md

@@ -0,0 +1,139 @@
+# Interchain tests for gaia.
+
+
+These tests use [interchaintest](https://github.com/strangelove-ventures/interchaintest/) to
+create, upgrade, and test chains.
+
+They dockerize the validators, so they depend on a `gaia` docker image being built.
+You can build a docker image using the [docker-push](../../.github/workflows/docker-push.yml) workflow.
+`docker-push` runs nightly on the `main` branch, and for all new releases, but you can also
+[run it manually on any branch](https://github.com/cosmos/gaia/actions/workflows/docker-push.yml)
+
+Once the `gaia` image is built, the `docker-push` action workflow automatically invoke the
+[interchain-test](../../.github/workflows/interchain-test.yml) workflow.
+
+Read on to learn how these tests work.
+
+## Upgrade testing
+
+The tests will make sure it's possible to upgrade from a previous version of
+`gaia` to the current version being tested. It does so by starting a chain from genesis
+on the previous version, then upgrading it to the current version.
+
+## Version selection
+
+The `interchain-test` workflow will start by selecting versions to test upgrading from.
+
+The [`matrix_tool`](./matrix_tool/main.go) tool will take the tag of the image
+being tested (e.g. `v18.0.0`, or `main`, or `some-feature-branch`), and figure
+out a corresponding semver. If the tag is already a valid semver, that's the
+version. Otherwise, it will take the major version from the module line in `go.mod`,
+and append `.999.0`. Given that semver, it'll figure out:
+
+* The previous rc (if the current version is itself an rc)
+* The previous minor version (if applicable)
+* The previous major version
+
+For instance, for `v15.1.0-rc1`, we'll test upgrading from:
+* `v15.1.0-rc0`
+* `v15.0.0`
+* `v14.2.0`
+
+The workflow will then test upgrading from each of those three to the current
+version. When it's a major upgrade, it will do so via governance proposal,
+otherwise it'll simply stop the old image and start the new one.
+
+## Test Suites
+
+Each of the *_test.go files in this directory contains a test suite.  These
+share some common scaffolding (a `SetupSuite`) to create and upgrade a chain,
+and then run a set of tests on that chain.
+
+So, for instance, a transactions suite:
+
+```go
+type TxSuite struct {
+	*chainsuite.Suite
+}
+```
+
+It extends `chainsuite.Suite,` so its SetupSuite will create and upgrade a
+chain (more on this later). The individual `Test*` methods then run the gaia
+version being tested:
+
+```go
+func (s *TxSuite) TestBankSend() {
+	balanceBefore, err := s.Chain.GetBalance(s.GetContext(), s.Chain.ValidatorWallets[1].Address, chainsuite.Uatom)
+	s.Require().NoError(err)
+
+	_, err = s.Chain.Validators[0].ExecTx(
+		s.GetContext(),
+		s.Chain.ValidatorWallets[0].Moniker,
+		"bank", "send",
+		s.Chain.ValidatorWallets[0].Address, s.Chain.ValidatorWallets[1].Address, txAmountUatom(),
+	)
+	s.Require().NoError(err)
+
+	balanceAfter, err := s.Chain.GetBalance(s.GetContext(), s.Chain.ValidatorWallets[1].Address, chainsuite.Uatom)
+	s.Require().NoError(err)
+	s.Require().Equal(balanceBefore.Add(sdkmath.NewInt(txAmount)), balanceAfter)
+}
+```
+
+Because of how testify works, we have to instantiate each suite to run it.
+This is also where we tell the suite to run an upgrade on Setup:
+
+```go
+func TestTransactions(t *testing.T) {
+	txSuite := TxSuite{chainsuite.NewSuite(chainsuite.SuiteConfig{UpgradeOnSetup: true})}
+	suite.Run(t, &txSuite)
+}
+```
+
+Of course, we can also parameterize the test suites themselves. This enables us
+to write tests once and run them a bunch of times on different configurations:
+
+```go
+type ConsumerLaunchSuite struct {
+	*chainsuite.Suite
+	OtherChain            string
+	OtherChainVersion     string
+	ShouldCopyProviderKey [chainsuite.ValidatorCount]bool
+}
+
+func TestICS40ChainLaunch(t *testing.T) {
+	s := &ConsumerLaunchSuite{
+		Suite:                 chainsuite.NewSuite(chainsuite.SuiteConfig{}),
+		OtherChain:            "ics-consumer",
+		OtherChainVersion:     "v4.0.0",
+		ShouldCopyProviderKey: noProviderKeysCopied(),
+	}
+	suite.Run(t, s)
+}
+
+func TestICS33ConsumerAllKeysChainLaunch(t *testing.T) {
+	s := &ConsumerLaunchSuite{
+		Suite:                 chainsuite.NewSuite(chainsuite.SuiteConfig{}),
+		OtherChain:            "ics-consumer",
+		OtherChainVersion:     "v3.3.0",
+		ShouldCopyProviderKey: allProviderKeysCopied(),
+	}
+	suite.Run(t, s)
+}
+```
+
+Notice also how `UpgradeOnSetup` isn't set here: the ConsumerLaunchSuite needs
+to be handed a pre-upgrade chain so it can make sure that a consumer chain that
+launched before the upgrade keeps working after the upgrade.
+
+
+## Writing new tests
+
+All you need to start writing new tests is a test suite as described above.
+The suite will have an `s.Chain` that you can test. Check out utilities in
+[`chainsuite/chain.go`](./chainsuite/chain.go) and
+[`chain_ics.go`](./chainsuite/chain_ics.go) for some convenience methods.
+
+In addition, the s.Chain object extends the `interchaintest` chain object, so
+check out [the docs](https://pkg.go.dev/github.com/strangelove-ventures/interchaintest/v7) to
+see what else is available.