Build a new binary from current branch and copy it to the tests/systemtests/binaries
folder by running system tests.
In project root:
make test-system
Or via manual steps
make build
mkdir -p ./tests/systemtests/binaries
cp ./build/simd ./tests/systemtests/binaries/
Switch to the tests/systemtests
folder to work from here.
If there is no test file matching your use case, start a new test file here.
for example bank_test.go
to begin with:
//go:build system_test
package systemtests
import (
"testing"
)
func TestQueryTotalSupply(t *testing.T) {
sut.ResetChain(t)
sut.StartChain(t)
cli := NewCLIWrapper(t, sut, verbose)
raw := cli.CustomQuery("q", "bank", "total-supply")
t.Log("### got: " + raw)
}
The file begins with a Go build tag to exclude it from regular go test runs.
All tests in the systemtests
folder build upon the test runner initialized in main_test.go
.
This gives you a multi node chain started on your box.
It is a good practice to reset state in the beginning so that you have a stable base.
The system tests framework comes with a CLI wrapper that makes it easier to interact or parse results.
In this example we want to execute simd q bank total-supply --output json --node tcp://localhost:26657
which queries
the bank module.
Then print the result to for the next steps
go test -mod=readonly -tags='system_test' -v ./... --run TestQueryTotalSupply --verbose
This give very verbose output. You would see all simd CLI commands used for starting the server or by the client to interact. In the example code, we just log the output. Watch out for
bank_test.go:15: ### got: {
"supply": [
{
"denom": "stake",
"amount": "2000000190"
},
{
"denom": "testtoken",
"amount": "4000000000"
}
],
"pagination": {
"total": "2"
}
}
At the end is a tail from the server log printed. This can sometimes be handy when debugging issues.
- Passing
--nodes-count=1
overwrites the default node count and can speed up your test for local runs
When we have a json response, the gjson lib can shine. It comes with jquery like syntax that makes it easy to navigation within the document.
For example gjson.Get(raw, "supply").Array()
gives us all the childs to supply
as an array.
Or gjson.Get("supply.#(denom==stake).amount").Int()
for the amount of the stake token as int64 type.
In order to test our assumptions in the system test, we modify the code to use gjson
to fetch the data:
raw := cli.CustomQuery("q", "bank", "total-supply")
exp := map[string]int64{
"stake": int64(500000000 * sut.nodesCount),
"testtoken": int64(1000000000 * sut.nodesCount),
}
require.Len(t, gjson.Get(raw, "supply").Array(), len(exp), raw)
for k, v := range exp {
got := gjson.Get(raw, fmt.Sprintf("supply.#(denom==%q).amount", k)).Int()
assert.Equal(t, v, got, raw)
}
The assumption on the staking token usually fails due to inflation minted on the staking token. Let's fix this in the next step
go test -mod=readonly -tags='system_test' -v ./... --run TestQueryTotalSupply --verbose
- Putting the
raw
json response to the assert/require statements helps with debugging on failures. You are usually lacking context when you look at the values only.
First step is to disable inflation. This can be done via the ModifyGenesisJSON
helper. But to add some complexity,
we also introduce a new token and update the balance of the account for key node0
.
The setup code looks quite big and unreadable now. Usually a good time to think about extracting helper functions for
common operations. The genesis_io.go
file contains some examples already. I would skip this and take this to showcase the mix
of gjson
, sjson
and stdlib json operations.
sut.ResetChain(t)
cli := NewCLIWrapper(t, sut, verbose)
sut.ModifyGenesisJSON(t, func(genesis []byte) []byte {
// disable inflation
genesis, err := sjson.SetRawBytes(genesis, "app_state.mint.minter.inflation", []byte(`"0.000000000000000000"`))
require.NoError(t, err)
// add new token to supply
var supply []json.RawMessage
rawSupply := gjson.Get(string(genesis), "app_state.bank.supply").String()
require.NoError(t, json.Unmarshal([]byte(rawSupply), &supply))
supply = append(supply, json.RawMessage(`{"denom": "mytoken","amount": "1000000"}`))
newSupply, err := json.Marshal(supply)
require.NoError(t, err)
genesis, err = sjson.SetRawBytes(genesis, "app_state.bank.supply", newSupply)
require.NoError(t, err)
// add amount to any balance
anyAddr := cli.GetKeyAddr("node0")
newBalances := GetGenesisBalance(genesis, anyAddr).Add(sdk.NewInt64Coin("mytoken", 1000000))
newBalancesBz, err := newBalances.MarshalJSON()
require.NoError(t, err)
newState, err := sjson.SetRawBytes(genesis, fmt.Sprintf("app_state.bank.balances.#[address==%q]#.coins", anyAddr), newBalancesBz)
require.NoError(t, err)
return newState
})
sut.StartChain(t)
Next step is to add the new token to the assert map. But we can also make it more resilient to different node counts.
exp := map[string]int64{
"stake": int64(500000000 * sut.nodesCount),
"testtoken": int64(1000000000 * sut.nodesCount),
"mytoken": 1000000,
}
go test -mod=readonly -tags='system_test' -v ./... --run TestQueryTotalSupply --verbose --nodes-count=1
Complexer workflows and tests require modifying state on a running chain. This works only with builtin logic and operations.
If we want to burn some of our new tokens, we need to submit a bank burn message to do this.
The CLI wrapper works similar to the query. Just pass the parameters. It uses the node0
key as default:
// and when
txHash := cli.Run("tx", "bank", "burn", "node0", "400000mytoken")
RequireTxSuccess(t, txHash)
RequireTxSuccess
or RequireTxFailure
can be used to ensure the expected result of the operation.
Next, check that the changes are applied.
exp["mytoken"] = 600_000 // update expected state
raw = cli.CustomQuery("q", "bank", "total-supply")
for k, v := range exp {
got := gjson.Get(raw, fmt.Sprintf("supply.#(denom==%q).amount", k)).Int()
assert.Equal(t, v, got, raw)
}
assert.Equal(t, int64(600_000), cli.QueryBalance(cli.GetKeyAddr("node0"), "mytoken"))
While tests are still more or less readable, it can gets harder the longer they are. I found it helpful to add some comments at the beginning to describe what the intention is. For example:
// scenario:
// given a chain with a custom token on genesis
// when an amount is burned
// then this is reflected in the total supply