TestTxnPanics need to cancel testcase from sending error to error channel

[Mskxn]

What happened?

This test has 6 testcases and run them one by one, with a select stmt to handle the timeout, but timeout can occurs no because the testcase is not panic, but is blocking for some reason.When this occurs, the errors in errc is not in the order as expected. Can we cancel the errc <- s.(string) in df when timeout?

etcd/client/v3/txn_test.go

Lines 95 to 105 in 0bdc660

	for i, tt := range tests {
	go tt.f()
	select {
	case err := <-errc:
	if err != tt.err {
	t.Errorf("#%d: got %s, wanted %s", i, err, tt.err)
	}
	case <-time.After(time.Second):
	t.Errorf("#%d: did not panic, wanted panic %s", i, tt.err)
	}
	}

What did you expect to happen?

If the testcase is failed because of timeout, do not send error to errc, so the remaining testcases can work as expected.

How can we reproduce it (as minimally and precisely as possible)?

Add some delay to any of these testcases to make the timeout error occurs.

Anything else we need to know?

No response

Etcd version (please run commands below)

Etcd configuration (command line flags or environment variables)

Etcd debug information (please run commands below, feel free to obfuscate the IP address or FQDN in the output)

Relevant log output

bug report:

txn_test.go:103: #0: did not panic, wanted panic cannot call If twice!
txn_test.go:103: #1: did not panic, wanted panic cannot call If after Then!
txn_test.go:103: #2: did not panic, wanted panic cannot call If after Else!
txn_test.go:103: #3: did not panic, wanted panic cannot call Then twice!
txn_test.go:103: #4: did not panic, wanted panic cannot call Then after Else!
txn_test.go:100: #5: got cannot call If twice!, wanted cannot call Else twice!

[lavacat]

Tried adding a delay and got failure as expected. @Mskxn am I reproducing not as you've intended?

make test-unit
...
? go.etcd.io/etcd/client/pkg/v3/verify [no test files]
% (cd client/internal/v2 && 'env' 'ETCD_VERIFY=all' 'go' 'test' './...' '-short' '-timeout=3m' '--race')
ok go.etcd.io/etcd/client/v2 0.115s
% (cd client/v3 && 'env' 'ETCD_VERIFY=all' 'go' 'test' './...' '-short' '-timeout=3m' '--race')
--- FAIL: TestTxnPanics (1.00s)
txn_test.go:104: #2: did not panic, wanted panic cannot call If after Else!
FAIL
FAIL go.etcd.io/etcd/client/v3 4.374s
ok go.etcd.io/etcd/client/v3/clientv3util 0.054s [no tests to run]
....

--- a/client/v3/txn_test.go
+++ b/client/v3/txn_test.go
@@ -61,6 +61,7 @@ func TestTxnPanics(t *testing.T) {
                {
                        f: func() {
                                defer df()
+                               time.Sleep(2 * time.Second)
                                kv.Txn(context.TODO()).Else(op).If(cmp)
                        },

[Mskxn]

Sorry for my too-brief description, I think the failed repro it caused by the 2-second delay, which is too long, so other testcases are finished and the test routine just exits. A 1-second delay can repro it with a higher possibility.

You can also delay some of them to gain a larger race window. For example, delay case 0 for 1.5 seconds and case 1 for 1 second, it is more stable to repro. In my case, it is :

=== RUN   TestTxnPanics
    txn_test.go:105: #0: did not panic, wanted panic cannot call If twice!
    txn_test.go:102: #1: got cannot call If twice!, wanted cannot call If after Then!
--- FAIL: TestTxnPanics (1.51s)

FAIL

The second error is not from the second case, instead, it is the late error from the first

[lavacat]

@Mskxn thanks I was able to reproduce with 1s delay.

=== RUN   TestTxnPanics
    txn_test.go:104: #2: did not panic, wanted panic cannot call If after Else!
    txn_test.go:101: #3: got cannot call If after Else!, wanted cannot call Then twice!
--- FAIL: TestTxnPanics (1.01s)

Do you want to submit a fix? Seams like you have a solution in mind.

[lavacat]

please close the issue