Ci race detector handle failing tests (#19143)

* swarm/storage: increase mget timeout in common_test.go

 TestDbStoreCorrect_1k sometimes timed out with -race on Travis.

--- FAIL: TestDbStoreCorrect_1k (24.63s)
    common_test.go:194: testStore failed: timed out after 10s

* swarm: remove unused vars from TestSnapshotSyncWithServer

nodeCount and chunkCount is returned from setupSim and those values
we use.

* swarm: move race/norace helpers from stream to testutil

As we will need to use the flag in other packages, too.

* swarm: refactor TestSwarmNetwork case

Extract long running test cases for better visibility.

* swarm/network: skip TestSyncingViaGlobalSync with -race

As panics on Travis.

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x7e351b]

* swarm: run TestSwarmNetwork with fewer nodes with -race

As otherwise we always get test failure with `network_test.go:374:
context deadline exceeded` even with raised `Timeout`.

* swarm/network: run TestDeliveryFromNodes with fewer nodes with -race

Test on Travis times out with 8 or more nodes if -race flag is present.

* swarm/network: smaller node count for discovery tests with -race

TestDiscoveryPersistenceSimulationSimAdapters failed on Travis with
`-race` flag present. The failure was due to extensive memory usage,
coming from the CGO runtime. Using a smaller node count resolves the
issue.

=== RUN   TestDiscoveryPersistenceSimulationSimAdapter
==7227==ERROR: ThreadSanitizer failed to allocate 0x80000 (524288) bytes of clock allocator (error code: 12)
FATAL: ThreadSanitizer CHECK failed: ./gotsan.cc:6976 "((0 && "unable to mmap")) != (0)" (0x0, 0x0)
FAIL    github.com/ethereum/go-ethereum/swarm/network/simulations/discovery     804.826s

* swarm/network: run TestFileRetrieval with fewer nodes with -race

Otherwise we get a failure due to extensive memory usage, as the CGO
runtime cannot allocate more bytes.

=== RUN   TestFileRetrieval
==7366==ERROR: ThreadSanitizer failed to allocate 0x80000 (524288) bytes of clock allocator (error code: 12)
FATAL: ThreadSanitizer CHECK failed: ./gotsan.cc:6976 "((0 && "unable to mmap")) != (0)" (0x0, 0x0)
FAIL	github.com/ethereum/go-ethereum/swarm/network/stream	155.165s

* swarm/network: run TestRetrieval with fewer nodes with -race

Otherwise we get a failure due to extensive memory usage, as the CGO
runtime cannot allocate more bytes ("ThreadSanitizer failed to
allocate").

* swarm/network: skip flaky TestGetSubscriptionsRPC on Travis w/ -race

Test fails a lot with something like:
 streamer_test.go:1332: Real subscriptions and expected amount don't match; real: 0, expected: 20

* swarm/storage: skip TestDB_SubscribePull* tests on Travis w/ -race

Travis just hangs...

ok  	github.com/ethereum/go-ethereum/swarm/storage/feed/lookup	1.307s
keepalive
keepalive
keepalive

or panics after a while.

Without these tests the race detector job is now stable. Let's
invetigate these tests in a separate issue:
https://github.com/ethersphere/go-ethereum/issues/1245

1 files changed, 16 insertions, 39 deletions

diff --git a/swarm/network/stream/snapshot_sync_test.go b/swarm/network/stream/snapshot_sync_test.go
index b45d0aed5..330f39712 100644
--- a/swarm/network/stream/snapshot_sync_test.go
+++ b/swarm/network/stream/snapshot_sync_test.go
@@ -19,7 +19,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"io/ioutil"
 	"os"
 	"runtime"
 	"sync"
@@ -42,8 +41,6 @@ import (
 	"github.com/ethereum/go-ethereum/swarm/testutil"
 )
 
-const MaxTimeout = 600
-
 type synctestConfig struct {
 	addrs         [][]byte
 	hashes        []storage.Address
@@ -80,37 +77,31 @@ func TestSyncingViaGlobalSync(t *testing.T) {
 	if runtime.GOOS == "darwin" && os.Getenv("TRAVIS") == "true" {
 		t.Skip("Flaky on mac on travis")
 	}
+
+	if testutil.RaceEnabled {
+		t.Skip("Segfaults on Travis with -race")
+	}
+
 	//if nodes/chunks have been provided via commandline,
 	//run the tests with these values
 	if *nodes != 0 && *chunks != 0 {
 		log.Info(fmt.Sprintf("Running test with %d chunks and %d nodes...", *chunks, *nodes))
 		testSyncingViaGlobalSync(t, *chunks, *nodes)
 	} else {
-		var nodeCnt []int
-		var chnkCnt []int
+		chunkCounts := []int{4, 32}
+		nodeCounts := []int{32, 16}
+
 		//if the `longrunning` flag has been provided
 		//run more test combinations
 		if *longrunning {
-			chnkCnt = []int{1, 8, 32, 256, 1024}
-			nodeCnt = []int{16, 32, 64, 128, 256}
-		} else if raceTest {
-			// TestSyncingViaGlobalSync allocates a lot of memory
-			// with race detector. By reducing the number of chunks
-			// and nodes, memory consumption is lower and data races
-			// are still checked, while correctness of syncing is
-			// tested with more chunks and nodes in regular (!race)
-			// tests.
-			chnkCnt = []int{4}
-			nodeCnt = []int{16}
-		} else {
-			//default test
-			chnkCnt = []int{4, 32}
-			nodeCnt = []int{32, 16}
+			chunkCounts = []int{1, 8, 32, 256, 1024}
+			nodeCounts = []int{16, 32, 64, 128, 256}
 		}
-		for _, chnk := range chnkCnt {
-			for _, n := range nodeCnt {
-				log.Info(fmt.Sprintf("Long running test with %d chunks and %d nodes...", chnk, n))
-				testSyncingViaGlobalSync(t, chnk, n)
+
+		for _, chunkCount := range chunkCounts {
+			for _, n := range nodeCounts {
+				log.Info(fmt.Sprintf("Long running test with %d chunks and %d nodes...", chunkCount, n))
+				testSyncingViaGlobalSync(t, chunkCount, n)
 			}
 		}
 	}
@@ -123,21 +114,7 @@ var simServiceMap = map[string]simulation.ServiceFunc{
 			return nil, nil, err
 		}
 
-		var dir string
-		var store *state.DBStore
-		if raceTest {
-			// Use on-disk DBStore to reduce memory consumption in race tests.
-			dir, err = ioutil.TempDir("", "swarm-stream-")
-			if err != nil {
-				return nil, nil, err
-			}
-			store, err = state.NewDBStore(dir)
-			if err != nil {
-				return nil, nil, err
-			}
-		} else {
-			store = state.NewInmemoryStore()
-		}
+		store := state.NewInmemoryStore()
 
 		r := NewRegistry(addr.ID(), delivery, netStore, store, &RegistryOptions{
 			Retrieval:       RetrievalDisabled,