go-ethereum/swarm/testutil/norace.go

27 lines
1.2 KiB
Go
Raw Normal View History

swarm: fix network/stream data races (#19051) * swarm/network/stream: newStreamerTester cleanup only if err is nil * swarm/network/stream: raise newStreamerTester waitForPeers timeout * swarm/network/stream: fix data races in GetPeerSubscriptions * swarm/storage: prevent data race on LDBStore.batchesC https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461775049 * swarm/network/stream: fix TestGetSubscriptionsRPC data race https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461768477 * swarm/network/stream: correctly use Simulation.Run callback https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461783804 * swarm/network: protect addrCountC in Kademlia.AddrCountC function https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-462273444 * p2p/simulations: fix a deadlock calling getRandomNode with lock https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-462317407 * swarm/network/stream: terminate disconnect goruotines in tests * swarm/network/stream: reduce memory consumption when testing data races * swarm/network/stream: add watchDisconnections helper function * swarm/network/stream: add concurrent counter for tests * swarm/network/stream: rename race/norace test files and use const * swarm/network/stream: remove watchSim and its panic * swarm/network/stream: pass context in watchDisconnections * swarm/network/stream: add concurrent safe bool for watchDisconnections * swarm/storage: fix LDBStore.batchesC data race by not closing it
2019-02-13 06:03:23 -06:00
// Copyright 2019 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
// +build !race
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
2019-02-20 15:57:42 -06:00
package testutil
swarm: fix network/stream data races (#19051) * swarm/network/stream: newStreamerTester cleanup only if err is nil * swarm/network/stream: raise newStreamerTester waitForPeers timeout * swarm/network/stream: fix data races in GetPeerSubscriptions * swarm/storage: prevent data race on LDBStore.batchesC https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461775049 * swarm/network/stream: fix TestGetSubscriptionsRPC data race https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461768477 * swarm/network/stream: correctly use Simulation.Run callback https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-461783804 * swarm/network: protect addrCountC in Kademlia.AddrCountC function https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-462273444 * p2p/simulations: fix a deadlock calling getRandomNode with lock https://github.com/ethersphere/go-ethereum/issues/1198#issuecomment-462317407 * swarm/network/stream: terminate disconnect goruotines in tests * swarm/network/stream: reduce memory consumption when testing data races * swarm/network/stream: add watchDisconnections helper function * swarm/network/stream: add concurrent counter for tests * swarm/network/stream: rename race/norace test files and use const * swarm/network/stream: remove watchSim and its panic * swarm/network/stream: pass context in watchDisconnections * swarm/network/stream: add concurrent safe bool for watchDisconnections * swarm/storage: fix LDBStore.batchesC data race by not closing it
2019-02-13 06:03:23 -06:00
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
2019-02-20 15:57:42 -06:00
// RaceEnabled is true when -race flag is provided to the go tool. This const
// might be used in tests to skip some cases as the race detector may increase
// memory usage 5-10x and execution time by 2-20x. That might causes problems
// on Travis. Please, use this flag sparingly and keep your unit tests
// as light on resources as possible.
const RaceEnabled = false