Refactor instance tests around a network harness - #508
Conversation
97c53ab to
d147e3d
Compare
4bb3bc9 to
5040547
Compare
| // order to do so. | ||
| // We then check does so by checking they participated in signing | ||
| // Equivalent test as the previous TestInstanceMixedNodeType. | ||
| func TestNonValidator_BecomesValidator(t *testing.T) { |
There was a problem hiding this comment.
equivalent to the previous: TestInstanceMixedNodeType
| // TestNonValidator_StaysNonValidator ensures that a non-validator does not restart when it is processing | ||
| // previous epoch changes. | ||
| // Equivalent to TestInstanceNonValidatorBootstraps | ||
| func TestNonValidator_StaysNonValidator(t *testing.T) { |
There was a problem hiding this comment.
previously was TestInstanceNonValidatorBootstraps
| // then it is no longer a validator, and finally it is | ||
| // Equivalent to: TestInstanceValidatorSkipsAnEpoch. This also starts a validator when the tip is a sealing block so it covers an edge | ||
| // case previously caught from the logging test. | ||
| func TestInstanceValidatorSkipsAnEpoch(t *testing.T) { |
There was a problem hiding this comment.
Equivalent to: TestInstanceValidatorSkipsAnEpoch. This also starts a validator when the tip is a sealing block so it covers an edge case previously caught from the logging test.
7ff7591 to
ed0e6d5
Compare
ed0e6d5 to
ac05817
Compare
ac05817 to
070a524
Compare
070a524 to
4ac8236
Compare
4ac8236 to
569e39c
Compare
569e39c to
274e913
Compare
dbea945 to
6287af1
Compare
6287af1 to
f933854
Compare
f933854 to
d4522b2
Compare
The file now holds only helpers. The new instance tests land in a fresh instance_test.go later in the series.
MockStorage parses stored blocks through an injected testInnerBlockDeserializer instead of the package level parseTestInnerBlock, and hands its WAL bookkeeping to a standalone walCreator. testPlatformChain wakes WaitForProgress waiters over a channel rather than a sync.Cond, looks up validator sets by exact height, and lets a test install a set at a height. reparseBlock inlines toRawBlock.
A network drives a set of nodes over an instanceComm that hands each message straight to the destination instance, and steps the chain one block at a time through a blockBuilderVM gated by a controlled block builder. Nodes start from a storage seeded with genesis, or from newChainStorage when a test needs the epoch defining block at the tip.
Each test builds a network and steps it a block at a time. Covers the epoch transition paths the old tests skipped: a non validator that becomes a validator, a validator dropped from the set, a node offline across a transition, and a validator that misses an epoch entirely.
adapters_test.go builds its node through the network harness instead of wiring an instance by hand, util_test.go follows the newTestPChain rename, and the instance logs the validator set it is notified of on an epoch change.
e86e081 to
48563fc
Compare
| select { | ||
| case i.queue <- m: | ||
| default: | ||
| i.n.t.Errorf("node %x dropped a message, queue is full at %d", i.id, maxInFlightMessages) |
There was a problem hiding this comment.
Why not just block until the message is enqueued?
There was a problem hiding this comment.
because blocking would probably mean a deadlock, since its being called from someones Instance lock. I think we should just set a reasonable capacity on the channel (in this case 10000 seems plenty), and if we drop one because the channel is full it probably means we are sending too many messages.
There was a problem hiding this comment.
I know it'll be a deadlock, but since we have a reasonable capacity, wouldn't that mean that reaching this capacity and having a deadlock is a sign that our tests or production code is wrong?
Just logging an error and continuing is IMO just avoiding a problem instead of facing it.
I think we should perhaps even panic if we are out of capacity.
| genesisNodes := pChain.GenesisValidatorSet().Nodes() | ||
| common.SortNodes(genesisNodes) | ||
| validatorSets[1] = genesisNodes | ||
| validatorSets[firstEverEpoch] = genesisNodes |
There was a problem hiding this comment.
I still don't understand why we need an epoch-->validator set mapping.
Why is having a single validator set not sufficient? The network can only advance to a single epoch at a time.
|
|
||
| // TestInstanceZeroBlockUsesLastNonSimplexPChainHeight asserts that the first ever Simplex block | ||
| // references the P-chain height of the last non-Simplex block. | ||
| func TestInstanceZeroBlockUsesLastNonSimplexPChainHeight(t *testing.T) { |
There was a problem hiding this comment.
do we have a test equivalent to this test?
2 participants
Final PR of the cleanup-instance stack (4/4), on top of the transition listener PR.
instance_testhelpers_test.go, rebuilt around anetworkharness:addNode,acceptNewBlock,waitUntilSealingBlock. The queue-basedinMemNetworkis replaced byinstanceComm, which delivers messages directly and re-parses blocks so recipients do not share mutable canoto state.instance_test.gobecomes short scenario tests. Epoch sealing tests now rely on production approval/aux info dissemination instead of hand-injecting approvals (removing the old TODO).TestParseBlockSizeMatchesBytesmoves verbatim toexternal_test.gonext toexternal.go.TestNonValidator_StaysNonValidator(replacingTestInstanceNonValidatorBootstraps) is skipped pending offline-node timeouts, andTestInstanceRestartAcrossEpochsis dropped (thenode.restart()helper remains unused).