Bump cardano-node to (almost) 1.28.0

[Anviking]

Issue Number

ADP-952

Overview

• Bump to cardano-node master, in preparation for 1.28.0 release (Add preliminary 1.28.0 snapshot input-output-hk/cardano-haskell#49)
• Update readme

Comments

Basic Alonzo Era support is a non-goal for this PR. I implemented fromAlonzoTx and fromAlonzoBlock as error "unimplemented"s for now. Will fill them in in a follow-up PR, perhaps with integration cluster support at the same time.

• Hydra jobset
• Cabal build

[Anviking]

Suggested change

	ShelleyBasedEraAlonzo -> error "todo: toAlonzoTxOut"
	ShelleyBasedEraAlonzo -> error "todo(ADP-952): toAlonzoTxOut"

[rvl]

If you use something like this:

error "toAlonzoTxOut unimplemented" -- TODO: [ADP-952] toAlonzoTxOut

then it's easy to find all your TODOs with lentil:

$ lentil -t ADP-952 lib
263 source files [****************] 100%
lib/shelley/src/Cardano/Wallet/Shelley/Compatibility.hs
   521  fromAlonzoBlock [ADP-952]
   923  fromAlonzoTx [ADP-952]
  1148  datumHash [ADP-952]
  1167  datumHash [ADP-952]
  1186  datumHash [ADP-952]
 
lib/shelley/src/Cardano/Wallet/Shelley/Transaction.hs
  1173  We presumably need to provide the protocol params if our tx uses
        scripts? [ADP-952]
  1220  toAlonzoTxOut [ADP-952]

[Anviking]

I now get

plutus-core/src/PlutusCore/Default/Builtins.hs:123:9: error:
--
  | • Could not deduce (GEq DefaultUni)
  | arising from a use of ‘makeBuiltinMeaning’
  | from the context: uni ~ DefaultUni
  | bound by the instance declaration
  | at plutus-core/src/PlutusCore/Default/Builtins.hs:120:10-60
  | or from: PlutusCore.Constant.Typed.HasConstantIn uni term
  | bound by the type signature for:
  | toBuiltinMeaning :: forall term.
  | PlutusCore.Constant.Typed.HasConstantIn uni term =>
  | DefaultFun
  | -> BuiltinMeaning term (CostingPart uni DefaultFun)
  | at plutus-core/src/PlutusCore/Default/Builtins.hs:122:5-20
  | • In the expression:
  | makeBuiltinMeaning
  | ((+) @Integer) (runCostingFunTwoArguments . paramAddInteger)
  | In an equation for ‘toBuiltinMeaning’:
  | toBuiltinMeaning AddInteger
  | = makeBuiltinMeaning
  | ((+) @Integer) (runCostingFunTwoArguments . paramAddInteger)
  | In the instance declaration for ‘ToBuiltinMeaning uni DefaultFun’
  | \|
  | 123 \|         makeBuiltinMeaning

for both stack and cabal, even though it seemed fixed locally (for stack build at least) 🤷‍♂️

[rvl]

Thanks @Anviking - I have done a squash and rebase.

The stack build error on Buildkite was because the auto-generated nix files weren't up-to-date, and this was because it failed to push back to github for some weird reason. But anyway, I regenerated them.

Let's see what happens with the integration tests...

bors r+

[iohk-bors]

Build failed:

• ci/hydra-build:required

#expected - integration tests passed, unit tests failed because a second swagger change was needed.

[rvl]

Again

bors r+

[iohk-bors]

Build failed:

• ci/hydra-build:required

#expected - Need to update golden json files, and "Time interpreter queried past the horizon" from integration tests.

[Anviking]

bors try

[iohk-bors]

try

Build succeeded:

• buildkite/cardano-wallet
• ci/hydra-build:required

[Anviking]

The stack build error on Buildkite was because the auto-generated nix files weren't up-to-date, and this was because it failed to push back to github for some weird reason. But anyway, I regenerated them.

Many thanks!

Not sure if I should squash everything, but merging.
bors r+

[iohk-bors]

Canceled.

[Anviking]

Fixed readme node revision and squashed

bors r+

[iohk-bors]

Build failed:

• buildkite/cardano-wallet

/tmp/stack-97b0a942bf6b3b39/test-ghc-env: openBinaryFile: resource busy (file is locked)
--
  |  
  | /tmp/stack-97b0a942bf6b3b39/test-ghc-env: openBinaryFile: resource busy (file is locked)
  | error: Command exited with code 1!

#2695 - but with test-ghc-env instead of weeder, but still likely related to weeder. Annotating it as such...

[Anviking]

bors r+

[Anviking]

bors r-

Saw the PastHorizon failure again...

[iohk-bors]

Canceled.

[Anviking]

bors try

[Anviking]

What is strange about https://hydra.iohk.io/build/6897950/nixlog/1 is that the PastHorizon failure come from reading a wallet in a part of a tests which doesn't use wallets. What actually caused the failure was CLI_NETWORK, which provided little details as of why it failed.

In https://buildkite.com/input-output-hk/cardano-wallet/builds/15481#6b969a8a-2a4a-4947-a21f-c1b41bae0882, there were PastHorizonErrors, but the retrying it ultimately made the test pass.

I can get PastHorizon errors locally using with stack test cardano-wallet:integration --ta '-j 25'.

I think there might be nothing in particular going on except that the tip-sync client had to compete with wallet chain-sync clients, and got out of date.

Things that might help down the line:

1. Use one chain-sync client for multiple wallets to reduce load
2. Chain-sync refactoring to handle rollbacks without re-negotiating (might reduce load)
3. Add logic to prioritize tip-sync client above wallets
4. Somehow get the TimeInterpreter from chain-sync, e.g. NodeToClient ChainSync with block metadata IntersectMBO/ouroboros-network#2901

Things we could do now:

1. Do nothing and merge for now
2. Reduce parallelism in CI and/or increase slot or epoch length
3. Attempt to recover from PastHorizon errors by waiting for the interpreter to be such up to date that all Tip block we have seen (from all chain-sync threads) are within the horizon.
   • Obviously more tricky than the above, but might be doable 🤔

[iohk-bors]

try

Build succeeded:

• buildkite/cardano-wallet
• ci/hydra-build:required

[Anviking]

bors r+

[iohk-bors]

Build failed:

• ci/hydra-build:required

[cardano-wallet.network:Error:47407] [2021-07-14 12:44:59.73 UTC] Time interpreter queried past the horizon. This should not happen because node tip is within safe-zone
Called from:
"src/Cardano/Wallet/Api/Server.hs:2941:15 in cardano-wallet-core-2021.6.11-5SxBQv4U3MgHTXQIA3DnMI:Cardano.Wallet.Api.Server"
Converting expression:
Some (ELet (ERelSlotToEpoch (EAbsToRelSlot (ELit (SlotNo 4411)))) (\x0 -> EPair (ERelToAbsEpoch (EVar x0)) (ESnd (EVar x0))))

 src/Test/Integration/Scenario/CLI/Network.hs:94:11:
  1) CLI Specifications, COMMON_CLI_NETWORK, CLI_NETWORK - cardano-wallet network information
       expected: ExitSuccess
        but got: ExitFailure 1

  To rerun use: --match "/CLI Specifications/COMMON_CLI_NETWORK/CLI_NETWORK - cardano-wallet network information/"

Randomized with seed 1647667095

Hopefully the likelihood of this failure has been reduced now, so perhaps can be labeled as:
#expected
although borderline sketchy to do so

[Anviking]

bors r+

[iohk-bors]

Build succeeded:

• buildkite/cardano-wallet
• ci/hydra-build:required