Cardano Updates

Added tracing points

Merge #937

937: Remove left-over appendix from LICENSE r=coot a=mrBliss

Remove left-over appendix from LICENSE

Merge #933

933: Remove Mux close operation r=karknu a=karknu

The mux close operation isn’t used, and leaving it around means that it could cause double closes if someone starts to use it in the future.

Merge #928

928: Minor additions to io-sim and io-sim-classes r=edsko a=edsko

• Add asyncThreadId to MonadFork
• Add forkWithUnmask to MonadTimer
• Implement derived timeoutAfter

Note that in order to correctly implement timeoutAfter, we needTimeouts to be comparable for equality; that was already the case, but we did not require it in MonadTimer; now we do.

Adding a few people as reviewer, not because it’s all that tricky but because I don’t know who’s around. However, implementation of forkWithUnmask would benefit from a careful check. I think it’s probably okay, but not 100% sure.

Make sure that ‘Async m’ is a ‘Functor’

Merge #935

935: Correctly derive the epoch size instead of hardcoding it to 21600 r=mrBliss a=mrBliss

Merge #934

934: Fixed a typo in haddock r=mrBliss a=coot

Correctly derive the epoch size instead of hardcoding it to 21600

For Byron, we now use kEpochSlots :: BlockCount -> EpochSlots (basically * 10) to derive the epoch size from k, which is extracted from the genesis config. In order to have access to the genesis config, nodeEpochSize now takes the NodeConfig as argument.

Fixed a typo in haddock

Remove Mux close operation

The mux close operation isn’t used, and leaving it around means that it could cause double closes if someone starts to use it in the future.

Merge #931

931: Update dependencies r=mrBliss a=erikd

consensus: avoid overly conservative ReadFutureSlot in newIterator

Nix voodoo

Update dependencies

fix CDDL definitions

In test/CDDL.hs there are some type synonyms for the protocols, including

type CS = ChainSync DummyBytes DummyBytes

That’s a chain sync codec where the point type is the same as the header type. What it always should have been is

type CS = ChainSync DummyBytes (Point DummyBytes)

but this was never noticed, because the CDDL tests never actually tested this part of the codec! The CDDL definition for a point was

point = bytes .cbor any

although there was a different one for block fetch (which I remember adding) which looks like

point = origin / [slotNo, blockHash]
origin = []
slotNo = uint
blockHash = null ; because the tests use () for hashes

I’ve consolidated this.

And finally, with this pull request, CBOR-in-CBOR is used for header and block encodings, so I had to update the CDDL file to use

\#6.24(bytes . cbor bytes)
```.

Merge #919

919: export default PBFT signature threshold r=avieth a=avieth

Merge #899

899: Check for file handle leaks in the consensus tests r=mrBliss a=mrBliss

For each node in the test setup, we return a NodeInfo record which stores the mock file-systems the node used for its ChainDB. This record can later be extended with other functionality, e.g., a way to shut down the node, or other information about the node.

We combine all useful output of a node in the new NodeOutput record (as soon as we have more than 2 values and we’re writing {fst,snd,thd}3, it’s a sign we have to switch to a record). In prop_general we check for file handle leaks.

Merge #930

930: update ouroboros-network/README.md for test-crypto relocation r=mrBliss a=nfrisby

#684 relocated test-crypto to a different package. This removes a stale reference to it from the README.

update ouroboros-network README: test-crypto is now elsewhere

Fixed nix setup

Fixed nix setup

Test state changes in subscriptionWorker

Test state changes in subscriptionWorker

Avoid redundant import

Merge #893

893: Fix a performance bottleneck in gkSigners r=mrBliss a=nc6

Merge #893

893: Fix a performance bottleneck in gkSigners r=mrBliss a=nc6

Comments and style tweaks

Comments and style tweaks

Add regular property/unit tests for pruneChainState

ghcjs: a stab at support

WIP

Merge #923

923: Reduce ChainDB memory usage r=mrBliss a=mrBliss

Before: After:

The profiled program adds the first 30,000 blocks from disk (so we add as fast we can read them) to the ChainDB.

Adjust RunNode’s CBOR decoders to provide the input bytes

Adjust RunNode’s CBOR decoders to provide the input bytes

Rewrite the Byron en/decoders to use the provided input bytes

The default de/encoder for a Byron block/header no longer uses CBOR-in-CBOR. These en/decoders are used by the storage layer, which now passes the decoded ByteString to the decoded value, which can use it to “annotate” the Byron block/header.

Rewrite the Byron en/decoders to use the provided input bytes

The default de/encoder for a Byron block/header no longer uses CBOR-in-CBOR. These en/decoders are used by the storage layer, which now passes the decoded ByteString to the decoded value, which can use it to “annotate” the Byron block/header.

WIP: Fix and test resource registry

Reduce ChainDB memory usage

After some heap profiling, these bangs provided the most bang for the buck (sorry for the pun).

For example, the memory usage of adding the first 30,000 blocks from disk (so as fast as we can read them) to the ChainDB reduced from +240 MiB (and growing) to 80 MiB. Note that these measurements were performed with block validation disabled.

Reduce ChainDB memory usage

After some heap profiling, these bangs provided the most bang for the buck (sorry for the pun).

For example, the memory usage of adding the first 30,000 blocks from disk (so as fast as we can read them) to the ChainDB reduced from +240 MiB (and growing) to 80 MiB. Note that these measurements were performed with block validation disabled.

Added CompletedSubscriptionLoop callback

Minor additions to io-sim and io-sim-classes

• Add asyncThreadId to MonadFork
• Add forkWithUnmask to MonadTimer
• Implement derived timeoutAfter

Note that in order to correctly implement timeoutAfter, we need Timeouts to be comparable for equality; that was already the case, but we did not require it in MonadTimer; now we do.

Minor additions to io-sim and io-sim-classes

• Add asyncThreadId to MonadFork
• Add forkWithUnmask to MonadTimer
• Implement derived timeoutAfter

Note that in order to correctly implement timeoutAfter, we need Timeouts to be comparable for equality; that was already the case, but we did not require it in MonadTimer; now we do.

bugfix: allow closed parent in forkLinkedTransfer

Make the slot duration part of RunNode

Merge #920

920: Make the StartTime part of RunNode r=Jimbo4350 a=mrBliss

When starting a node, the nodeStartTime method can be used to obtain the start time. This method will return the right time for each block type, and thus protocol. For example, for real Byron blocks, it will extract it from the Genesis config. For mock blocks, it will just use some hard-coded dummy date, since the start time is not important for mock blocks.

Added SocketStateChange Callback

Added SocketStateChange Callback

Added SocketStateChange Callback

Added SocketStateChange Callback