Integrate ouroboros-network

[neilmayhew]

Description

Integrate ouroboros-network in preparation for cardano-node-10.2.0 release

See IntersectMBO/cardano-node#6040 for overall progress

[jasagredo]

FYI, the cardano-ledger commit in the SRP is gone. I suspect the new one is IntersectMBO/cardano-ledger@3fe0221 or IntersectMBO/cardano-ledger@12f6aa6 (the merge commit of the PR).

[lehins]

IntSet is more efficient at handling Set of Ints

Suggested change

	let is = Set.fromList is'
	let is = IntSet.fromList is'

[lehins]

Configuration should always try and use the types that will be used in the end. This way we can protect against invalid values when parsing the config instead of allowing for dangerous values that can lead to overflows. For example, by using SlotNo, we here we can protect against a very large value being placed in the config that will end up overflowing

Suggested change

	, gcfBlockFetchGracePeriod :: Maybe Integer
	, gcfBucketCapacity :: Maybe Integer
	, gcfBucketRate :: Maybe Integer
	, gcfCSJJumpSize :: Maybe Integer
	, gcfGDDRateLimit :: Maybe DiffTime
	, gcfBlockFetchGracePeriod :: Maybe DiffTime
	, gcfBucketCapacity :: Maybe Integer
	, gcfBucketRate :: Maybe Integer
	, gcfCSJJumpSize :: Maybe SlotNo
	, gcfGDDRateLimit :: Maybe DiffTime

[lehins]

@facundominguez I have a couple of suggestions and some minor concerns about this:

• csbcCapacity is already an Integer, so fromInteger is redundant
• using maybe instead of fromMaybe allows for default values to have the expected type.
• All conversions with to/fromInteger and fromIntegral should have type annotations, because that way it is easier to spot dangers of overflow
• SlotNo is backed by Word64, so there is danger of overflow here, so it would be better to use that type instead in gcfCSJJumpSize, as I suggested above.

Suggested change

	gbfcGracePeriod = fromInteger $ fromMaybe defaultBlockFetchGracePeriod gcfBlockFetchGracePeriod
	csbcCapacity = fromInteger $ fromMaybe defaultCapacity gcfBucketCapacity
	csbcRate = fromInteger $ fromMaybe defaultRate gcfBucketRate
	csjcJumpSize = fromInteger $ fromMaybe defaultCSJJumpSize gcfCSJJumpSize
	gbfcGracePeriod = maybe defaultBlockFetchGracePeriod (fromInteger @DiffTime) gcfBlockFetchGracePeriod
	csbcCapacity = fromMaybe defaultCapacity gcfBucketCapacity
	csbcRate = maybe defaultRate (fromInteger @Rational) gcfBucketRate
	csjcJumpSize = maybe defaultCSJJumpSize (fromInteger @SlotNo) gcfCSJJumpSize

[lehins]

@nbacquey List comprehensions seem to be more readable, more compact and won't rely on list fusion to be efficient.

Also switching to IntSet will be more efficient and there are special functions in containers notMember x s, that can be used instead of not $ member x s

Suggested change

	let is = Set.fromList is'
	in map fst $ filter (\(_, i) -> not $ i `Set.member` is) (zip xs [0..])
	let is = IntSet.fromList is'
	in [x | (x, i) <- zip xs [0..], i `IntSet.notMember` is]

[jasagredo]

@lehins This branch should incorporate the latest commits in blockfetch/milestone-1. In particular it is missing:

9e001f766 * origin/blockfetch/milestone-1 Make `ChainSelStarvation` carry an `Enclosed`
a36fd875d * Enrich comment about disabled `mustReplyTimeout`
513f34756 * Fix tests that relied on default timeouts
4e561806d * Rework default ChainSyncTimeouts in peer simulator
bee2c8445 * Log ChainSync mini-protocol events if need be
7b04c2822 * Make the `DynamoStarting` trace more explicit
ef092ca75 * Add a `TraceDrainingThePipe` event
3f8da3b60 * Improve and clarify CSJ documentation

I created #1351 which you can merge right away.

[jasagredo]

I removed the no changelog label. Please only apply it at the very end when merging this iff there are changes in non-testing libraries (those whose name is not prefixed by unstable-) that do not require a changelog entry. For tests/benchmarks/unstable-*/executables, the CI check should already ignore changes in those.

If we apply this label too early in the process, we risk introducing changes in the meantime which don't end up in the Changelog.