Bump nim-serialization to potentially solve #435

[mratsim]

status-im/nim-serialization#10 and maybe solve #435

[mratsim]

Raised status-im/nim-serialization#11 for a loadFile smoked test in nim-serialization.
The error is strange:

../beacon_chain/eth2_network.nim(241, 9) template/generic instantiation from here
../vendor/nim-serialization/serialization.nim(92, 16) Error: type mismatch: got <ByteStreamVar>

but expected one of: 

proc close(f: File)
proc close[TMsg](c: var Channel[TMsg])
proc close(s: var ByteStream)
template close(s: AsciiStream | UnicodeStream | ObjectStream)
proc close(transp: StreamTransport)
proc close(server: StreamServer)
proc close(rw: AsyncStreamRW)
proc close(socket: Socket)
proc close(transp: DatagramTransport)
proc close(stream: P2PStream): Future[void]
proc close(api: DaemonAPI): Future[void]

However according to Nim auto-dereferencing rule, Nim should be able to autodereference ByteStreamVar to ByteStream:

https://github.com/status-im/nim-faststreams/blob/e7a34b74f298eaabd7bdbe2f080bf9747ed6bfd9/faststreams/input_stream.nim#L14-L166

type
  # ...
  ByteStream* = object of RootObj
    head*: ptr byte
    bufferSize: int
    bufferStart, bufferEnd: ptr byte
    cursorsList: ptr StreamCursor
    reader: StreamReader
    asyncReader: AsyncStreamReader
    closeStream: CloseStreamProc
    bufferEndPos: int

  # ...

  # TODO: ByteStreamVar should become a `var` short-lived pointer once this is supported
  ByteStreamVar* = ref ByteStream
 
proc openFile*(filename: string): ByteStreamVar =
  new result
  var memFile = memfiles.open(filename)
  result.head = cast[ptr byte](memFile.mem)
  when debugHelpers:
    result.bufferStart = result.head
  result.bufferEnd = result.head.shift memFile.size
  result.bufferEndPos = memFile.size
  result.closeStream = proc = close memFile

# ...

# TODO: use a destructor once we migrate to Nim 0.20
proc close*(s: var ByteStream) =
  if s.closeStream != nil:
    s.closeStream()

Solutions:

• either expose close for ByteStreamVar
• or do stream[].close() in nim-serialization loadFile
• or use a finalizer in nim-faststreams (which is undeterministic but ensure that we do not forget to close the stream)

[jangko]

cannot solve #435

[mratsim]

turns out the illegal storage access is in the Nim compiler, it probably doesn't like defer

compiling beacon_node crashes

sync_protocol.nim(51, 6) Hint: 'sync_protocol.fromHeaderAndBody(b: var BeaconBlock, h: BeaconBlockHeader, body: BeaconBlockBody)[declared in sync_protocol.nim(51, 5)]' is declared but not used [XDeclaredButNotUsed]
sync_protocol.nim(42, 6) Hint: 'sync_protocol.toHeader(b: BeaconBlock)[declared in sync_protocol.nim(42, 5)]' is declared but not used [XDeclaredButNotUsed]
Hint: request_manager [Processing]
Hint: validator_keygen [Processing]
Hint: interop [Processing]
beacon_node.nim(222, 41) Hint: conversion from string to itself is pointless [ConvFromXtoItselfNotNeeded]
SIGSEGV: Illegal storage access. (Attempt to read from nil?)

[mratsim]

So now #435 is triggered on a month old test that never failed before: https://ci.appveyor.com/project/nimbus/nim-beacon-chain/builds/27346548/job/gki7ew7tkda8ou8w#L242

[Suite] Official - Shuffling tests [Preset:  [Preset: minimal]
  [OK] Shuffling a sequence of N validators [Preset: minimal]
[Suite] Official - BLS tests
    Unhandled exception: The parameter is incorrect.
 [OSError]
  [FAILED] Private to public key conversion
  [OK] Message signing
  [OK] Aggregating signatures
  [OK] Aggregating public keys

[mratsim]

Can be rebased and merged once #441 is merged into master as #440 requires the same fix anyway.

[tersec]

Apparently no rebasing required.