Cannot instantiate Future got: <T> but expected: <T> after bumping nim-chronos

[KonradStaniec]

After bumping nim-chronos to latest version when trying to run make fluffy following error manifests:

/nimbus-eth1/fluffy/fluffy.nim(112, 27) template/generic instantiation of `installDiscoveryApiHandlers` from here
/nimbus-eth1/fluffy/rpc/rpc_discovery_api.nim(103, 12) template/generic instantiation of `rpc` from here
/nimbus-eth1/vendor/nim-json-rpc/json_rpc/rpcproxy.nim(92, 23) template/generic instantiation of `rpc` from here
/nimbus-eth1/vendor/nim-json-rpc/json_rpc/server.nim(19, 16) template/generic instantiation of `rpc` from here
/nimbus-eth1/vendor/nim-json-rpc/json_rpc/router.nim(158, 56) Error: cannot instantiate Future
got: <T>
but expected: <T>

Imo this error is composed of two parts:

1. In https://github.com/status-im/nim-json-rpc/blob/master/json_rpc/router.nim#L150, correct return type for seq[Record] should be:

BracketExpr
  Ident "seq"
  Ident "Record"

but what really is there during compilation is:

Call
  OpenSymChoice
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
    Sym "[]"
  Sym "seq"
  Sym "Record"

Which makes me think we hit issue: nim-lang/Nim#11091

2. Second part of the issue is most probably the fact that after Improve ram usage nim-chronos#243 , our async macro generates iterator which takes generic future as arguments i.e has paramater Future[T].
   New version:
   iterator nameIter(chronosInternalRetFuture: Future[T]): FutureBase
   Old version:
   iterator nameIter(): FutureBase {.closure.}

and we hit some type of the issue : nim-lang/Nim#19358

So even the issue 1 existed from what I can tell always, the old version of chronos macro was not affected, but after the update
correct type cannot be instantiated.

One workaround this is to introduce type alias:

type Records = seq[Record]

then everything builds successfully, of course it does not solve root cause.

I still do not have enough knowledge of the compiler to propose the fix to the root cause, but imo it would be nice if proper BracketExpr was generated in our rpc macro. Wdyt @zah @kdeme ?

[Menduist]

Hey, stumbled onto this issue
https://github.com/status-im/nim-chronos/blob/87197230779002a2bfa8642f0e2ae07e2349e304/chronos/asyncmacro2.nim#L91-L94
This part of the transformation is here to handle the compiler issue with all the syms, so if it doesn't work anymore, it's probably a regression in chronos

I'll look around to see if status-im/nim-chronos#243 caused it

[mratsim]

You can workaround it by rebuilding the untyped AST.

Example for in transformed into contains: https://github.com/mratsim/weave/blob/71dc2d7/weave/parallel_macros.nim#L19-L71
or bound symbol that needs to be transformed back into identifiers: nim-lang/Nim#8531 (comment) and you need:

proc replaceNodes*(ast: NimNode): NimNode =
  # Replace NimIdent and NimSym by a fresh ident node
  proc inspect(node: NimNode): NimNode =
    case node.kind:
    of {nnkIdent, nnkSym}:
      return ident($node)
    of nnkEmpty:
      return node
    of nnkLiterals:
      return node
    else:
      var rTree = node.kind.newTree()
      for child in node:
        rTree.add inspect(child)
      return rTree
  result = inspect(ast)

[Menduist]

Fixed here: status-im/nim-chronos#261

Not clear why it worked at some point, will investigate later

[kdeme]

Fixed here: status-im/nim-chronos#261

Great, thanks! I can confirm the fluffy build works again with that chronos branch.