[FIRRTL] LowerAnnotations rejects Reset/UInt<1> Wiring Problem source/sink pairs

[fabianschuiki]

I stumbled across the following use of SourceAnnotation/SinkAnnotation in the wild which tries to wire an abstract Reset signal to a concrete UInt<1>:

circuit Foo : %[[
  {
    "class":"firrtl.passes.wiring.SourceAnnotation",
    "target":"~Foo|Bar>y",
    "pin":"xyz"
  },
  {
    "class":"firrtl.passes.wiring.SinkAnnotation",
    "target":"~Foo|Foo>x",
    "pin":"xyz"
  }
]]
  module Bar :
    wire y : Reset
    y is invalid

  module Foo :
    inst bar of Bar
    wire x : UInt<1>
    x is invalid

Running this through firtool --parse-only produces the following error:

input.fir:14:5: error: Wiring Problem source type '!firrtl.reset' does not match sink type '!firrtl.uint<1>'
    wire y : Reset
    ^
input.fir:19:5: note: The sink is here.
    wire x : UInt<1>
    ^

Related to #4496 which added support for lowering some of these legacy wiring annotations as WiringProblems.

The issue is that the LowerAnnotations pass already checks for the types of the source/sink to line up. But the abstract Reset type only gets eliminated in InferResets, which runs after this pass. So we either have to make the anno lowering more lenient (accepting compatible types), or move the type checking to a later point when we actually deal with materializing XMRs or the wires themselves.

[dtzSiFive]

Making wiring more lenient for compatible types (especially if no truncation/extension required) seems reasonable.

We already look ahead and insert a wire + cast for reset taps in similar situation:

circt/lib/Dialect/FIRRTL/FIRRTLAnnotationHelper.cpp

Line 641 in 2efe829

// For resets, sometimes inject a cast between sink and target 'sink'.

.