cudaq.translate(format='openqasm2') fails for adjoint modifier

[arulandu]

Consider the following cudaq kernel.

  def s_kernel():
      k, q = cudaq.make_kernel(cudaq.qubit)
      k.s(q)
      return k

  kernel = cudaq.make_kernel()
  reg = kernel.qalloc(1)
  kernel.adjoint(s_kernel(), reg[0])
  
  print("CUDAQ", kernel)
  print("QASM", cudaq.translate(kernel, format="openqasm2"))

Output:

CUDAQ module attributes {quake.mangled_name_map = {__nvqpp__mlirgen____nvqppBuilderKernel_NBJIEOE85U = "__nvqpp__mlirgen____nvqppBuilderKernel_NBJIEOE85U_PyKernelEntryPointRewrite"}} {
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_NBJIEOE85U() attributes {"cudaq-entrypoint"} {
    %0 = quake.alloca !quake.veq<1>
    %1 = quake.extract_ref %0[0] : (!quake.veq<1>) -> !quake.ref
    quake.apply<adj> @__nvqpp__mlirgen____nvqppBuilderKernel_FVE4B53IYY %1 : (!quake.ref) -> ()
    return
  }
  func.func @__nvqpp__mlirgen____nvqppBuilderKernel_FVE4B53IYY(%arg0: !quake.ref) {
    quake.s %arg0 : (!quake.ref) -> ()
    return
  }
}

QASM // Code generated by NVIDIA's nvq++ compiler
OPENQASM 2.0;

include "qelib1.inc";

gate nvqpp__mlirgen____nvqppBuilderKernel_FVE4B53IYY q0 {
  s q0;
}

qreg var0[1];
@__nvqpp__mlirgen____nvqppBuilderKernel_FVE4B53IYY var0[0];

As you can see, the MLIR has a quake.apply<adj> as intended but the translated OpenQASM 2.0 does not have the adjoint.

Produced on Ubuntu 24.04.1 LTS.