`cuda-translate` segfault

[IlanIwumbwe]

Required prerequisites

• Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

Running the compiler with -fno-array-conversion causes segfault in QIR translation phase

Steps to reproduce the bug

Run the code below

#include <cudaq.h>
#include <iostream>
#include <cmath> 
#include <vector> 
#include <complex> 

using namespace std::complex_literals;

CUDAQ_REGISTER_OPERATION(
unitary_1, 
    2,    // Number of qubits
    0,    // Number of params
    (std::vector<std::complex<double>>{
           {0.48963027692339345, -0.25497750785839352},{-0.22290322074166435, 0.30788011301354312},{ 0.50693167801629657, -0.036611223283494282},{ 0.30200841137366813, 0.44860167656042738},
           {-0.51738122275537601, -0.091208540881094272},{-0.83122487092786412, 0.079376387676733509},{ 0.088739148266466686, 0.010140598124231683},{ 0.082370385022818904, -0.10954417805917306},
           {0.088186930618414469, 0.62478914049322731},{-0.097310507557154757, -0.3128004789016916},{ -0.035253085563191573, -0.27404382928458881},{ 0.64664643631039709, 0.0073284857955460466},
           {-0.1010243030649756, 0.10441273962889276},{0.19981159839759674, -0.10512109154564615},{ 0.72831094545570485, 0.35626224934256495},{ 0.088607619728219733, -0.51254855126160204} 
    })
);

struct main_circuit { 
uint64_t operator()() __qpu__ {
	cudaq::qubit q0;
	cudaq::qubit q1;

	unitary_1(q0, q1);
	
	return 0;
}};

int main(int argc, char* argv[]) {
	auto kernel = main_circuit{};	
	return 0;	
}

Compile:

external/cuda-quantum/build/bin/nvq++ -fno-array-conversion test.cpp -o test

Error:

external/cuda-quantum/build/bin/nvq++: line 27: 259349 Segmentation fault      (core dumped) "$@"
failed: "/home/ilan/programming/QFpp/external/cuda-quantum/build/bin/cudaq-translate --convert-to=qir:0.1 test.qke.iCJ0fu -o test.ll.EHZxYU"

Replacing the custom unitary with some other primitive 2 qubit gate works. Compiling array conversion enabled works. This is presumably because the custom unitary vectors are treated as though they need dynamic heap allocation instead of creating the unitary matrix and embedding it in the quake call operation.

Expected behavior

Compiler should not segfault. If array conversion is expected, this should be communicated gracefully to the user.

Is this a regression? If it is, put the last known working version (or commit) here.

Not a regression

Environment

• CUDA-Q version: b04dee51
• Python version: 3.12.3
• C++ compiler: clang++ 18.1.3
• Operating system: Ubuntu 24.04

Suggestions

No response