Delitmus parameter ordering is wrong

[MattWindsor91]

In a particular delitmusify run, I saw the local variables r0 and r1 on thread 0 get assigned to parameter slots 9 and 8 respectively, even though the actual parameters generated in the output were in the opposite order.

Methinks there are two issues:

• the parameter generator is somehow considering these variables in the wrong order;
• the delitmusifier isn't ensuring it outputs parameters in the same order as the parameter generator.

[MattWindsor91]

It turns out that the first one is not the case: the input test appears to have them in the wrong order to begin with! The second bit is the real issue.

[MattWindsor91]

For viewers at home, an example test case is this:

C test_10
// Hint: try simulating with herd7 <test_10.litmus>
// WARNING: C litmus output is experimental!

// Declaring global variables.
{
  x = 0;
  y = 0;
}

// Thread 0
void P0(atomic_int *x, atomic_int *y) {
  int r1 = 0;
  int r0 = 0;

  r0 = atomic_load_explicit(y, memory_order_acquire);
  r1 = atomic_load_explicit(x, memory_order_relaxed);
}

// Thread 1
void P1(atomic_int *x, atomic_int *y) {
  atomic_store_explicit(x, 1, memory_order_relaxed);
  atomic_store_explicit(y, 1, memory_order_release);
}

exists (0:r0 == 1 /\ 0:r1 == 0 /\ x == 1 /\ y == 1)

Note that r1 is declared before r0, and should ideally appear in the parameter list before it.