Philox generators don't increment second counter

[maleadt]

I'm trying to understand the implementation, so maybe I'm missing something, but it doesn't seem that this implementation correctly handles the second counter of, e.g., Philox2x? In the reference implementation, there is:

static unsigned long int NAME##_get(void *vstate){                      \
    NAME##_state *st = (NAME##_state *)vstate;                          \
    const int N=sizeof(st->ctr.v)/sizeof(st->ctr.v[0]);                 \
    if( st->elem == 0 ){                                                \
        ++st->ctr.v[0];                                                 \
        if( N>1 && st->ctr.v[0] == 0 ) ++st->ctr.v[1];                  \
        if( N>2 && st->ctr.v[1] == 0 ) ++st->ctr.v[2];                  \
        if( N>3 && st->ctr.v[2] == 0 ) ++st->ctr.v[3];                  \
        st->r = CBRNGNAME(st->ctr, st->key);                                 \
        st->elem = N;                                                   \
    }                                                                   \
    return 0xffffffffUL & st->r.v[--st->elem];                          \
}       

Furthermore, when the first counter runs out, the code crashes:

julia> using Random123

julia> rng = Philox2x(UInt32)
Philox2x{UInt32, 10}(0xfce15e42, 0x8ffc34da, 0x88bd3fad, 0x00000000, 0x00000000, 0)

julia> rng.ctr1 = typemax(UInt32)
0xffffffff

julia> rand(rng)
0.9878138457424939

julia> rand(rng)
ERROR: InexactError: trunc(UInt32, 4294967296)
Stacktrace:
 [1] throw_inexacterror(f::Symbol, #unused#::Type{UInt32}, val::Int64)
   @ Core ./boot.jl:602
 [2] checked_trunc_uint
   @ ./boot.jl:632 [inlined]
 [3] toUInt32
   @ ./boot.jl:716 [inlined]
 [4] UInt32
   @ ./boot.jl:756 [inlined]
 [5] convert
   @ ./number.jl:7 [inlined]
 [6] setproperty!
   @ ./Base.jl:34 [inlined]
 [7] rand
   @ ~/Julia/depot/packages/Random123/cb3eU/src/common.jl:49 [inlined]
 [8] rand (repeats 2 times)
   @ ~/Julia/depot/packages/RandomNumbers/m57pA/src/common.jl:25 [inlined]
 [9] top-level scope
   @ REPL[4]:1

[sunoru]

You are right. I don't remember why only the first counter was used. I'm gonna fix it in an update.

[sunoru]

Fixed