Nested associate statements internal compiler error

[ivanp]

Nesting associate statements leads to ICE's in both ifort and ifx. Here is a MWE which aborts:

associate(d => sin(42.0))
    associate(g => d)
        print *, g, loc(g) == loc(d)
    end associate
end associate
end

 Output of other compilers can be found in Compiler Explorer: https://godbolt.org/z/TE7z37vne

[ivanp]

It looks like this has been resolved with version 2025.3.0.

[Igor_V_Intel · Accepted solution]

It was a bug in ifx 2025.2 and earlier. Fixed in 2025.3.

[JFH]

With my version of ifx the ICE has gone away but the output still looks wrong because the result of loc(g) == loc(d) is F. Gfortran gives T.

```

(lf) john:~$ ifx --version
ifx (IFX) 2025.3.0 20251010
Copyright (C) 1985-2025 Intel Corporation. All rights reserved.

(lf) john:~$ ifx associate.f90 && ./a.out
-0.9165215 F
(lf) john:~$ gfortran associate.f90 && ./a.out
-0.916521549 T
(lf) john:~$

```

[ivanp]

We'd have to read the standard "fine-print" here. I'm not sure the results of `loc` extension are relevant here. I used it to determine if compilers are creating copies or not. Note that you cannot use `c_loc` on `g` or `d`, because they don't have the pointer or target attribute. 

[JFH]

There is no standard "fine print" because the F2023 standard has no LOC intrinsic, and so gfortran and ifx must have provided it as extensions. The different outputs suggest that ASSOCIATE  was implemented in different ways that both work.