Re: [PATCH v3 00/11] Improve Fortran intrinsic types and procedures

[Luis Machado <luis.machado@arm.com>]

On 4/25/22 14:19, Kempke, Nils-Christian wrote:
> Hi,
> 
> Sorry for that - I think the assert should have been removed anyway.
> 

Not a problem. I just want to make sure I understand the failure and 
what needs to be done to address it.

> I might have missed removing it - but the two lines below are what I wrote instead - an
> error and gdb should not crash..
> 
>    if (value_type (arg1)->code () == TYPE_CODE_COMPLEX
>        || value_type (arg2)->code () == TYPE_CODE_COMPLEX)
>      error (_("Types of arguments for CMPLX called with more then one argument "
> 	     "must be REAL or INTEGER"));
> 
> Does removing the assert do that? I'd assume the tests then just FAIL.
> 

Just for informational purposes, dropping the assertion makes the code 
assert somewhere else.

We could stop with an error like "Unsupported kind" if kind_arg->code () 
is unexpected, but I'm wondering if we could've caught this somewhere 
else before calling into this function with something we know is wrong 
already. If we could do that, then we would fail early, and possibly 
more gracefully.

I see errors of the following type:

"unsupported kind 1 for type complex*4"

This gives a bit more information, but I don't think a code == 
TYPE_CODE_ERROR gives us a chance to show anything more.