[Bug fortran/64980] [5 Regression] ICE in trans-expr.c

["mikael at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64980

--- Comment #11 from Mikael Morin <mikael at gcc dot gnu.org> ---
Created attachment 34755
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34755&action=edit
Another proposed  fix

(In reply to Bernd Edlinger from comment #10)
> Yes, but I have no idea how to know how the result is going to be
> used, in gfc_conv_procedure_call, line 4544.  If you have an idea
> how to detect this case, that would be great.

It is used as argument to the procedure, and also added to the interface
mapping mechanism through gfc_add_interface_mapping.
>From there, it is dropped if the resulting size can be determined by other
means; otherwised it is used to set the array result bounds.

The attached patch (untested so far) makes a copy of the reference before
passing it through VIEW_CONVERT_EXPR.
It avoids the second V_C_E at least.

> 
> Until yesterday, I thought the result is always used in a procedure call
> context, and I thought we only have aliasing violations on different
> variations of the same principal class, but that was also wrong.
What I don't understand is why there isn't something alike needed for non-class
arrays?  They also use different variations of the same principal class, don't
they?