[Patch, fortran] PR29422 and PR29428 - allocatable component wrinkles

[Brooks Moses]

Richard E Maine wrote:
> On Oct 11, 2006, at 2:03 PM, Dominique Dhumieres wrote:
>>"this should be equivalent to x%a = y (for an unallocated x%a)"
>>
>>Do you mean that "x = mytype(y)" could be replaced by
>>"x%a = y"?

In this case (where x has no other components, and x%a is not already 
allocated), yes.

>>  If I do so, ifort aborts and gfortran and g95
>>seg fault.
> 
> Well, then they have a bug. That's completely independent of the  
> question of what the bounds should be. There isn't even any  
> "interesting" question about whether it is valid code (given either  
> f2003 or the allocatable TR).

I think it's more an "unimplemented feature" than a "bug" -- unless I'm 
quite missing something, allocate-on-assignment is only part of F2003, 
not part of TR-15581, and none of the compilers in question are claiming 
to have full F2003 support.

- Brooks