Re: [PATCH] Fix tree-data-ref.c ICE on VIEW_CONVERT_EXPR<type>(0) (PR tree-optimization/33856)

[Jakub Jelinek <jakub at redhat dot com>]

On Sat, Oct 27, 2007 at 06:33:55PM +0200, Richard Guenther wrote:
> While there might be cases we (currently) do not fold
> VIEW_CONVERT_EXPR<>(cst), did you look at why we do not in this
> particular case (what's some_type?).  Possibly a separate PR for

Yes.  some_type in this case is a RECORD_TYPE, and folding VIEW_CONVERT_EXPR
(particularly in native_interpret_expr) handles only integral, floating
and vector types.  Not sure if it is a good idea to change that (or perhaps
just special case 0 there and turn that into a zero element CONSTRUCTOR of
the corresponding type).  Or perhaps SRA could avoid creating
VIEW_CONVERT_EXPR for RECORD_TYPE/UNION_TYPE.

> not folding this case is appropriate so we can address this during next
> stage1.
> 
> Does this PR happen because key is uninitialized?

Yes, that VIEW_CONVERT_EXPR is created during early SRA (and it a struct
of two ints, so on LP64 it is the same size as long)

	Jakub