[Bug middle-end/52693] Wrong code with SRA and arrays

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52693

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|mjambor at suse dot cz      |jamborm at gcc dot gnu.org

--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-03-23 22:30:08 UTC ---
(In reply to comment #0)
> With the attach testcase, SRA does the following replacement which seems wrong:
>   value = stack.l[D.4470_102];
> With
>   value_77 = MEM[(struct exprList_s *)&stack];
>   value$u_620 = MEM[(struct exprList_s *)&stack].u;
> 

It is confusing but I do not think it is wrong.  When you disable the
special handling of enumeral types in analyze_access_subtree, you get
(in this case IMHO equivalent):

  value$type_621 = MEM[(struct exprList_s *)&stack].type;
  value$u_622 = MEM[(struct exprList_s *)&stack].u;

which is exactly what we'd expect when breaking value into the two
components it consists of.  And the testcase still fails.

It is equivalent because the first MEM_REF is of type unsigned int,
whereas the second one loads the whole structure and then there's a
COMPONENT_REF to load only the second integer from it.