Re: [PATCH] Fix PR optimization/11210

[Ulrich Weigand <weigand at i1 dot informatik dot uni-erlangen dot de>]

Eric Botcazou wrote:

>I got caught by expr.c:get_inner_reference(), which is advertised as handling 
>NOP_EXPR and the likes, but doesn't return the value you would have expected 
>for NOP_EXPRs wrapping up bitfields. Not sure if this is a bug or a feature, 
>but this is certainly not what we want in decode_field_reference(). It's 
>amazing that neither IA-32 nor PPC caught the problem.

However, it did cause bootstrap failure on s390x due to miscompilation of
the following snippet from getattrtab.c:

struct attr_desc
{
  char *name;
  struct attr_desc *next;
  unsigned is_numeric : 1;
  unsigned negative_ok : 1;
  unsigned unsigned_p : 1;
  unsigned is_const : 1;
  unsigned is_special : 1;
  unsigned func_units_p : 1;
  unsigned blockage_p : 1;
  int lineno;
};

int
test (struct attr_desc *attr)
{
 if (! attr->is_special && ! attr->is_const)
     write_attr_get (attr);
}

fold_truthop would merge the two struct member accesses into a single
check, but miscompute the necessary mask because it though the members
were 32 bits wide instead of a single bit ...

>        * fold-const (decode_field_reference): Record outermost type in
>        case the expression is a NOP. Strip all NOPs. Set the signedness
>        to that of the outermost type (if any) when the bitsize is equal to the
>        size of the type.

This fixes the bootstrap problem on s390x-ibm-linux.

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  weigand@informatik.uni-erlangen.de