[Bug c++/24996] [4.0/4.1/4.2 Regression] ICE on throw code

[mmitchel at gcc dot gnu dot org]

------- Comment #12 from mmitchel at gcc dot gnu dot org  2006-01-23 05:10 -------
I don't fully understand stabilize_expr; that's Jason's invention, IIRC. 
However, I think that the problem is that for a COND_EXPR we can't pre-evaluate
both arms of the conditional, because only one of them is supposed to be
executed at runtime.

I think this is a bug in the gimplifier.  I certainly don't see anything in the
documentation for CLEANUP_EXPR that suggests the restriction you've described.

I suppose that we could try to turn:

  b ? T(x) : T(y)

into:

  SAVE_EXPR b,
  TARGET_EXPR (tmp, b ? T(&tmp, x) : T(&tmp, y), b ? cleanup1 : cleanup2)

but that seems difficult, and I'm not sure it would work.

By the way, here's a simpler test case:

struct A {
  A();
  ~A();
};

struct S {
  S(const A&);
  ~S();
};

void foo(bool b)
{
  throw b ? S(A()) : throw 0;
}


-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at redhat dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24996