[Bug rtl-optimization/60769] New: [4.8 Regression] ICE: Max. number of generated reload insns per insn is achieved (90)

[glisse at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60769

            Bug ID: 60769
           Summary: [4.8 Regression] ICE: Max. number of generated reload
                    insns per insn is achieved (90)
           Product: gcc
           Version: 4.8.3
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: glisse at gcc dot gnu.org

Hello,

the following seems already fixed in 4.9, but it would be great to backport it
(I don't know which patch) to 4.8. For many bugs, it is often possible to find
a workaround, but for allocation bugs, it is quite mysterious what fragile
changes I have to make to my code to work around it. I tested with debian's
current gcc-4.8, which is essentially r207828 from gcc-4_8-branch.

Compile with g++-4.8 -c -O ouin.ii:
ouin.ii: In function ‘void fn3(N&)’:
ouin.ii:38:1: internal compiler error: Max. number of generated reload insns
per insn is achieved (90)

 }
 ^




template <class T> void fun(T);
struct B {};
struct R {
  int *x;
  B f;
};
R v(int &, R);
void rfun(R &);
struct A {
  void m_fn2(R p1) {
    R a = p1;
    rfun(p1);
    fun(this);
    fun(a);
  }
};
struct J {
  A ep;
  A ap;
  int c2a;
  void m_fn1(R &p2) {
    R d, e, b;
    v(c2a, p2);
    e = v(c2a, b);
    ap.m_fn2(e);
    v(c2a, p2);
    d = v(c2a, b);
    ep.m_fn2(d);
  }
};
struct N {
  int &p_;
  J cfo;
};
void fn3(N&n) {
  R h;
  n.cfo.m_fn1(h);
}
extern N &c;
void fn1() { fn3(c); }