Re: Fix stack overflow problem when bootstrapping on Darwin

[Geoffrey Keating <geoffk at apple dot com>]

On Friday, May 2, 2003, at 08:56  AM, Richard Sandiford wrote:

> Geoffrey Keating <gkeating@apple.com> writes:
> > 2003-04-29  Geoffrey Keating  <geoffk@apple.com>
> >
> > 	* c-semantics.c (expand_unreachable_stmt): Return a tree.
> > 	(expand_stmt): Update for change to expand_unreachable_stmt.
> > 	(expand_unreachable_if_stmt): Likewise.
>
> This patch caused an ICE on the test case below.  It missed some
> callers of expand_unreachable_stmt, so the else clause was no
> longer being expanded.
>
> Patch bootstrapped & regression tested on i686-pc-linux-gnu.
> OK to install?

I have a similar patch that doesn't use a ?: in a function's operand, 
which I think is nicer, but please commit the testcase since it tests a 
codepath that isn't tested at present.

(It turns out gcc.dg/switch-2.c also did fail, but I couldn't tell 
because I couldn't run a 'before' because I couldn't bootstrap.)

> (Maybe the function should be renamed?  "expand" doesn't seem all that
> appropriate now.  It does so some expansion, though, and I couldn't
> think of a better alternative.)

I actually considered changing it to expand_unreachable_stmts, but 
decided not to bother...

> 	* c-semantics.c (genrtl_if_stmt): Pass the return value of
> 	expand_unreachable_stmt to expand_stmt.
> 	(genrtl_switch_stmt): Likewise.
>
> Index: c-semantics.c
> ===================================================================
> RCS file: /cvs/gcc/gcc/gcc/c-semantics.c,v
> retrieving revision 1.56
> diff -c -d -p -F^[(a-zA-Z0-9_^#] -r1.56 c-semantics.c
> *** c-semantics.c	29 Apr 2003 20:39:12 -0000	1.56
> --- c-semantics.c	2 May 2003 13:18:39 -0000
> *************** genrtl_if_stmt (t)
> *** 416,434 ****
>     expand_start_cond (cond, 0);
>     if (THEN_CLAUSE (t))
>       {
> !       if (cond && integer_zerop (cond))
> ! 	expand_unreachable_stmt (THEN_CLAUSE (t), warn_notreached);
> !       else
> ! 	expand_stmt (THEN_CLAUSE (t));
>       }
>
>     if (ELSE_CLAUSE (t))
>       {
>         expand_start_else ();
> !       if (cond && integer_nonzerop (cond))
> ! 	expand_unreachable_stmt (ELSE_CLAUSE (t), warn_notreached);
> !       else
> ! 	expand_stmt (ELSE_CLAUSE (t));
>       }
>     expand_end_cond ();
>   }
> --- 416,434 ----
>     expand_start_cond (cond, 0);
>     if (THEN_CLAUSE (t))
>       {
> !       expand_stmt
> ! 	(cond && integer_zerop (cond)
> ! 	 ? expand_unreachable_stmt (THEN_CLAUSE (t), warn_notreached)
> ! 	 : THEN_CLAUSE (t));
>       }
>
>     if (ELSE_CLAUSE (t))
>       {
>         expand_start_else ();
> !       expand_stmt
> ! 	(cond && integer_nonzerop (cond)
> ! 	 ? expand_unreachable_stmt (ELSE_CLAUSE (t), warn_notreached)
> ! 	 : ELSE_CLAUSE (t));
>       }
>     expand_end_cond ();
>   }
> *************** genrtl_switch_stmt (t)
> *** 687,693 ****
>
>     emit_line_note (input_filename, lineno);
>     expand_start_case (1, cond, TREE_TYPE (cond), "switch statement");
> !   expand_unreachable_stmt (SWITCH_BODY (t), warn_notreached);
>     expand_end_case_type (cond, SWITCH_TYPE (t));
>   }
>
> --- 687,693 ----
>
>     emit_line_note (input_filename, lineno);
>     expand_start_case (1, cond, TREE_TYPE (cond), "switch statement");
> !   expand_stmt (expand_unreachable_stmt (SWITCH_BODY (t), 
> warn_notreached));
>     expand_end_case_type (cond, SWITCH_TYPE (t));
>   }
>
> *** /dev/null	Thu Apr 11 15:25:15 2002
> --- testsuite/gcc.c-torture/compile/20030502.c	Fri May  2 14:16:32 2003
> ***************
> *** 0 ****
> --- 1,12 ----
> + void foo ()
> + {
> +   if (1)
> +     goto foo;
> +   else
> +     for (;;)
> +       {
> +       foo:
> + 	bar ();
> + 	return;
> +       }
> + }