This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RF[CT]: IMA repairs 2/n: maybe_apply_renaming_pragma rawrite

> I attempted to bootstrap this patch on sparc-solaris2.8 but failed; I
> got an ICE in calls.c on libgcc2.c::_trampoline, here:
> 3301              /* Verify that we've deallocated all the stack we used. 
> */ 3302              if (! (flags & (ECF_NORETURN | ECF_LONGJMP))
> 3303                  && old_stack_allocated != stack_pointer_delta
> 3304                                            - pending_stack_adjust)
> 3305                abort ();
> This appears to be a case of the stage 1 compiler miscompiling the
> stage 2 compiler -- given the same input, the stage 1 compiler does
> not crash.


2004-05-25  Paolo Bonzini  <>

	* combine.c (gen_binary): Remove.
	(known_cond, simplify_shift_const
	find_split_point, combine_simplify_rtx,
	simplify_if_then_else, simplify_set,
	simplify_logical, expand_field_assignment,
	extract_left_shift, force_to_mode,
	if_then_else_cond, apply_distributive_law,
	simplify_and_const_int, simplify_shift_const,
	gen_lowpart_for_combine, simplify_comparison,
	reversed_comparison): Replace with
	simplify_gen_binary, simplify_gen_relational or
	(distribute_and_simplify_rtx): New function.
	* simplify-rtx.c (simplify_binary_operation):
	Use nonzero_bits to simplify ANDs where we are
	turning off bits already known to be off in OP0.

(which additionally has a wrong date).

Now trying to come up with a testcase...

Eric Botcazou

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]