(This happens when you break on a Sel creation and the firm debugger
 tries to tell you about the created node)

- We now always output dwarf3 so we can use DW_OP_call_frame_cfa and avoid
  construction location lists, but just reuse the callframe info lists.
- Backends have to emit debug info as callframe calculation changes:
  The ia32 backend has a preliminary implementation which assumes esp
  offset of frame_type_size at the beginning of a block (currently
  always true), the no-omit-fp mode assumes ebp relative addressing
  (which is correct except for the prolog/epilogue insns)

Remove the stabs debugoutput code and with it all the complexity of a
modular debug output system. The only remaining debug output module is
dwarf.

It's an internal thing and nothing from the dwarf spec

- the "create" command now catches new nodes, entities and types as
  their numbers are unique anyway

DWARF: Actually emit the list of files, not just the first filename over and over again, i.e. use index i, not 0.

this fixes a bug, because match_arguments determines the operations
mode by looking at the passed operands, so they must still have the
original mode and not the mode after a skip_downconv.

Merge a block with its only control flow predecessor,
if it contains no phi nodes and the block is the only successor of the predecessor.

This adds support for C99/gcc style compound types where the last
element is an array of unspecified size. The compound size is the
minimum size (assuming the array has no elements), or is calculated
by the size of an initializer.

ftst is always available and faster than fldz+fucom.

The ftst from the other branch cannot be commutative (it only has 1 operand).