[Bug tree-optimization/71881] [4.9/6/7 Regression] ICE on valid code at -O3 with -g enabled on x86_64-linux-gnu: cannot update SSA form

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71881

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
So in theory release_ssa_name()s insert_debug_temp_for_var_def should have
saved
us here.  But!  We have a single debug-use left plus the definition stmt is

  i_21 = i_29 + 1;

and thus i_29 + 1 counts as sth we can use as replacement.

BUT!

This only works if i_29 isn't yet removed which it is because destroy_loop
doesn't care about the BB order in its BB removal process (it has DOM order
but walks forward).

With that fixed the debug stmt ends up as

  <bb 9>:
  b = 0;
  # DEBUG i => 0 + 1

which is correct but also unfolded.

          /* If we didn't replace uses with a debug decl fold the
             resulting expression.  Otherwise we end up with invalid IL.  */
          if (TREE_CODE (value) != DEBUG_EXPR_DECL)
            {
              gimple_stmt_iterator gsi = gsi_for_stmt (stmt);
              fold_stmt_inplace (&gsi);
            }

but fold_stmt_inplace doesn't really do much to debug stmts.