[tree-ssa] COND_EXPR lowering

[law@redhat.com]

In message <1065632546.2629.182.camel@p4>, Andrew MacLeod writes:
 >I also have some comments/questions:
 >
 >>   
 >> *************** remove_useless_stmts_and_vars (tree *fir
 >> *** 1178,1249 ****
 ><...>
 >> ! 	  else if (TREE_CODE (then_clause) == GOTO_EXPR
 >> ! 	      && TREE_CODE (else_clause) == GOTO_EXPR
 >> ! 	      && (GOTO_DESTINATION (then_clause)
 >> ! 		  == GOTO_DESTINATION (else_clause)))
 >> ! 	    {
 >> ! 	      *stmt_p = then_clause;
 >> ! 	      repeat = 1;
 >> ! 	    }
 >
 >This has ben removed. Is cleanup_cfg() going to catch this and turn it
 >into a straight goto?
Note, I doubt it matters for this particular hunk, but we should keep in
mind that remove_useless_stmts_and_vars is called from more than one
location these days.  It's called after gimplification, but before building
the CFG (to clean up obviously useless crud), then again after we convert
from SSA back to normal form.

Again, I doubt it's particularly important for this case, but it's worth
remembering.



Jeff