[tree-ssa] Crashes in UNSSA

[tree-ssa] Crashes in UNSSA

[Jan Vroonhof]

Yes, I know it isn't finished.
Yes, I know I am living dangerously by enabling copy propegation
Andrew probably knows this already anyway, but just in case it
saves him some time:

When I actually enable enough optimisation for unssa to be used in
earnest, I see crashes in bsi_commit_first_edge_insert.

It actually is in this fragment

      bsi = bsi_last (src);
      /* If it is an empty block, simply insert after this bsi, and the new stmt
	 will become the only stmt in the block.  */
      if (bsi_end_p (bsi))
        {
	  bsi_insert_after (&bsi, stmt, BSI_NEW_STMT);
	  return bsi;

What seems to be happening is that src is "quasi-empty", it only
contains (a list of) empty_stmt_node's (tree-ssa-dce has eliminated
all the expressions). In this case bsi_last returns a bsi with .pp =
0, which bsi_insert_after does not handle.

At this point I am not quite sure what the correct fix is.

Hope this is useful.

Jan