http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54971
--- Comment #2 from Martin Jambor <jamborm at gcc dot gnu.org> 2012-10-18 17:37:25 UTC --- I already have a work-in-progress patch based on your suggestions that works for the testcase but need to think a bit more about less obvious cases that might happen. However, I have to leave the office now and so will continue tomorrow. I will probably have some questions about what I can put to the debug statements. To answer your question, apart from the change in analyze_access_subtree you described, all code checking the grp_to_be_replaced flag that might be looking at a LHS has to be extended to look at this flag too. But as I said, I already have a WIP patch. Thanks.
