On Sun, Jul 9, 2017 at 9:52 AM, Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> wrote: > > It could have made a difference in the test suite because killing > unreachable code can create new simplification opportunities. Yes. As far as I can tell, when the kill instruction get invoked, it should have cause the REPEAT_CSE. "Kill unreachable bb" shouldn't need to set CSE flag. > What do you mean exactly by 'pass full kernel check'? That you get > exactly the same warnings with & without the patch? Yes, that is what I usually do to give sparse some workload other than the test-suite. > >> From 484a3a27d95b4bf3be9ac4b9bcf1aca1abe3ac19 Mon Sep 17 00:00:00 2001 >> From: Christopher Li <sparse@xxxxxxxxxxx> >> Date: Sat, 8 Jul 2017 19:34:49 -0700 >> Subject: [PATCH] move kill_unreachable_bbs to outer cse stage >> >> The current way of kill_unreach_bbs in insert_branch() >> cause delete entry in ptrlist that the upper level >> caller is looping on. > > I know what you're talking about but otherwise, I really can't parse > this sentence. > > Anyway, I think that this bug merit a much better commit message. > Of course. Can you take my patch and merge it with your validation test? Do what you see fit. I think it'd better commit together, and get a better commit message. I was overwhelming by the ptr ref counting. I need some thing in my branch to stop die on those ep->bbs. > It would also be nice to know why the patch that came with the bug > report have been discarded without a single comment. Sorry I haven't been more specific. I think the original patch is a bit too complicated. Also just as I suspected, there is more than the bb list get affected. If possible, I think we should try to avoid get into nested loop delete in the first place rather than make the nested loop delete work. Chris -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
