On Thu, Jul 10, 2014 at 04:30:13PM +0200, Richard Biener wrote:
> Compromise "hack" below. It simply avoids the transform for
> sources that c_strlen can compute a length of. That "fixes" all
> strlenopt testcase apart from strlenopt-8.c which does
> memcpy (, flag ? "a" : "b"); which then still folds
> during gimplification. I have XFAILed that.
>
> I've tried to comb my way through strlenopt but failed to quickly
> add support for generic stores (it has very rough support for
> char stores, see also PR61773).
>
> Does the below look ok?
I can look at the tree-ssa-strlen.c stuff and removing the !c_strlen
hack incrementally.
Jakub
