------- Comment #5 from roger at eyesopen dot com 2007-01-04 22:34 ------- Can you reduce a test case for the loop, now that the code in the PR description is fixed? One thing that looks a bit odd is that in the condition you describe the constant term "1", isn't on the far right, which I believe should be the correct canonical form. Hence it looks like some of the trees/subtrees in your expression haven't been folded. With a small testcase it'll be easier to see whats going on.
I agree that you might be right, and it may be profitable to disable these BIT_NOT_EXPR forms in fold, and identify them during RTL expansion or during the RTL optimizers. But I'd like to see the sequence of events before I admit defeat. Thanks in advance. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30322
