On Thu, 19 Nov 2015, Marek Polacek wrote: > On Wed, Nov 18, 2015 at 09:14:46PM +0000, Joseph Myers wrote: > > remove_c_maybe_const_expr doesn't seem to be quite what you want. Apart > > from this not being a case covered by the comment on the function, you're > > ignoring the possibility of the side effects in the > > C_MAYBE_CONST_EXPR_PRE. So I think you want something else: if either > > argument is a C_MAYBE_CONST_EXPR whose C_MAYBE_CONST_EXPR_PRE is non-NULL > > and has side-effects, don't run the comparison, and otherwise it's OK to > > go down to the C_MAYBE_CONST_EXPR_EXPR. > > Indeed, thanks. The following variant does what you suggest. Perhaps > I should've introduced a new helper function, but I couldn't find a > proper name. > > Bootstrapped/regtested on x86_64-linux, ok for trunk?
OK. -- Joseph S. Myers jos...@codesourcery.com
