On Tue, Dec 16, 2025 at 10:48:40PM +0700, Jason Merrill wrote:
> > +fail_ret:
>
> Why is this label here? The use below when looking up "data" fails would
> seem to fall through again, as *non_constant_p hasn't been set, and other
> uses check the same condition before the goto. Why don't the current gotos
> just return NULL_TREE?
That is from an earlier version where I've been trying to distinguish
between
if (*jump_target)
return NULL_TREE;
if (*non_constant_p)
return call;
or so. But later on we've changed reflect.cc not to care about that
distinction (i.e. don't dereference result of functions before we check
for (if (*jump_target || *non_constant_p)) and constexpr.cc to do:
if (*jump_target)
return NULL_TREE;
if (*non_constant_p)
return t;
So I guess I should also change other leftovers of this and replace
return call; with return NULL_TREE;
Jakub
