On 11/16/2015 04:12 PM, Joseph Myers wrote:
On Fri, 13 Nov 2015, Jason Merrill wrote:
In the future we probably want to move many of these warnings into
language-independent code so we can avoid folding for them in the front end;
we also shouldn't need to fold during genericization, but apparently not doing
that leads to optimization regressions.
We have similar issues for C: various places in the C front end call
c_fully_fold because folded trees are needed for warnings rather than any
language semantics needing folding. (Others call it when a tree is about
to be wrapped in something that c_fully_fold doesn't unwrap, which could
be avoided by handling more C-specific trees at genericization time.)
Right. The plan certainly includes pushing folding back for the C
front-end as well.
Interestingly enough this issue has come up in the context of the kernel
security hardening discussion as well. Presumably the early folding in
the C front-end is causing them to lose information they want to see in
their plugins.
jeff
