On 3/29/22 02:02, Pokechu22 wrote:
On Thu, Mar 24, 2022 at 1:53 PM Jason Merrill <ja...@redhat.com> wrote:
Thanks! For future reference, the patch doesn't apply easily because
gmail wrapped lines; for sending patches via gmail you'll need to use
attachments. Or you can use another MUA, or git send-email. This time
I fixed the wrapping by hand, but that's a pain (especially since
there's no "try again" flag to git am).
Thanks. I was vaguely aware that gmail did something wrong, but I
thought that enabling plain-text mode was enough to work around it.
It might be useful to add that information to the contributing
webpage. I've attached a copy of the patch here in case anyone else
wants to test it without dealing with the mangling.
Out of curiosity, do you know if this issue could have impacted
anything beyond the false warning? I had previously determined that
ensure_literal_type_for_constexpr_object completes the type, so for
constexpr variables it wouldn't have caused any further issues, but I
discovered that regular const variables were also affected later on
(and possibly other unused variables could be affected, but non-const
ones would trigger the warning normally).
It also affects whether the variable goes into .data or .rodata. For
#include <initializer_list>
const auto x = { 1, 2 };
without the change, x goes into .data; after the change it goes in .rodata.
Also, like other DCO projects, we can't normally accept pseudonymous
contributions. But as you say in the PR, this patch is trivial enough
that I'm content to apply it myself; I want to make some adjustments anyway.
Also out of curiosity, what do you want to adjust with it? I'd like
to know in case it's relevant for any other patch that I submit
(though there probably won't be a situation where I'm writing another
patch for a while).
I'm adding a diagnostic if the deduced type remains incomplete, not
anything more generally relevant.
Lastly, I'd appreciate my pseudonym being credited (at least as a
co-author) in the final patch,
Yes, that was my plan.
Jason
