[libcxx-commits] [libcxx] [libc++]libcxx constexpr deque (PR #200929)

Hristo Hristov via libcxx-commits libcxx-commits at lists.llvm.org
Sat Aug 29 23:05:45 PDT 2026


Zingam wrote:

> > > What about #129368. Should we close it?
> > 
> > 
> > I believe the approaches used in #129368 for product code is better than those this PR, and #129368 can be considered more completed, except for failing tests. So I want to reduce this PR to mere addition of lightweight tests, and then merge #129368 after this PR.
> 
> @frederick-vs-ja Do you still consider the above as a viable option? I had a quick look at the tests in this PR and I think I don't like what I see. If that's still the (possible) plan. I'd propose to close this PR and copy over the relevant bits to the original PR: #129368
> 
> What I mean we keep the changes to the tests in #129368 e.g.
> 
> Converting everything to the newer pattern:
> 
> ```c++
> constexpr bool test() {
>    ...
>    return true;
> }
> 
> int main(...) {
>    test();
> }
> ```
> 
> but then insert the simplified tests:
> 
> ```c++
> constexpr bool test() {
>    ...
>    return true;
> }
> 
> constexpr bool test_constexpr() {
>    ...
> }
> 
> int main(...) {
>    test();
>    ...
>    test_contexpr();
>    static_assert(test_constexpr());
>    ...
> }
> ```
> 
> I think that's what you propose, right? Alternatively we can create a new PR with just the updates to the tests.
> 
> I'd expect that similar issues could uncover themselves in the dependent `<stack>` and `<queue>`.

Then we have this:

> We don't have to stay within the default Clang limit.


Have you investigated that option?

https://github.com/llvm/llvm-project/pull/200929


More information about the libcxx-commits mailing list