[cfe-dev] Spaceship operator oddity with optional

[Nuno Lopes via cfe-dev]

Thank you!

I will investigate your lead and report the bug to libstdc++. I’ve been hitting quite a few bugs in synthesized <=> operators lately..

 

Thanks,

Nuno

 

From: Arthur O'Dwyer <arthur.j.odwyer at gmail.com> 
Sent: 26 January 2021 16:14
To: Nuno Lopes <nunoplopes at sapo.pt>
Cc: Clang Dev <cfe-dev at lists.llvm.org>
Subject: Re: [cfe-dev] Spaceship operator oddity with optional

 

On Tue, Jan 26, 2021 at 8:58 AM Nuno Lopes via cfe-dev <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> > wrote:


I just bumped into this mysterious behavior with the spaceship operator over
optional:

struct expr {
  std::strong_ordering operator<=>(const expr &rhs) const;
  //bool operator==(const expr &rhs) const;
};

int f() {
  return std::is_eq(std::optional<expr>() <=> std::optional<expr>());
  //return std::optional<expr>() == std::optional<expr>();
}

Function f() returns 0 both on gcc & clang trunk
(https://gcc.godbolt.org/z/fv85eP).
However, if you uncomment the operator==() line, the function starts
returning 1.

The documentation of
https://en.cppreference.com/w/cpp/utility/optional/operator_cmp just says:
    7) If bool(lhs) && bool(rhs) is true returns *x <=> *y
      Otherwise, returns bool(lhs) <=> bool(rhs)

So no word about operator=='s existence.
Is this a bug in both clang & gcc or just some unobvious 3-way comparison
rule I'm not aware?

 

I haven't dug deep, but based on your description it's an issue with GNU libstdc++, not an issue with either of the compilers themselves.

Neither libc++ nor Microsoft STL support C++20 optional::operator<=> yet; libstdc++ is the only library vendor who's tried to implement it, and if their implementation doesn't match your expectations, then my guess is that it's a bug in their implementation. (Or, perhaps, a bug in your expectations — but I give it more than 50/50 it's an implementation bug in libstdc++. The details of operator-spaceship have changed many times in the past 3 years and maybe GNU libstdc++ is implementing some older version.)

 

https://gcc.godbolt.org/z/ofxPss

My guess is that libstdc++ std::optional hasn't caught up to the compiler changes from

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1185r2.html

 

HTH,

Arthur

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20210126/8d9695d7/attachment-0001.html>