domain: delete broken but unused __has_common_domain

[romintomasetti]

__has_common_domain incorrectly returns true for domain pairs whose common type is indeterminate_domain<...>.

For example:

struct Whatever {};

struct SomeDomain : public STDEXEC::default_domain {};
struct WontBeGoodDomain : public STDEXEC::default_domain, public Whatever {};

// Passes, but shouldn't.
static_assert(STDEXEC::__has_common_domain<SomeDomain, WontBeGoodDomain>);

// Here is the common type.
static_assert(std::same_as<STDEXEC::__common_domain_t<SomeDomain, WontBeGoodDomain>, STDEXEC::indeterminate_domain<WontBeGoodDomain, SomeDomain>>);

It could be fixed by checking that the result is not an instance of indeterminate_domain, but since __has_common_domain is never used, let's get rid of it.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[romintomasetti]

@ericniebler I am not sure I actually understand what common_domain is supposed to do.

In the following, I expected that SomeDomain and SomeOtherDomain would have stdexec::default_domain as common domain. Given that it's using std::common_type under the hood, it obviously cannot understand that SomeDomain and SomeOtherDomain have stdexec::default_domain as a base.

Are you expecting people writing new domains to specialize std::common_type for pairs of domains that make sense?

  struct Whatever {};

  struct SomeDomain : public STDEXEC::default_domain {};
  struct SomeOtherDomain : public STDEXEC::default_domain {};
  struct WontBeGoodDomain : public STDEXEC::default_domain, public Whatever {};

  static_assert(STDEXEC::__has_common_domain<SomeDomain, SomeOtherDomain>);
  static_assert(STDEXEC::__has_common_domain<SomeDomain, WontBeGoodDomain>);
  static_assert(!std::same_as<STDEXEC::__common_domain_t<SomeDomain, SomeOtherDomain>, STDEXEC::default_domain>); // unexpected... ?
  static_assert(std::same_as<STDEXEC::__common_domain_t<SomeDomain, WontBeGoodDomain>, STDEXEC::indeterminate_domain<WontBeGoodDomain, SomeDomain>>);

[romintomasetti]

And just to clarify, it seems that __has_common_domain says yes, no matter what you feed it with:

  struct Whatever {};
  struct ThisIsNotADog {};

  struct SomeDomain : public STDEXEC::default_domain {};
  struct SomeOtherDomain : public STDEXEC::default_domain {};
  struct WontBeGoodDomain : public STDEXEC::default_domain, public Whatever {};

  static_assert(STDEXEC::__has_common_domain<Whatever, SomeOtherDomain>);
  static_assert(STDEXEC::__has_common_domain<Whatever, ThisIsNotADog>);
  static_assert(STDEXEC::__has_common_domain<Whatever, STDEXEC::default_domain>);
  static_assert(STDEXEC::__has_common_domain<SomeDomain, SomeOtherDomain>);
  static_assert(STDEXEC::__has_common_domain<SomeDomain, STDEXEC::default_domain>);
  static_assert(STDEXEC::__has_common_domain<SomeDomain, SomeDomain>);
  static_assert(STDEXEC::__has_common_domain<SomeDomain, WontBeGoodDomain>);
  static_assert(STDEXEC::__has_common_domain<int, double>);
  static_assert(STDEXEC::__has_common_domain<int, int>);

[ericniebler]

yeah, __has_common_domain is vestigial since i added indeterminate_domain. deleting it is the right thing.

struct SomeDomain : public STDEXEC::default_domain {};
struct SomeOtherDomain : public STDEXEC::default_domain {};

static_assert(!std::same_as<STDEXEC::__common_domain_t<SomeDomain, SomeOtherDomain>, STDEXEC::default_domain>); // unexpected... ?

yes, that's not expected. this is what happens when i rush. i suppose it's to be expected that i would make mistakes with my 11th-hr fix for customization.

i'll address that problem in my own pr, and file an LWG issue.

[ericniebler]

/ok to test 969f7e6