[temp.constr.atomic]/2 Meaning of "the same *expression*" is unclear

[CaseyCarter]

[temp.const.atomic]/2 says:

Two atomic constraints are identical if they are formed from the same expression and the targets of the parameter mappings are equivalent according to the rules for expressions described in [temp.over.link].

Not all readers will note that "expression" is italic here, so that it denotes the grammar-term expression, and if they do they may interpret it to mean that "the same expression" means "generated in the same manner from the grammar". Those who don't arrive at this conclusion will likely follow the cross-reference to [temp.over.link] looking for clarification, and will find [temp.over.link]/5:

Two expressions involving template parameters are considered equivalent if two function definitions containing the expressions would satisfy the one-definition rule [basic.def.odr], except that the tokens used to name the template parameters may differ as long as a token used to name a template parameter in one expression is replaced by another token that names the same template parameter in the other expression.

"expressions are equivalent" is pretty close to the phrasing "the same expression", and nothing else in [temp.over.link] tells me what it means for two expressions to be "the same", so paragraph 5 must be why [temp.const.atomic]/2 told me to look at [temp.over.link], right?

[jensmaurer]

So, should we add "(same sequence of tokens)" after expression?

[CaseyCarter]

So, should we add "(same sequence of tokens)" after expression?

The intended meaning is stronger than "same sequence of tokens." Post-P0717, the intent is that two atomic constraints must be generated from the same utterance in order to be equivalent. For example, given:

template<class T> concept C1 = is_same_v<T, int>;
template<class T> concept C2 = is_same_v<T, int>;
void f(C1 auto) {}
void f(C2 auto) {}

the call expression f(42) is ambiguous despite that the two atomic expressions are both spelled is_same_v<T, int> and both have the same parameter mappings. I'm not sure how to phrase "the same utterance" in such a manner that it doesn't suggest that concept definitions in headers are broken when included in multiple TUs.

[jensmaurer]

So, maybe we should say "the same lexical expression", although I'm still unclear how that translates to linkage across translation units.

[jensmaurer]

Editorial meeting: This phrasing is only used for partial ordering, not for redeclarations (which indirectly uses the ODR). We should have a CWG issue to maybe add "... appearing within the same declaration".

[xmh0511]

Similar things are

template<class T>
requires (sizeof(T)>0) void fun(T){}  //#1

template<class T>
requires (sizeof(T)>0) && true void fun(T){}  //#2

when partial ordering #1 and #2, although the atomic constraint (sizeof(T)>0) of #1 is literally identical to that of #2 such that it seems to obey "formed from the same appearance of the same expression" and has the same mapping. However, they're not be considered as identity.