[intro.execution] make function evaluation actually not interleave CWG2466

[RedBeard0531]

While the intent was clear and explicitly stated in the footnote, the prior wording didn't require B to be consistently sequenced before or after every A in F, only that they be individually pairwise sequenced.

[jensmaurer]

Since sequenced-before is a transitive order, isn't your reformulation just creating the transitive hull, which is (specification-wise) a no-op?

Put differently, can you show an example execution with As and Bs that satisfy the current normative wording, but violates the footnote?

[zygoloid]

Example:

int f() { A1(); A2(); return 0; };
int k = f() + B();

Prior to this change, what prevents A1 < B < A2 (where < is "sequenced before")?

Regardless, these words are subtle, and we shouldn't change them without CWG (and possibly SG1) consultation.

[languagelawyer]

what prevents A1 < B < A2 (where < is "sequenced before")?

I have more crazy question: what prevents A2(); < A1(); or even return 0; < A1();?

[RedBeard0531]

what prevents A2(); < A1(); or even return 0; < A1();?

http://eel.is/c++draft/intro.execution#9

Every value computation and side effect associated with a full-expression is sequenced before every value computation and side effect associated with the next full-expression to be evaluated.

[languagelawyer]

Well, this doesn't say which full-expression is "next". But prolly I'm nagging to much. Continuing this way, we may start asking "where is it specified that the program is parsed from left to right and from top to bottom".

[jensmaurer]

Addressed by CWG2466.

[jensmaurer]

Fixed by CWG2466 in commit da92321.