On Fri, 9 Oct 2026, Alejandro Colomar wrote: > Joseph, would it be acceptable to you to have it unspecified?
You could do that *with documentation and tests*. (That includes testing with different values for the two length expressions that the result is one of the values, in various cases with both constant, both variable or one constant and one variable. Also in cases where one gives a length and one uses []. Also in cases where one gives a length and one declares it as a pointer.) In the latter two cases, you'd need to be clear in the documentation about whether it's guaranteed valid, or guaranteed invalid - or unspecified whether it's valid or not in the case where some declarations give a length and some don't, but I don't like that option of being unspecified affecting validity rather than just which length expression is chosen. -- Joseph S. Myers [email protected]
