https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95287
--- Comment #4 from Raul Tambre <raul at tambre dot ee> --- Per Richard Smith's comment here [0], which I think has basically slightly more complicated examples of this same problem, this is valid. While pathological, his explanation makes sense to me. He's also the one that submitted DR2233. I think the key piece here is 7.6.1.3p4 in DR2233: "when a function is called". j<int> can't be called without overriding the first parameter so it's not invalid despite it inheriting a default argument for i and thus requiring also values for the others. Thus [1]: 1. j() calls the parameter pack one 2. j(1) calls the parameter pack one 3. j(1, 2) calls the specialization because it specializes the parameter pack just being a single int 4. j(1, 2, 3) calls the parameter pack one Though this wording is different in the latest expr.call#6 [2] and doesn't mention being ill-formed. But I'm not too great with the legalese. [0] https://reviews.llvm.org/D79800#2058985 [1] https://godbolt.org/z/4bKTojKde [2] https://eel.is/c++draft/expr.call#6
