From: Andres Freund <and...@anarazel.de> Sent: Wednesday, 13 November 2019 6:01 AM
>Peter Smith: > > Is there a reason to not just make StaticAssertExpr and StaticAssertStmt be > the same? I don't want to proliferate variants that users have to understand > if there's no compelling > need. Nor do I think do we really need two different fallback implementation > for static asserts. > > As far as I can tell we should be able to use the prototype based approach in > all the cases where we currently use the "negative bit-field width" approach? I also thought that the new "prototype negative array-dimension" based approach (i.e. StaticAssertDecl) looked like an improvement over the existing "negative bit-field width" approach (i.e. StaticAssertStmt), because it seems to work for more scenarios (e.g. file scope). But I did not refactor existing code to use the new way because I was fearful that there might be some subtle reason why the StaticAssertStmt was deliberately made that way (e.g. as do/while), and last thing I want to do was break working code. > Should then also update > * Otherwise we fall back on a kluge that assumes the compiler will complain > * about a negative width for a struct bit-field. This will not include a > * helpful error message, but it beats not getting an error at all. Kind Regards. Peter Smith --- Fujitsu Australia