I think this is a spec bug. I think it's correct for the for statement to use the universe declaration of int.

It's not an aliasing issue as such. Your argument works just as well for type int int32 as it does for type int = int32.

Your argument works just as well for type int int32 as it does for type int = int32.

Good point, well spotted.

@ianlancetaylor Agreed. Changed title accordingly. That said, it's not clear to me what the best way to address this is in the spec. There's a lot of places where we talk about type 'int' and we always mean universe's 'int' no matter what kind of shadowing declarations of 'int' are active.

Perhaps a paragraph under "Predeclared identifiers" mentioning that whenever the spec names a predeclared identifier, that it refers to the universal declaration?

I've looked at this a bit. Adding a paragraph to the "Predeclared identifiers" section may help if one reads the spec sequentially and keeps that paragraph in mind at all times.

But that's not what people do when they have a specific question. Instead they look for the respective section in the spec. The clarification needs to be in those sections, if at all.

There's many mentions of the type int, and many other types as well (rune, string, etc.), all over the spec. We don't want to sprinkle the text with "predeclared int type" etc. Another option might be to link all these types to the "Predeclared identifiers" section. That will only help if people click on the link.

But more generally, we've now gone for almost 8 years and this has never come up as a problem: Readers tend to think of the predeclared types when we mention int type, string type, etc. It really only leads to confusion if people define their own names that happen to shadow predeclared identifiers and even that it needs a somewhat contrived example (as is the case in the example used for the issue).

We have a more wider issue related to confusing nomenclature. Closing this one in favor of #4532.