Improve two errors

[wooster0]

This improves the error when you do for example pointerof(123) which can be fixed by putting 123 in a variable. That's needed because you can't just modify and operate on the raw value 123, that doesn't make sense. The value of a variable however can be modified.

And this also makes the stack overflow error slighlty more detailed to make sure there's no confusion. Even after stack overflow detection there was still some confusion about the error: #7032.

[bew]

Also, it surprises me that pointerof(self) (and other pointerof(X) checks) is a parse error instead of a check done later in the processing of the parse tree.

[straight-shoota]

@bew Only pointerof(self) is a parser error. self can't be a variable and it can't be used in pointerof, so its definitely an error at this stage.

[straight-shoota]

Thanks, @r00ster91!

[bcardiff]

Isn't it misleading that the message says

pointerof only works with variables

but pointerof(obj.@ivar) and pointerof(LibFoo.external_var) are allowed?

[wooster0]

The error message doesn't specify what kind of variable. It could be for example an instance variable, a class variable or an external variable. It would be misleading if it would say "local variables".

[asterite]

I agree with @bcardiff . It's misleading because at the syntax level obj.@ivar and LibFoo.external_var don't look like variables, even though they point to variables.

What's wrong with the current error message?

[wooster0]

What's wrong with the current error message?

I don't think there's something wrong with the error. I think at the syntax level obj.@ivar isn't confusing because you can clearly see that it's an instance variable. Same for LibC.variable. If you know that it's a variable, the error makes sense. The error doesn't specify what kind of variable it needs to be, it can be any variable.

[asterite]

People usually read "variable" as "local variable". Ask anyone "what's a variable in Ruby" and they will say "the 'x' in 'x = 1'". They will never say "oh, and variable could also mean instance or class variable".

I think, if it's not broken, we shouldn't fix it.

[wooster0]

Okay but maybe at least turn can't take pointerof(self) into can't take address of self for consistency with the other errors then?

[Sija]

I vote to close this PR.

[wooster0]

@Sija why? I'm trying to do something out of this PR.

[Sija]

@r00ster91 because the actual error message is imo good enough.