Re: [Emacs-diffs] emacs-25 504696d: Etags: yet another improvement in Ruby tags

[Dmitry Gutov]

On 02/04/2016 06:48 AM, Eli Zaretskii wrote:

> > Anyway, the example shouldn't put attr_X calls inside a method
> > definition, or it gives an impression that we handle this situation
> > intentionally, or somehow differently from the usual case. Whereas we
> > could as well skip those tags altogether (but we don't really have to,
> > as long as we only generate non-qualified tags, and check that every
> > argument is a Symbol literal, i.e. it starts with a colon).
>
> Please modify the test files as you see fit, and tell me what the tags
> should be.

I've updated the example and the tags. No further action needed, thanks.

> > > +      alias_method :qux, :tee, attr_accessor :bogus
> >
> > This one is a bit weird as well:
> >
> > - An alias_method call with three arguments will raise an ArgumentError.
>
> It's there to test the algorithm, which should not tag the bogus
> accessor.

Why is it bogus, though? If it were not a syntax error (because of 
alias_method using a paren-less call as well, not because of preceding 
comma), it would be a valid attr_accessor call.

For instance, these examples are syntactically valid and would result in 
the generation of the method 'foo':

class C
  puts(attr_accessor :foo)
end

class C
  1 + 2; attr_reader :bar
end

We don't really need to support them, but actively fighting against them 
seems odd.

> I'm not sure how this simplifies things.  The point was that a comma
> doesn't reset the mini-state machine to the state where it is once
> again ready to see attr_accessor.  If you are saying there are other
> situations like that, please describe them.
>
> IOW, does etags handle the above intentionally invalid code correctly?
> It should.

Point is, the example code is syntactically invalid. etags doesn't need 
to handle that kind of code at all, right? It would be better to get a 
syntactically-valid example, if we an find one. Not sure if we can; 
these are also invalid:

1, 2, attr_accessor :bogus
puts(1, attr_accessor :bogus)

Ok, so it it doesn't simplify things, don't worry about it.