Re: Three Flymake backends Was Re: Two issues with the new Flymake

[Dmitry Gutov]

On 11/6/17 1:08 PM, João Távora wrote:

> Then probably a good idea to do what flycheck does because gremlins.

Sure.

> > > Hey I didn't spend more than 5 minutes with it :-)
> >
> > However long, I'm glad you did. Now, the version I've sent is pretty
> > much working.
> >
> > Will you push it to emacs-26?
>
> I think you should test it a tiny bit more and then you push it :-)

It seems to be working well, but I'm getting bogged down by minor details.

Do we add defcustoms for the program name and the config file name? Do 
we really need the ruby-flymake-command defcustom? I doubt there is an 
alternative program that gives the same output as 'ruby -wc'.

> > Where will it live?
>
> ruby-mode.el so it at least locally shares some code with ruby-flymake.

OK, code sharing will be step two.

> > How/when will Flymake choose between rubocop-flymake and ruby-flymake?
>
> Flymake can use both at the same time. Just
>
>     (add-hook 'flymake-diagnostic-functions 'ruby-flymake nil t)
>     (add-hook 'flymake-diagnostic-functions 'rubocop-flymake nil t)
>
> The user can remove-hook if he wants to.
>
> If rubocop does everything "ruby -w" does and more,

It does.

> then maybe a single
> backend. One that uses a ruby-flymake-use-rubocop-if-available
> defcustom, and then checks for (executable-find "rubocop").

Do we check for (executable-find "rubocop") once inside the major mode 
function, or every time the checker is called?

In the latter case we'll have a function ruby-flymake-auto calling one 
or the other.