Re: Small shell-mode patch to handle auto-cd (e.g. shells like zsh)

[Jason Kim]

Working on it now,  Expect the git format-patch shortly
But just to set the stage.... a few comments.

On 5/6/21 1:23 AM, Eli Zaretskii wrote:
>   . A few comments on the code itself:
>
> > +(defvar shell-has-auto-cd nil
> > +  "if set to true, shell-mode assumes that a command that resolves to a 
> > directory
> > +is an implicit cd command. Useful for shells like zsh that has this feature.")
>
> This should be a defcustom, so that users could customize it easily.
done.

> The first line of the doc string should be a single complete
> sentence.  Something like this:
>
>    If non-nil, `shell-mode' will automatically process implicit \"cd\" commands.

Yep that's lot better :-) thanks!

> (Then in the following text of the doc string explain what does
> "implicit cd command" mean in this context.)
>
> Should this variable's value be computed automatically based on the
> shell's name?

I don't think so... If only because its an option that needs enabling in 
zsh as well?

> > -               end cmd arg1)
> > +               end cmd arg1 cmdS)
>
> I wonder if you could find a more descriptive name for the cmdS
> variable.

cmdS --> cmd-subst-fn

> michael.albinus@gmx.de writes
>- Several different shells could be used in an Emacs session, depending
>  where you run the shell, locally or on a remote host. So it might be
>  problematic to have just one fixed global value.

I'm confused...?
Would making the variable buffer-local in a user's .emacs address this?

dir tracking (for things like auto completion) never worked for me if I 
ssh into a different host from emacs-shell.
(maybe because pattern globbing rules are more complex (optionally) in zsh?)
What am I missing here?

Can we skip this issue for now? (or does making the var buffer-local get 
around it?)

Thank!!!
-jason