bug#30190: 27.0.50; term run in line mode shows user passwords

[Tino Calancha]

> > @@ -2288,7 +2289,8 @@ term-send-invisible
> >  \\[view-lossage]."
> >    (interactive "P") ; Defeat snooping via C-x esc
> >    (when (not (stringp str))
> > -    (setq str (term-read-noecho "Non-echoed text: " t)))
> > +    (let ((read-hide-char ?*))
> > +      (setq str (read-passwd "Non-echoed text: "))))
> >    (when (not proc)
> >      (setq proc (get-buffer-process (current-buffer))))
> >    (if (not proc) (error "Current buffer has no process")
>
> Why do we need to bind `read-hide-char` here?
I made this binding so that the user observes same characters (?*)
to hide the input as with command:
M-x  term-send-invisble RET

> More specifically, shouldn't `read-passwd` do that for us (hence if it
> doesn't yet, then the right patch is to add this let-binding to
> `read-passwd`)?
I don't think so.  `read-passwd' uses ?. as default.  The docstring 
suggest us to let-bind `read-hide-char' in case we wish another char.
Alternatively we could use ?. always as default, and change 
`term-send-invisble'.
Personaly, I prefer ?* because my vision is quite poor and ?. looks too 
small :-|


> > @@ -2297,6 +2299,17 @@ term-send-invisible
> >      (term-send-string proc str)
> >      (term-send-string proc "\n")))
> >
> > +;; TODO: Maybe combine this with `comint-watch-for-password-prompt'.
>
> Would be nice, yes.
Indeed, one of my favourite Beach Boys songs.

> > @@ -3152,6 +3165,9 @@ term-emulate-terminal
> >           (term-handle-deferred-scroll))
> >
> >         (set-marker (process-mark proc) (point))
> > +        (when (stringp decoded-substring)
> > +          (term-watch-for-password-prompt (prog1 decoded-substring
> > +                                            (setq decoded-substring nil))))
>
> I suggest you add a comment explaining why we set decoded-substring to nil.
Agreed.  It's not obvious at first glance.