On 2/5/2023 7:33 AM, Stefan Monnier via Bug reports for GNU Emacs, the 
Swiss army knife of text editors wrote:
> BTW, a low-tech way to get similar results is to change your completion
> table.  Instead of having entries like
> 
>     user-login-name
> 
> make it contains entries like:
> 
>     user-login-name" ...

I'd thought about this, but it gets a little uglier for more-complex 
cases like bug#53371, which wants to complete things like the "foo" in 
"#<buffer foo>".

> Regarding `pcomplete-exit-function`:
> - It's ugly and I don't like it, but given the general design of the
>    current Pcomplete API, I think it's OK.

Yeah, I'd rather do it by returning some structured object, but that 
would probably require some significant incompatible changes to pcomplete...

> - Please add a good docstring to that variable (the fact that the other
>    vars nearby don't have any is no excuse, instead it's a(nother)
>    problem that should be fixed).

Done.

> - Maybe `pcomplete-exit-function` should override the default exit
>    function (i.e. the handing of `pcomplete-termination-string`) rather
>    than being added to it?
>    E.g. we could start with `pcomplete-exit-function` bound to the
>    default (the one that handles `pcomplete-termination-string`) and let
>    the rest of the code modify it either with `setq` (to completely
>    override) or `add-function`.

I like this a lot better than my previous solution. Done.

I also changed the last patch to return a programmed completion function 
that mimics "~user" completion in 'completion-file-name-table'. That 
seems like a more-consistent way to handle that case, since now Eshell 
"~user" completion works the same as completing any file name.