Hi Michael,

On 30/05/2024 11:49, Michael Albinus wrote:

>> Thank you. I haven't tested this personally either, but it seems like
>> this is the behavior we will want by default.
>>
>> So the proposed solution should include it, I think.
> In order to not misunderstand: you agree the solution based on
> tramp-histfile-override now, don't you?

Just to be frank: I don't like the approach, but I do think it's better 
than the status quo. In that sense, I agree with it, if none other is 
accepted.

But, well, in the meantime I came up with a couple of new solutions:

1) [first patch] We can add a new major mode, for 'M-&' to use instead 
of the full-blown 'shell-mode' - it could be very simple: just apply 
font-lock keywords and maybe set list-buffers-directory.

Problems? I suppose someone might be using shell-mode-hook to do 
something in the async-shell-command output buffer, and it won't fire 
anymore. Seemingly very minor concern.

2) [second patch] async-shell-command could set shell--start-prog (it's 
permanent-local) to `none', and then shell-mode will check for that 
value, and if set to this value, skip the input ring setup.

Downsides? Pretty ad-hoc approach. And any external code looking up this 
(private) variable could be surprised by the new value.