Re: INSIDE_EMACS and Tramp

[Stefan Monnier]

>> In the original situation, IIUC what happens that Eshell sets up
>> INSIDE_EMACS with ",eshell" then creates the process, which gets
>> delegated to Tramp, at which point Tramp would add ",tramp" to
>> INSIDE_EMACS so we'd get what we need.
> This is the scenario indeed. But what happens now, w/o the patch:

w/o which patch?

> - start-file-process (for example) is called.
>
> - Tramp is invoked. It starts a remote shell, passing "
>   "INSIDE=EMACS=28.0.50,tramp:2.5.0-pre".
>
> - The command given by start-file-process is executed, including
>   "INSIDE_EMACS=28.0.50,foo" (for example).

I'm not sure I understand.  Do you mean to say that the code which calls
`start-file-process` runs a command of the form

    /usr/bin/env "INSIDE_EMACS=28.0.50,foo" CMD

[ or something morally equivalent ] ?

If so indeed my approach won't work, but is there a good reason
why the caller would want to do that instead of binding
process-environment around the call to `start-file-process`
(IOW, could we simply say that if the caller does that, they get what
they deserve?).

The problem with your approach is that it means calling `start-process`
from within a Tramp directory could launch a process with INSIDE_EMACS
that contains ",tramp" even tho the process is running locally.


        Stefan