bug#72426: 29.2.50; comint-pager doesn't affect async-shell-command

[Spencer Baugh <sbaugh@janestreet.com>]

Jim Porter <jporterbugs@gmail.com> writes:

> On 8/3/2024 8:38 AM, Eli Zaretskii wrote:
>>> From: Spencer Baugh <sbaugh@janestreet.com>
>>> Date: Sat, 3 Aug 2024 06:47:10 -0400
>>> Cc: 72426@debbugs.gnu.org
>>>
>>> comint-terminfo-terminal affects async-shell-command, why not this?
>> Ugh!  A mistake, IMNSHO.  But that ship sailed a long time ago, so
>> we
>> cannot fix the mistake.  We can avoid enlarging the mistake, though.
>> 
>>> If the fact that the variable is in comint is the problem, I can rename it and move it elsewhere.
>> I don't think functions that are almost primitives should pay
>> attention to application-level features such as this one.
>
> Perhaps we should be setting the pager in a similar way to how TERM is
> set in startup.el?
>
>     ;; Subprocesses of Emacs do not have direct access to the terminal, so
>     ;; unless told otherwise they should only assume a dumb terminal.
>     ;; We are careful to do it late (after term-setup-hook), although the
>     ;; new multi-tty code does not use $TERM any more there anyway.
>     (setenv "TERM" "dumb")
>
> I think the reasoning in that comment applies to PAGER as well: unless
> told otherwise, subprocesses probably shouldn't use pager like "less";
> it's very unlikely to work correctly.
>
> In that case, would it make sense to add something along these lines
> to startup.el?
>
>     (when (executable-find "cat")
>       (setenv "PAGER" "cat"))

Yes, I'd be very in favor of that.  Fixing this is exactly my
motivation.

> The 'comint-pager' variable could still be useful though, since you
> can set it to a pager that's fancier than "cat" but that actually
> works in Comint (unlike "less"). For example, you could set a pager
> that does automatic syntax highlighting with ANSI escapes; Comint
> would be able to handle that.

Maybe, but I'd rather delete it in the meantime, since it doesn't serve
its main purpose.